5 myths about portfolio construction

Misperceptions are rife when it comes to portfolio optimizers

Optimization has been used as a portfolio management tool for many years, mainly by quantitative managers. Optimizers allow users to combine risk models with return expectations to create efficient portfolios – those that either maximize expected return for a given level of risk or minimize risk for a fixed level of expected return. Optimization requires an objective (e.g. minimize risk) and constraints (e.g. remain fully invested, no more than x% in a sector, and so on). The intended solutions can be relatively simple, or can be far more complicated, taking into account taxes, ESG considerations, liquidity requirements, etc. There is no doubt that they are an indispensable tool within the portfolio manager’s toolkit, but because of the quantitative nature of optimizers, many myths have proliferated around their capabilities and use.  

Here we set out to set the record straight.

 

Myth #1: Non-quants/discretionary fundamental portfolio managers don’t need to use optimization software

They certainly should! Any manager can use an optimizer to ensure a desirable tradeoff between risk and return, regardless of whether their approach is top-down or bottom-up. The challenge for discretionary managers is stating their views in a way the optimizer can use. So, for example, one can’t input a simple “buy” into the optimizer, but a “buy” can be translated into a relative ranking versus all other assets. Ranks can be used to constrain lists of assets to be bought and sold, or they can be processed further to create returns. The user can use rankings to create an expected return relative to the market. Perhaps you expect a 10% market return, and average cross-sectional volatility is 10%. Therefore, the top-ranked stock could translate into an expected return of 20%. This is a simple process that can easily be done within a purpose-built optimizer like the Axioma Portfolio Optimizer.  

A portfolio manager could also incorporate a level of conviction, perhaps by assuming a two-standard-deviation return above the average (i.e. market) return for the highest conviction names. Once the portfolio is created, the manager can run a “sanity check”, to make sure the size of the bets corresponds to their level of conviction.  

Once the expected return (aka alpha) is defined, the optimizer can do a much better job of trading off risk and return than most (if not all) people can do heuristically. Importantly, the optimizer can help the manager avoid unintended and potentially harmful bets. (See Myth #4, below).

 

Myth #2: Portfolio optimizers are only for equity portfolios

Although common use of optimization in the financial world started with equities, optimization is well-suited for other asset classes such as bonds, or indeed for multi-asset-class portfolios. Bond portfolio managers have shied away from optimization because of the far larger number of potential instruments vis a vis equities, as well as a dearth of available risk models. In equities, a typical benchmark and investable universe might include a few hundred to a few thousand securities. Fixed income benchmarks often include tens of thousands of securities. Although fixed income factor models have not been around as long as their equity counterparts, they have become more popular recently and can be used in optimization.  

Currently, the predominant use of optimization in fixed income is for creating index-like, or passive portfolios that closely match the characteristics of an underlying benchmark. One issue bond managers face that is less acute in equities is the lack of liquidity in the bond market. It is not uncommon for many of the constituents of major fixed income benchmarks to be illiquid. An optimizer can easily incorporate liquidity measures including broker axes to determine individual asset selections.

 

Myth #3: You need a PhD to use an optimizer

While one may need a PhD to build an optimizer, most optimization problems in finance are relatively simple to articulate and implement. Many optimization users can state the problem in terms such as “I want to maximize expected return as defined by my alpha forecasts after accounting for transaction costs, while not exceeding a tracking error of 3%, having no more than 5% active weight in any sector or individual stock, and no more than 5% turnover”. It only takes a bit longer to load those parameters into the optimizer than it did to type them out.  

Of course there are many more complex requirements – for example, adding a minimum or maximum number of names, which makes the problem more difficult for the optimizer to solve, but setting up and running the problem remains relatively straightforward. 

There will always be problems that the optimizer can’t solve (i.e., the solution is infeasible because some of the constraints conflict). A purpose-built optimizer will include functionality to help resolve these cases, but there may still be some situations where more explanation is needed. This is where using an optimizer offered by a reputable and knowledgeable firm with optimization experts can come in handy – still no PhD required for the user, just the phone number or email address for customer support!

 

Myth #4: It will not improve my performance so why bother?

We noted above that an optimizer can help a manager avoid unintended bets. The optimizer typically does so via constraints (e.g. “no positive exposure to Downside Risk”). Many constraints are required by the investment mandate, while others are not required but helpful. The main reason for imposing such constraints is to enhance performance by avoiding unintended exposures in the portfolio that may have a negative payoff. Many managers should be able to do this without substantially changing the portfolio. Our paper “Adding Alpha by Subtracting Beta” illustrates how this can work, by using the optimizer to allow small changes in active weights of a fundamental equity portfolio to reduce exposure to harmful factors. The optimizer reallocated the risk budget from unproductive bets to ones that allowed the manager’s stock-picking ability to shine.

 

Myth #5: It’s a black box and I want transparency in my investment process

When portfolio managers add inputs such as expected return, risk forecasts, trading cost parameters, etc. into an optimizer, it will return the portfolio that maximizes their stated objective (e.g., expected return, or risk adjusted return or risk adjusted return minus transaction costs) while satisfying all of their stated constraints. But here is where the black box perception originates: In general, it’s not always clear why the optimizer chose stock A rather than stock B, or why the weight of an asset was 1.3% rather than 1.6%. This does not mean that the process is opaque.  

One does not have to know all the inner workings of an optimizer to trust it. Take for example if you were going under the knife. You do not need a medical degree, but will instead trust the years of training and experience the surgeon has. Similarly, a portfolio manager can trust that the output of an optimizer is the most efficient portfolio possible given the objectives and constraints she provided.  

A good optimizer will in fact provide insight into the core tradeoffs that drove the optimization decision: Which constraints were most limiting on the portfolio? Which piece of the objective (e.g., risk, return, or transaction cost) was most dominant? How much would the objective improve if I were to loosen a constraint (e.g., allow more turnover, or add more names to the portfolio)?  

We would also argue that while a fundamental manager may point to price targets or extensive knowledge about a company so their stock picking seems transparent, often the holdings may not reflect the narrative (i.e. the manager's bias has crept into the portfolio) and the models built to determine those targets remain "proprietary". In other words, full transparency is lacking in all forms of active management.

Portfolio optimization fake news!

To sum up, optimization is a fancy word for a tool that allows managers to create efficient portfolios quickly. Just like most tools, there is a learning curve and a need to use them correctly. But in the same way one would not conclude that hammers are only for carpenters or fancy stoves are only for chefs, optimizers are not only for quant equity managers with PhDs.

 

Learn more about the Axioma Portfolio Optimizer and request a demo below.