Approaches to Portfolio Construction

By now you know that a lot of factors go into portfolio management. From factor exposures, timing, position sizing, and breadth—a manager can add value in lots of different ways and incorporate various statistical and philosophical approaches. So how might you describe portfolio construction?

Exactly. Portfolio construction is part statistics and analytics, part art and part science, and part estimation and part implementation. These competing views impact the choices made in the portfolio construction process, as managers can choose between four quadrants: systematic versus discretionary and bottom up versus top down. Managers can also be described as benchmark aware versus benchmark agnostic, meaning that a manager's approach is implemented with a framework that specifies the level of active risk and active share. Active share is the measure of how similar a portfolio is to its benchmark. Some maintain a very low active share while claiming to actively manage, and so are really "closet indexers." Now, think about the differences between fundamental versus statistical investment approaches—or put another way, systematic versus discretionary. As managers seek to add exposure to rewarded factors, a systematic approach will add balanced factor exposure through research-based rules across a broad spectrum of securities. In contrast, discretionary strategies build greater depth of understanding within their positions through analysis of firm governance, business model, and competitive landscape or will seek higher quality or better timing strategies (even though it's difficult to produce results).

Given this detailed analysis, which approach do you think relies upon the manager's judgment more in constructing the portfolio?

Incorrect. Systematic approaches rely upon more models and data to make decisions.

Yes! Discretionary strategies use the manager's judgment within the portfolio, usually on a smaller subset of the portfolio. Usually, the manager uses financial metrics, along with non-financial variables (like quality of management, competitive landscape, and pricing power of the firm) to study securities, but the manager must then decide how important these metrics are and weight the securities appropriately. This results in more concentrated portfolios that reflect the manager's opinions, thus the portfolio approach for a discretionary manager is a less formal approach that builds a portfolio of attractive securities subject to risk constraints.

Not quite. One of the approaches incorporates more of the manager's opinions into the portfolio.

It's the opposite situation for systematic strategies, which work to limit idiosyncratic risk and use portfolio optimization processes to achieve the desired factor exposure. So which strategy do you think results in the more diversified portfolio?

Right! Systematic strategies are generally more diversified as the optimization process ensures that risk is diversified. Discretionary portfolios, meanwhile, are more concentrated as the manager uses their opinions to build specific exposures. Speaking of concentration, when it comes to bottom up versus top down, both of these approaches can lead to concentrated portfolios. However, the type of concentration will vary.

No. Discretionary managers use more concentrated weights to build portfolios with attractive securities.

Incorrect. One strategy uses a more formal approach to ensure that risk is diversified.

For example, top-down concentrated portfolios will be more heavily weighted in macro factor exposures, and this strategy can also incorporate factor timing. This is especially true for "sector rotators" who move in and out of sectors with concentrated positions attempting to front-run the cycle, essentially. Since top-down strategies are focused at the country, sector, and style levels, these managers will shift the portfolio between favored and unfavored factors. So timing becomes important. Bottom-up managers, on the other hand, focus on security-specific factors like value, growth at a reasonable price, momentum, and quality. These factors can lead to concentrated portfolios depending upon the manager's opinion.

Given these portfolio construction approaches, it's important to keep in mind the risks associated with active management. That's why it's good to know the risk measures of active share and active risk. Although these measures don't always move in sync, active share and active risk will help identify key risks. Active share measures the extent to which the number and sizing positions in a manager's portfolio differ from the benchmark. $$Active~ share = \frac{1}{2} \sum^{n}_{i=1} |Weight_{portfolio,i} - Weight_{benchmark,i}|$$ where $$n$$ represents the total number of securities that are either in the portfolio or the benchmark. You can see that there are two sources of active share. One is holding benchmark securities at different weights than the index. What's the other source of active share?

No.

That's right!

Yet this approach has a significant flaw. What's the flaw in looking back at historical standard deviations?

Exactly. These historical deviations might not be applicable in the future, which makes the realized active risk a flawed measure. That's why predicted active risk is the better measure because it requires a forward-looking estimate of correlations and variances. For this reason, the variance/covariance matrix of returns is critical for the calculation of active risk. As mentioned, active share and active risk can differ, and it's important to understand why. Essentially, active risk is affected by the degree of cross correlation, while active share is not, meaning that active share doesn't take into account the efficiency of diversification. So, a manager can take complete control of active share but can't control active risk due to variables outside of their control.

No. That's not a flaw; that's a benefit of using a historical approach.

Incorrect. The historical standard deviations don't capture the future weightings of the portfolio.

So what variables associated with active risk are outside the manager's control?

Incorrect. Correlations between assets are another important factor that managers must study.

That's not it. Variances between assets are crucial for the manager to understand.

That's correct. Managers can't control correlations and variances, so active risk can't be controlled. The calculation for active risk is: $$\sigma_{R_A} = \sqrt {\sigma^{2} (\sum \beta_{pk} - \beta_{bk}) \times F_{k} + \sigma^{2}_{e}} $$ where $$\sigma_{R_A}~$$ is the active risk of the portfolio, $$\sigma^{2} (\sum \beta_{pk} - \beta_{bk}) \times F_{k}$$ is the variance attributed to factor exposure, and $$\sigma^{2}_{e} $$ is the idiosyncratic risk. Even though active risk can't be controlled, it can be predicted over short time periods. Plus, research has shown that high net exposure to a risk factor will lead to a high level of active risk. If factor risk is fully neutralized, then the active risk is entirely attributed to active share. If the number of securities is large and/or the average idiosyncratic risk is small, then active share will be small, and the level of active risk will rise with an increase in each factor and in idiosyncratic volatility. From these points, you can then classify portfolios as either factor neutral, factor diversified, or factor concentrated. You can also label managers as diversified or concentrated.

Knowing more about these terms, which manager do you think has more idiosyncratic risk?

Incorrect. Diversified mangers have low security concentration and low idiosyncratic risk.

That's right! Concentrated portfolios have high security concentration and high idiosyncratic risks due to the manager's weights to specific factors. So this means that, in general, a multi-factor manager will be a diversified manager with low idiosyncratic risks, while an active manager or concentrated stock picker will have higher active risk and greater idiosyncratic risk. However, the active risk can be reduced by the active manager through adding additional positions to dilute the security concentration.

Incorrect. Factor-neutral portfolios don't take much specific factor risk, so there's limited idiosyncratic risk.

In addition, these labels have also had an impact on the fees charged to investors, as funds with low active shares are essentially index funds. So these funds shouldn't be charging full active management fees when in reality they're actually closet index funds.

The other source of active share is holding securities within the portfolio that aren't in the benchmark. Essentially, if two portfolios are managed against the same benchmark, the portfolio with fewer securities will have a higher level of active share than the highly diversified portfolio. The second risk measure is active risk, which is the more difficult calculation. Although the calculation still involves the benchmark and portfolio weights, there are two ways to calculate active risk. The first is realized active risk, which is the actual, historical standard deviation between the portfolio returns and benchmark returns.

To sum it up: [[summary]]

Part art and part science

Part statistics and part analytics

Part estimation and part implementation

A systematic approach

A discretionary approach

Both equally rely on manager's judgment

Systematic strategy

Discretionary strategy

Equal diversification

Holding securities at the same weight as the benchmark

Holding securities within the portfolio that are not in the benchmark

They might not be applicable in the future

They can show volatility within the strategy

They can mimic future weightings within the portfolio

Variances

Correlations

Correlations and variances

Diversified

Concentrated

Factor neutral

Continue