Our initial efforts in portfolio construction began with establishing a platform of funds. This platform of approved funds is the result of due diligence efforts combined with deep market insights and experience in selecting candidates from the vast array of funds available to investors. By establishing and monitoring this group of funds, the investment firm provides a set of excellent opportunities for its portfolio managers, whose task is to turn an asset allocation into a portfolio that delivers the appropriate strategy, while providing a reasonable expectation of exceeding the client’s expectations of return and risk.

We propose a “** team of funds**” approach, which is a departure from the traditional method of identifying the ideal fund within each asset segment, and then throwing these “beauty contest” winners together and calling this a portfolio. At best, one could consider this to be an “All-Star” team – but our goal is to create a “Winning” team!

We believe that true portfolio construction is rooted in the fundamental idea that the interaction between funds in the portfolio is as important as their individual performance (relative to their benchmarks.) After all, isn’t this the basis for the asset allocation process that precedes portfolio construction? Just as diversification between asset classes reduces market return volatility, we will show that “*alpha diversification*” substantially reduces active return volatility (i.e., “*tracking error*.”) The result will be a more efficient active process that delivers an excess return with the lowest possible active risk. Taken together, this provides a higher level of confidence in the active process.

** Another innovation of our approach is to customize the fund selection process to the client’s asset allocation. **Many firms identify a favorite set of funds, and then pro-rate these fund exposures across all strategies. This is convenient for the firm, but it is not the best solution for the client. Why not?

First, it seems contradictory for a conservative investment strategy to invest in funds that employ aggressive active strategies. Is a highly-risk-averse client who cannot tolerate high levels of market volatility going to be comfortable with a fund manager who adds high levels of active return volatility? And would a client seeking high returns be happy with a manager who delivers lower levels of active return to avoid active risk? In both cases, we should consider matching the risk aversion of the funds to their market exposures within the portfolio.

Second, our approach of managing the interaction between the funds is partly driven by the exposures dictated by the asset allocation. The allocation limits act as constraints to the fund selection process. These concepts are validated by optimizations that select across the funds for a given strategy. *Our customized approach selects the portfolio’s funds in the context of its asset allocation.*

We will use the same strategy as for our platform evaluation: a diversified “70-30” strategy of stocks and bonds, choosing from among the platform’s 57 funds across 11 asset segments:

** Another unique characteristic of our portfolio construction process is to identify a target excess return.** This aligns with each client’s total return goal, which is the

Our initial portfolio construction focuses on doubling the alpha of the platform, from about 45 bps to about 90 bps. We switched our terminology from “*excess return*” to “*alpha*” - and this is much more than just a word choice.

Excess return is simply the difference in return between a fund and its benchmark. Any fund that earns a return lower than the benchmark is typically considered an “underperformer” – but what if that fund took much less volatility risk than the benchmark? That fund is an “outperformer” – but only after accounting for its lower risk**. Risk matters**, and so we adjust all excess returns to account for differences in risk, and *we refer to “volatility-adjusted excess return” as “alpha.”*

We also calculate a new Information Ratio, calculated as alpha divided by the fund’s tracking error. *Our goal in this first portfolio construction exercise is to generate an alpha of 90 bps at the highest volatility-adjusted information ratio.*

** Another unique characteristic of our portfolio construction process is to identify a target excess return.** This aligns with each client’s total return goal, which is the

With an investment context rooted in client goals, robust risk measures and rigorous diversification theory, the drivers of our portfolio construction process become clear:

- Funds that round out each other’s market exposures
- Funds whose excess returns have low correlations to the other funds in the portfolio
- Funds whose alpha patterns have a low correlation to the portfolio’s overall alpha pattern

The first point recognizes that funds often have long-term exposures that may differ from their benchmarks. This produces a degree of “misfit risk” that is a source of tracking error. One solution is to identify funds with complementary “misfit” characteristics, so that the pair of funds has a higher correlation to their benchmark. This results in higher excess returns with less tracking error risk.

Our second concept is the greatest source of lowering tracking error while retaining alpha. Since active processes are unique to each fund manager, the correlations between excess returns across funds tend to be lower than comparable correlations between asset groups. This means that ** alpha diversification is much more powerful than the diversification found in the asset allocation process** – but only if the portfolio construction process focuses on building a “team” of funds!

We simplify the alpha diversification process by calculating the *correlation of excess return between each fund and the total portfolio excess return*. This is essentially the average correlation between one fund and every other fund in the portfolio. ** This key statistic shows the percentage of a fund’s individual tracking error that remains in the alpha-diversified portfolio. **In some cases, we see funds that deliver alpha while subtracting tracking error. That is an amazing concept!

We selected 21 funds that covered the 11 segments in our asset allocation. The portfolio delivered a 7-1/4% return, outperforming the benchmark return of about 6-1/2%. It delivered this higher return with lower volatility, so that its volatility-adjusted alpha was almost 90 bps. Our goal was to outperform the benchmark on a risk-adjusted basis, and our portfolio accomplished this over its 12-year evaluation period. This is reflected in its higher Sharpe Ratio, indicating that our portfolio was better compensated for its risk.

The portfolio tracking error was 100 bps, which is substantially lower than the 400 bps of tracking error found in the funds used in the portfolio. **Alpha diversification eliminated 75% of the individual active risk of the funds we used.** This changes how we evaluate funds. Instead of rejecting funds with high individual tracking error, we evaluate them in the context of the portfolio, where we retain their alpha, while eliminating most of their individual active risk through active diversification.

The portfolio’s most powerful and persuasive performance statistic is its information ratio (IR.) The portfolio’s IR of 0.76 is about 3.7x greater than the individual funds’ average IR of only 0.21. This high IR implies that this team of funds can deliver at least 32 bps of excess return at a 95% level of confidence.

Any active portfolio can be viewed as a 2-component mix of both market and active exposures. These performance components contribute to both return and risk. Our “*risk aware*” approach to investing requires attribution of both return and risk, and this top-level attribution analysis reflects some impressive (and surprising) results:

The active process contributed 76 bps of excess return (portfolio return minus benchmark return) while **subtracting**
24 bps of total return volatility. Given the high correlation between the portfolio and the benchmark, the benchmark contributes almost all its volatility to the portfolio. However, the tracking error of the funds (100 bps) contributes nothing to the portfolio’s volatility – instead, it lowers the portfolio’s volatility by 24 bps. Clearly, the portfolio’s volatility is not lower than the benchmark due to differences in asset allocation or tactical positioning. Its lower volatility is the result of the effectiveness of our portfolio construction process. *Adding to return while lowering overall risk… priceless!*

__Pop quiz__*:
Q: Why did the portfolio’s alpha decrease the portfolio’s total return volatility?
A: Because its alpha was negatively correlated to the portfolio’s total return.*

Any diversification exercise examines the correlation between EVERY pairing of assets. A portfolio of 21 funds has 210 combinations. We summarized the correlations of these pairings into deciles, where we see the range between +.2 and -.2. These are very low values, indicating a very high degree of alpha diversification (reflected in the elimination of about three-quarters of the funds’ individual tracking error.) The average alpha correlation was zero.

We summarized each fund’s ** “dual-diversification potential” **as shown by its

*Contribution to Risk = Weight * Individual Risk * Correlation to Portfolio*

Statistics such as the return and volatility (or excess return and tracking error) are simply average values measured over the investment performance period. No strategy works all the time, nor does it work consistently over any investment horizon. Therefore, it is useful to monitor the intra-period changes in results to understand the trends in performance. Hopefully, this provides insight into the typical behavior of the strategy.

We applied a 24-month rolling average to the alpha earned by the portfolio over its 12-year horizon. Its trend line indicates a degree of cyclicality in its performance. We also see that the team of funds delivered outperformance for most of the time, with a brief period of underperformance that was a fraction of its more-frequent outperformance. This analysis helps to establish reasonable expectations of the portfolio’s performance, relative to its benchmark.

We reviewed the portfolio’s absolute and relative *performance results in terms of its major decision components*: its market exposure and its active process. Our analysis provided insight into their contributions to return and to risk. Our next steps are to present the results of the asset segments and then evaluate the contributions of the portfolio’s individual funds.

These asset segments correspond to the components of the asset allocation. Their performance results are expressed in both total return and benchmark-relative excess return and alpha. The bond segment illustrates the need to adjust returns for their volatility. Had this been ignored, we would have mistakenly classified high quality (HQ) bonds as the outperformer, while high yield bonds (HY) would have been called the worst underperformer in the portfolio. We see that the opposite was true: on a risk-adjusted basis, high quality subtracted alpha while high yield contributed to portfolio alpha.

We introduce our innovative risk attribution, combining it with our use of volatility-adjusted excess return (alpha.) We see a reasonable correspondence between alpha and tracking error for the 11 segments in our asset allocation.

While gaining insight into the individual characteristics of the asset segments is useful, our greater interest is in* how these segments contribute to the portfolio’s total alpha and tracking error.* This is the heart of the matter*: how do these segments contribute to the portfolio’s “efficiency?” *

We explained in prior articles that ** efficiency is achieved when each asset contributes the same proportion of return as it contributes to risk**. In a perfect portfolio, these amounts are the same, and our efficiency measure

Our 11 asset segments are impressive in their efficiency, with 8 of them lying very close to the efficiency line. We see an outlying outperformer that helps to offset the inefficiency of the pair of downside outliers. As noted earlier, the factors contributing to risk contribution are: a) segment weighting, and b) segment tracking error and c) correlation of segment excess return to portfolio excess return.

We hope to see a strong relationship between risk and return. This can be measured by the square of the familiar correlation statistic; this *“R-Squared” explains the percent of return that is explained by risk.* For our asset segments in isolation, we see that 41% of their alpha is explained by tracking error. But when we consider these segments in the context of the portfolio, (as opposed to the traditional “in isolation” view of the prior exhibit) we see the full value of our team approach to portfolio construction. In the context of portfolio contribution, almost three-quarters of the portfolio’s alpha is explained by its tracking error. *This provides additional confidence in the fund selection process.*

We come now to the fun part of our portfolio evaluation: how each of the 21 funds contributed to the overall alpha and tracking error. At first glance, we see a set of diamonds that are not quite as “tight” as we observed in our chart of segment contributions – with the overall explanatory power between return and risk declining from 73% to 65 percent.

This is what we expect to see as we move from an asset grouping to the individual assets within each group. This is proof of “alpha diversification” between the funds in the portfolio. This is especially true within the funds of each asset segment, where a more-efficient fund helps to support a less-efficient one. In our context of the total portfolio, every fund plays a part in the overall result, and each contributes to the cumulative result. In practical terms, we pair our weaker contributors with complementary contributors to produce an efficient pairing of active contributors.

** Our optimization process selected those funds from our platform that were collectively most suited to delivering the target alpha with the minimum of active risk.** We see that most of the portfolio’s funds cluster closely to the active efficiency line, with the weaker performers balanced out by their stronger counterparts. The result is the greatest efficiency possible, given our strategy, goals, constraints, and our platform.

Finally, we examine each of our 21 funds in terms of their efficiency, as measured by their *%*** contribution to alpha minus % contribution to tracking error**. Our hope is that most of these funds have results close to zero, where they contribute an equal proportion of the portfolio’s active return and risk. We also hope that for a small set of offsetting outliers, where inefficiency is balanced out by “super-efficiency.” This is our expectation, given our efforts at “alpha diversification.”

Our results did not disappoint. Our greatest source of inefficiency is our large cap growth segment. Fortunately, our portfolio construction process paired these underperforming funds with a large value fund that more than offset the large growth inefficiency. Together, they form an efficient large cap segment. The mid cap value funds offset each other, producing an efficient segment. However, the mid growth funds were inefficient, making the overall segment a source of inefficiency in the portfolio. Fortunately, small cap and non-US investments provided substantial active efficiency, resulting in a super-efficient total equity asset class.

*This hierarchical evaluation of portfolio efficiency is possible because we have a robust risk attribution process drives our team approach to portfolio construction. *

With a goal of creating superior multi-asset, asset owner portfolios, the investment industry has an opportunity to advance its active process substantially - but this requires a change in direction: redirecting portfolio construction from “*running a beauty contest*” to “*creating a winning team*” of funds.

The focus must move from evaluating individual funds in isolation, to creating a portfolio of funds, using a robust portfolio construction process. We began this process by creating a credible platform of funds through a strong due diligence and monitoring process. Our next step was to select our funds in the context of their contribution to the overall portfolio. This process integrated the asset allocation and investment selection processes, creating a more holistic and cohesive approach.

A robust risk attribution process is the key to our success. Once established, this provides insights into the key to active efficiency: equalizing the contributions to active return and risk across all the portfolio’s assets. Our initial results were quite encouraging, and our subsequent articles will apply our portfolio construction process to more challenging alpha goals and more broadly-based strategies.

.

.

.

*Written in Partnership with Stephen Campisi*