Introducing a portfolio analysis and construction philosophy, consistent with the academically validated principles of factor investing.

The Hunt for a Suitable Equity Portfolio

Portfolio construction is a powerful financial technology that investors can leverage to access more efficient investment opportunities in the face of uncertainty and to implement their financial and extra-financial preferences. Efficiency is largely obtained thanks to the benefits of diversification, allegedly the only free lunch in finance. However, as useful as diversification may be, it alone is not sufficient to construct a portfolio because some sources of common risk cannot be fully diversified away, as evidenced in Figure 1 for a hypothetical equity portfolio.

The level of the plateau reached in Figure 1 is, by design, the portion of equity risk that could not be entirely eliminated through diversification, also called systematic risk (as opposed to the idiosyncratic risk that is fully diversified away). It is equal to the average covariance of returns, across every stock pair in the portfolio. Understanding the source and structure of covariance (or correlation) between stocks is therefore a prerequisite to identifying and quantifying the systematic risks of equity portfolios. Additionally, investors searching for efficient investment opportunities need to also identify robust sources of long-term performance to inform their portfolio selection and allocation processes. Absent any market anomaly, the robust sources of performance should be found precisely amongst the aforementioned systematic risks, because investors require a premium for bearing those risks that cannot be fully diversified away and that hurt a portfolio mostly in bad times.

One might be tempted to select a diversified portfolio based on simple historical measures of risk and return such as past volatility, performance or Sharpe ratio. However, these metrics are heavily sample dependent and may not persist in the future (out of sample), so a thorough analysis of the drivers of risk and return is required to form an educated view on the level of efficiency of a portfolio.

Fortunately, a large body of financial research on asset pricing and factor investing is available to guide investors in their search for efficient portfolios. However, one may rightly ask: why use factors to analyze and construct a portfolio rather than simply look at the stock-level allocation?

Why Do Equity Investors Need to Consider Factors?

The reason is that diversified equity portfolios are often made up of hundreds of individual securities, which makes it difficult to understand what drives risk and return. A traditional “asset allocation” view can hide risk in portfolios because seemingly different securities may have similar behaviors. This means they may generate gains or losses at the same time, thus creating sizeable gains or losses at the portfolio level. For example, it is commonly accepted that stocks in the same sector often have comparable risk and return profiles. Similarly, practitioners and academics know that small caps, or cheap “value” stocks, or “quality” stocks, regardless of their sector, do share common risk and return features. In other words, equity securities exhibit all sorts of correlations, and these correlations are explained by a limited number of underlying factors. By viewing portfolio risk through the lens of risk factors, investors get a synthesized dashboard of their portfolio and a better understanding of what is driving risk and return. This leads to more informed and accurate investment decisions.

Going back to the concepts developed in the previous section, remember that systematic risk is the primary source of performance of a diversified portfolio, and it is directly dependent on the level of correlation prevailing between the stocks in the portfolio. Put another way, the risk your portfolio is still left with even after diversification is… driven by factors! This is why factors are a natural ingredient of efficient equity portfolio construction. An image often used (by BlackRock’s head of factor investing is that factors are the foundation of investing, just as nutrients are the foundations of the food we eat, and that putting together a balanced diet means understanding what nutrients are contained in our food.

The Scientific Portfolio Risk Factors

In line with academic literature, we consider factors as the optimal means to clearly visualize the drivers of both risk and potential long-term performance in an equity portfolio. Every systematic risk factor is characterized by its ability to explain the common variations in stock returns, but academia distinguishes between those that are expected to generate excess returns in the long run and those that are not known to attract a risk premium.

The first category is often referred to as fundamental risk factors or rewarded risk factors and they are backed by an extensive body of research that recognizes them as persistent drivers of expected returns (i.e., long term performance). Indeed, these risk factors represent sources of common risk that cannot be fully diversified away and that are compensated by a risk premium. They have been subject to a high degree of academic scrutiny and challenge, and their economic rationale has been extensively documented. As a result, investors can avoid the trap of data mining and construct portfolios that are likely to remain robust out of sample. The seven generally accepted fundamental risk factors include Market, Value, Size, Low Risk, Investment, Profitability and Momentum.

The second category is unrewarded risk factors, because while they are deemed sources of common risk and contribute to the risk and short-term performance of a portfolio, they are not considered by the finance literature to contribute to the expected excess returns and long-term performance of the portfolio. Sector/industry factors typically fall into this second category of factors that remain important for risk management purposes. The third and final contribution to portfolio volatility is the collection of stock-specific idiosyncratic risks which are considered fully diversifiable and are naturally unrewarded.

In the absence of a particular view on the market, and provided an investor is willing to accept the risk of deviating from the broad-based cap-weighted strategy, our investment philosophy is that over the long term, well-constructed systematic investment portfolios are those that are mostly exposed to rewarded risks and that have managed to reduce most of their idiosyncratic risk thanks to diversification. To assess the extent to which a portfolio is exposed to fundamental (and thus rewarded) risk factors and whether such exposures are spread out appropriately, we use a measure called Factor Quality. Therefore, a balanced exposure to several fundamental factors results in a high Factor Quality score and enhances the potential for long term risk-adjusted returns because factors are not perfectly correlated. Indeed, while fundamental factors are expected to individually outperform on average, their outperformance does not generally occur at the same time.

When modifying or constructing a portfolio to meet a specific objective, it is therefore important to keep in mind the principle of maintaining well-diversified, rewarded factor exposures. This can be particularly relevant for investors wishing to also consider a sustainable objective while continuing to meet financial constraints related to their fiduciary responsibility. How should investors put the principles of factor investing into practice? The tool commonly used by the investment management industry to implement these principles is a risk model.

The Scientific Portfolio Risk Model

The ideal risk model should be comprehensive and largely explain the observed risk and return of a portfolio, yet parsimonious (in its design) and, finally, actionable.

Scientific Portfolio recognizes that building a risk model is perhaps as much of an art as it is a science. There are three common types of factor models used by academics and industry practitioners: i) statistical models, where both factors and factor exposures, i.e., betas, are assumed unobservable and need to be estimated, ii) fundamental time series models which are popular with academics and where factors are assumed observable but not betas, and iii) fundamental cross-sectional models, used by popular risk engines in the industry and where betas are assumed observable but not factors. Exhibit 1 provides a more detailed description and a comparison of the strengths and weaknesses of the three approaches for designing a risk model.

Having reviewed the various approaches to factor modeling, we have designed our risk model to have the intuitive appeal of a time series approach, the flexibility of a cross-sectional approach, and the parsimony of a statistical approach, along with a very strong ability to explain risk. We first scan a portfolio through the intuitive lens of a fundamental time series model, using the seven fundamental time series factors and then, separately, ten unrewarded sector-based risk factors3. The seventeen estimated betas (factor exposures) represent a full risk identification card that is fed into the cross-sectional engine, which in turn, appropriately disentangles the informational overlaps between fundamental betas and sector betas. However, the cross-sectional disentangling does not address the fact that there are arguably too many betas in the risk identification card, and we are likely over-specifying the model. That is why our cross-sectional engine also includes a critical feature designed to avoid over specification (in many ways similar to what Principal Component Analysis (PCA)-based statistical models do) and reduce (behind the scenes) the dimensionality of the model. This ensures there is no “double counting” and makes the model results more robust. The main output is an exhaustive decomposition of the systematic risk of the portfolio across seventeen risk contributions, each associated with an intuitive and meaningful risk dimension.

The overall architecture of our model is therefore based on a combination of standard approaches in equity risk modeling, the technical details of which are available on our platform4. The cross-sectional architecture works in tandem with the time series framework to provide transparency and flexibility without sacrificing robustness. This is achieved by reducing the dimensionality of the model while disentangling the informational overlaps. This allows our model to combine a high level of actionability (based on economically meaningful intuitive factors) and a high level of explanatory power.

APPENDIX: Construction of the Scientific Portfolio risk factors

Fundamental (or style) risk factors
The figure below provides a summary of the characteristics used to build each of the Scientific Portfolio fundamental (or style) risk factors, also referred to as rewarded risk factors in this document. The market factor is simply a market cap-weighted portfolio of the relevant investment zone (e.g., for the US it would be the cap-weighted portfolio comprising the 500 largest US stocks).

While the market factor is cap-weighted, stocks in the long and short leg of each regressor are equally weighted. Each regressor (with the exception of the market factor) is constructed by longing the 20% of stocks with the highest performance expectations with respect to a given characteristics (overperformers) and shorting the 20% of stocks with the lowest performance expectation (underperformer). For instance, the long leg of the “Size” factor contains the 20% smallest companies with equal weights. The relative weight of each leg in the final portfolio is calculated so as to cancel the exposure of the factor to the market factor. At the end of each quarter, the market exposure is computed for both the long and short leg using daily returns. The relative weight of each leg is then adjusted as follows so as to neutralize the market exposure of the factor:

Sector-based risk factors

We use the TRBC classification standard of economic sectors to construct our sector-based risk factors. Each of our ten sector-based risk factors corresponds to the performance track record of a portfolio that is long the sector’s cap-weighted index and short the market index in such a way that the resulting time series has no exposure to the market.