The Big Interview: Feifei Li and Joseph Shim at Research Affiliates

A recent addition to the debate surrounding the implementation of multi-factor strategies came from analysts Feifei Li and Joseph Shim at Research Affiliates. Looking at the issue of the market impact and what it does for investment returns, the pair come up with a multi-factor strategy that is aimed at limiting such costs. ETF Stream caught up with Li and Joseph to ask them to explain their thinking and to ask them how their research has been received by Research Affiliates’ client base. But we start with the basics of factor definitions and how this forms a solid base of all their subsequent calculations.

ETF Stream: We have covered something of the debate around the proliferation of factors on the part of providers, and the shifting definitions. With this research you stick very much to the academic literature. Is it vital that the factor definitions are rock solid before you start thinking about transaction affects?

Feifei Li and Joseph Shim: You are right – we think sticking to academic literature to identify factors that are backed by strong economic intuition is important. The academic scrutiny helps to build the ex-ante belief that those return drivers (either as compensation for undiversifiable risk, or due to investors die-hard behavioural bias) are persistent in the future, instead of being results of data-mining. In this particular article ‘Strike the right balance in Multi-factor Strategy Design’, we take the factor selection as given, and only emphasizes on portfolio construction perspectives.

ETF Stream: How well understood do you think is the concept of market impact?

FL & JS: The concept of market impact is poorly understood among the investing public because it is not directly observable (because the index tracking portfolio’s value changes simultaneously with the index’s). Explicit items such as management fees and commissions are often the only costs that an investor consider when he or she makes an investment decision, and implicit costs such as market impact are easily overlooked. However, market impact can take a much heavier toll on investment returns especially when the size of the investment is substantially large — expense ratio is like a tip of an iceberg relative to the hidden market impact cost for many smart beta strategies. Also, even though it is hard to measure market impact costs of a strategy accurately, it is fairly easy to compare and understand the implementation advantages of one strategy over another. It is persistent until the strategy changes its methodology — very different from performance numbers which often is worse than useless (negative in predicting future performance).

ETF Stream: When it comes to portfolio concentration in each individual factor, can you explain what the dangers are of over-fitting? 

FL & JS: If we only focus on empirical evidence, the optimal cut-off level may differ slightly across individual factor portfolios and across different sample periods. For instance, profitability factor may perform the best at 20% cut-off, while momentum factor has the highest risk-adjusted return at 30% cut-off in the sample period. However, the results are dependent upon your sample, and there is no guarantee that the same cut-off will be the optimal one in the future. In other words, it is also possible that profitability and momentum factors may perform the best at 30% and 20% cut-offs respectively for the following years. What is important is to understand the trade-off between including fewer securities with strong signal strength vs. the higher transaction cost due to holding a concentrated portfolio. So we recommend a narrow range, and consistent coverage across all factors unless there is a strong economic rationale for the individual factor portfolios to be constructed differently.

ETF Stream: Are your findings regarding the inclusion of momentum and size factors quite surprising?

FL & JS: We think so, to some extent. Most investors focus on factor portfolios’ stand-alone features to decide inclusion or not. We are here emphasizing on how those factors compliment your overall holdings. Without considering implementation issues, given the significant diversification benefit momentum and size factors bring to the table (negative or low positive correlation with other factors), including them in a multifactor strategy seems a no-brainer. It is a bit surprising that the inclusion of momentum and size factors does not increase the trading cost of a multi-factor strategy. It is primarily due to the trades initiated by uncorrelated factors cancel each other out at the headline multi-factor portfolio level. This observation appears to be informative regarding the decision an investor pursuing a multifactor strategy must make: To capture the desired factor premium effectively, there seems to be benefit of hiring a single manager who cover all the factors from a trading perspective.

ETF Stream: Have you seen any evidence of take-up of your strategy among ETF providers? What criticisms do you have of the offerings that are currently available?

FL & JS: PIMCO is the ETF provider for RAFI Dynamic Multi-factor index. Some of the offerings use 50% cut-off to select the constituents for individual factors or just tilt the portfolio weights within the entire universe without any selection. These examples focus on low-cost implementation, while missing opportunities for better performance. Also, many capture size by focusing on cap-weight small cap part of the universe which is subject to the criticism that size premium has vanished after correction of delisting return reporting in data bases and also post-publication crowding. Some ETFs also use macro regimes to do factor timing. We think macro regime is hard to be forecasted with the precision level that can be useful for timing style factors. And it does add to the risk of over-fitting, adds complexity, and requires more turnover and trading thus higher implementation costs. We always prefer simplicity in our smart-beta strategy design. We also favour smart-beta multi-factor offerings based on mixing approach (combing single factor portfolios together). The integrating approach (which uses the composite scores based on multiple characteristics to select the stocks ) can be a good candidate for a quant active strategy when high active risk is allowed, capacity is not the primary concern, and sophisticated implementation can mitigate trading cost concerns. However, Investors who value full transparency, diversification, minimal governance oversight, and low fees, should find a mixing multi-factor index strategy a sensible choice.

ETF Stream: Broadly, what reaction do you get when people comment on your findings?

FL & JS: Our client facing team has received a number of inquiries related to this subject over the past couple of years — it is part of the motives for us to summarize our findings in product design process and make it publicly available. Clients/Prospects see what is on the table, but they want to know the analysis leads to those decisions. Our article ‘Strike the Right Balance in Multi-Factor Strategy Design’ answers the questions and explains the conscious and deliberate decisions that we made in our product design: to identify the most advantageous balance between effectively harvesting the factor premium and implementation cost. The feedback we receive from our clients and prospects is mostly positive and they find the analysis helpful.

Leave a Reply

Your email address will not be published - please refer to our privacy policy to see how your details are handled.