MPI ties up with Eurekahedge & BarclayHedge

Dynamic factor models power new hedge fund indices

HAMLIN LOVELL

Markov Processes International (MPI) and their clients contend that to stay relevant, indices should be representative and investable. There are reasons to question whether some existing indices meet either or both criteria, and clearly conflicts exist between the two objectives. Most existing indices are all inclusive, equally weighted and comprised of hard closed funds and a long tail of small funds, but it is not possible to access hard closed funds and not practical to maintain miniscule investments in transient cohorts of thousands of tiny funds.

MPI are also of the opinion that selection/non-reporting bias, survivorship bias, and backfill or instant history bias can all serve to artificially inflate index returns, which are often higher for non-investable than for investable hedge fund indices. According to MPI, these biases can be overcome by building a representative index comprised of a selective group of the largest funds.

“There still remains the challenge of making the index investable,” argues MPI’s Executive Vice President and Head of Institutional Solutions, Rohtas Handa. Historically, that is the step that has been missing, he adds. “Some of these indices, were not originally designed to be investable.” He elaborates that “they have often lacked stability with too much turnover, have been equally weighted, and have had over-representation of smaller managers, which is not a good model for how institutions actually invest.”

All of this has been widely understood for many years. The genesis of MPI’s move into hedge fund indices was a reverse enquiry from a client who read MPI’s research piece analysing the returns of the Bridgewater Pure Alpha strategy. MPI work with over 300 asset managers worldwide, including pension funds and sovereign wealth funds, producing research that informs clients’ risk management and analytical processes. “We help clients to analyse investment through the prism of factors so they can understand the drivers of returns,” says Handa.

The Eurekahedge 50 Index  
MPI’s first step into hedge fund indices was to join forces with Eurekahedge to “create cleaner, more stable and more accessible hedge fund indices based on the largest 50 hedge funds,” says Handa. MPI wanted to stick to larger funds with longer track records. Starting with the largest hedge fund managers, MPI selected one fund per firm, based on a set of rules. The rules included five equally-weighted broad strategy buckets: long/short equity, global macro, event-driven, relative value and multi-strategy, to produce a more balanced benchmark. The final 50 have median assets of $1.9 billion and index membership is typically rebalanced annually.

The Eurekahedge 50 Index (Bloomberg: EHFI400) is not itself investable, partly because nine of the chosen funds may be hard closed or soft closed to investment (as of September 2017), but MPI has never had any intention of investing into any of the funds.

Instead, their market and factor exposures are statistically inferred, to create the investable ‘Eurekahedge 50 Tracker’ (Bloomberg: EHFI401). The process is also informed by qualitative analysis of the managers’ stated strategies: “We use our knowledge, expertise and modelling capability to find an optimal fit between the dynamic nature of the funds and their factor exposures. This works better for larger funds because more information is available on them,” states Handa.

The strategy buckets are broken down into nine or 10 clusters of factor exposures, including geographic weightings, to arrive at more granular factors. MPI’s back-test “went back at least 10 years, for more stability, history and to assess results over longer periods,” explains Handa.

The intention is to latch onto relatively broad factor exposures that represent liquid market drivers, and not to try and reverse engineer any alpha from idiosyncratic security- or instrument-specific risk, nor to pick up any illiquidity risk premia. “The liquid market drivers are by far the most important in explaining returns,” claims Handa.

Between November 2014 and September 2017, MPI’s factor-based index – the ‘Eurekahedge 50 Tracker’– has generated a pretty strong fit with the actual performance of the 50 managers – The Eurekahedge 50 Index. Tracking error has been only 1.65%, assuming fees and costs of 1.5% for the index. Argues Handa, “we are identifying the consensus view of the largest 50 hedge funds.” He reiterates the importance of the dynamic approach, which can adjust as new data comes in: “Historically, static approaches have been much slower to react, so their tracking error can grow over time.”

MPI Barclay Elite Systematic Traders Index
The MPI Barclay Elite Systematic Traders Index (MBEST 20) has a similar end goal to the Eurekahedge 50 Index. The objective is to provide a measure of the performance of the largest 20 CTAs, which make up the ‘Elite Systematic Traders Index.’

“To build an investable tracker for the MBEST 20, we had to pick up short-term, medium-term and long-term signals using proprietary factors,” explains Handa. “We spent nine months of research understanding the various strategies,” he says. He acknowledges that the systematic CTA factor index will have “higher tracking error as CTAs are more volatile than other managers. Nonetheless, the in-sample test has done well so far.” He adds “choosing 20 systematic CTAs was determined to provide the index optimal diversification.”

MPI feels particularly comfortable partnering with BarclayHedge, partly as the veteran index provider has taken steps to mitigate biases in indices: its coverage of CTAs is over 90% by number (and a higher percentage by AUM); the BarclayHedge ‘graveyard database’ of nearly 20,000 funds minimises survivor bias, and timestamping minimises backfill bias.

Markov and Handa’s factor heritage
MPI was an early adopter of factor-based analysis and claims to have pioneered dynamic regression analysis for hedge funds. “In the 1990s, MPI was the first to launch commercial software using William Sharpe’s style analysis. This was based on a sophisticated model using return streams to glean insights into the behaviour of strategies,” explains Handa.

For many years, a relatively simple constrained multiple regression had been used for style analysis. MPI founder, Michael Markov, established that traditional regression models did not react fast enough to changing strategies, however. He, therefore, sought to improve responsiveness by patenting a Dynamic Style Analysis technique that could cope with more complex strategies that use leverage and change exposures. “The key is to be very reactive and pick up changes as they happen from one month to the next,” stresses Handa.

Handa was previously Head of Global Sales to Asset Owners and Consultants at index provider FTSE International, where he helped to launch and market smart beta indices, and was earlier also working with risk factors at MSCI Barra. The objective at MPI now is “to build institutional quality benchmarks, that can act as a good gauge for hedge fund performance,” says Handa.  Arguably, the MPI hedge fund indices can also be seen in the context of growing interest in ARP (Alternative Risk Premia) and liquid alternative products, which have seen multiple launches over the past two years.

Investable products
“One bank has already offered structured products based on the Eurekahedge 50, providing proof of concept,” says Handa. MPI is in active discussions with other parties who may commercialise the indices into investable products. The field is wide-open and the architecture is also open. “Asset managers can choose how they implement the factor indices, in terms of physical or synthetic replication, total return swaps, ETFs etc.,” points out Handa. Readers should watch this space for announcements on investable products – and on further launches of factor-based indices from MPI.