Returns come mainly from a mix of dynamic and tactical style and factor exposures and single stock alpha – but not beta. Rolling three-month equity market correlations, using daily data, have been very close to zero over the life of the strategy and this is very much by design: “The portfolio is rebalanced to a zero-forecast beta, and dollar neutrality, on a daily basis,” says Alentorn. Gross exposure around 200% is less leveraged than many market neutral or multi-strategy hedge funds and net returns since inception work out at an average low single digit alpha percentage per turn of leverage.

Recently, the strategy has generated profits in all four quarters of 2022, from different styles and factors. “For example, the first quarter captured profits from value, and later in the year quality came to the fore,” says Alentorn.

Dynamically reweighting factors

Nimble and timely rebalancing was needed to navigate 2021 and 2022, which saw a lot of very extreme factor rotations in a short space of time. “Historically, such rotations might have only occurred once a year. As cycles have become shorter, alpha generation needs to embrace cyclicality by deploying different styles and we expect this will continue as central banks raise interest rates and try to avoid recession, which will create more dispersion,” says Alentorn. 

Dynamic factor reweighting is driven by various signals that are sensitive to different dimensions of the changing market climate. “The models identified a regime change when value performed well during a down market in 2022. This was unusual,” points out Alentorn. Rebalancing is also calibrated to the volatility regime because drawdowns in the momentum models have historically occurred during periods of higher volatility. This lesson has informed changes that helped to protect recent performance: “We have avoided the momentum factor drawdown of the last few quarters, thanks to dynamic factor weighting,” reveals Alentorn. In addition to heeding volatility, reweighting is based on a current assessment of risk appetite: “We were cautiously positioned for high uncertainty, pessimism and dispersion in March 2023,” he adds. 

The factor rebalancing has become bolder over time and the strategy will sometimes cut one or more factor families down to nearly zero exposure. “The weighting range for each of the five factors is between 0 and 40%. Value has fallen as low as 2% while management quality has risen as high as 28%,” says Alentorn. 

Sometimes two factors can be simultaneously reduced. GEAR’s dynamic value model blends value with quality metrics around financial strength, yet there are times when neither value nor quality is driving markets. “Whereas historically either value or quality would do well, the 2019 style “dash for trash” or “growth bubble” led both value and quality to underperform. Therefore, GEAR now has the freedom to reduce both value and quality weights,” says Alentorn.

Proprietary and orthogonal factor families 

GEAR is broadly seeking longs with benign, and shorts with malign, value, growth, momentum and quality characteristics, all of which are ultimately based on well documented behavioural biases. GEAR is distinguished by how these factors are defined, measured and combined.

GEAR’s five broad factor titles – dynamic valuation, sustainable growth, strong management, sentiment and market dynamics – include some generic names but are far from generic style premia. “We have spent 18 years developing proprietary ways of capturing anomalies and styles, and performance is very different from more simplistic styles. External academic advisers help to define and refine the factors,” says Alentorn. 

“Sustainable” growth really means durable or persistent growth rather than any ESG related concept, based on a long historical lookback of growing revenues and earnings. 

Sentiment signals are shorter term measures of over- and under-reaction, in contrast to longer term momentum, which is anyway not pure trend for GEAR. 

Momentum is one half of market dynamics, which also includes mean reversion. Over a multi-year period, exposure is evenly split between momentum and mean reversion, but the mix can change a lot at different points of the cycle. 

“In 2022, momentum was down-weighted and long term mean reversion became a much bigger weight,” observes Alentorn.

He views company management as another form of quality going beyond balance sheet quality. The models monitor historical management decisions to gauge whether they were more aggressive and reckless or more conservative. Assessing the quality of capital allocation entails a mix of forward and backward-looking analysis.

These five families are conceptually independent and over time they have been deliberately constrained and neutralised to avoid correlation. “Each family can be used as an independent prism to view the investment universe of 7,000 or so stocks,” says Alentorn. 

Positive factor exposures

All five factor families have positive expected returns. “We would not attempt to time factors that have a negative expected return on average, because very high timing accuracy would be needed for something like low volatility, which is a very volatile and asymmetric factor that is also difficult to implement on a long/short basis,” explains Alentorn.

Factor evolution and 2019

Factors and processes are constantly evolving with ongoing projects, and sometimes negative performance can also prompt rethinking and innovation.

Returns of -11% in 2019 marked the worst ever year for the strategy and were in fact the only significant negative year over the fund’s 14-year track record. The post mortem exercise, which involved the in-house team and external academics, examined why models underperformed and the findings contributed to various changes. 

The portfolio optimization process has been enhanced since 2019, to place more emphasis on the higher moments of the expected return distribution, such as skewness and kurtosis as well as left tail and downside risk: “Factor weights are optimized versus expected Sortino ratios or per unit of downside risk,” says Alentorn.

The drawdown was mainly caused by being short of expensive, unprofitable tech in the US and overlooking flows was identified as a key culprit of losses. “2019 was a market driven mostly by retail flows in daily dealing instruments, including ETFs and active funds, inflating valuations and not by fundamental metrics such as price,” says Alentorn. This insight led to a new data project on flows, globally tracking synchronized buys and sells of stocks and sectors. “The flow signal identified selling pressure in unprofitable tech companies at the start of 2022. This then snowballed into synchronized selling over days and weeks as pressure built into capitulation,” says Alentorn. Thus, the sentiment factor family has been enhanced with flows, enriching the factor family rather than replacing it or creating a new one.

A new project on transitory risk factors, picked up by the principal component analysis (PCA) models, has been useful in complementing more traditional factors by identifying exogenous factors driving individual stocks. This has augmented the sustainable growth model: post the Covid growth bubble, GEAR has profited from shorting names that saw a one-off boost to growth from stay at home or Covid trends. 

Evolution aided by Jupiter and academics

Meanwhile, natural language processing analysis of earnings transcripts from quarterly earnings calls is another addition to the company management module, which has also been enhanced with director deals. “A great deal of granular detail on director deals, which include country by country information on option expires, taxes and so on is analysed to identify meaningful sales. The director deals signals have contributed to shorts in Silicon Valley Bank where the CFO was selling for a number of months,” says Alentorn.

ESG at a reasonable price

The company management factor also includes ESG ratings, which are becoming more granular and informing both long and short books. The ESG signals introduced in 2020 have been additive to performance. “The approach could be summed up as “ESG at a reasonable price” because applying a pure ESG approach to a market neutral strategy can lead to shorting value and being long technology, which is a structural bet with no alpha expectation,” points out Alentorn. Firms with deteriorating ESG scores can become shorts while improving ESG ratings are one input for longs. 

The strategy is migrating to disclosures under SFDR 8 and will be excluding sectors including UN Global Compact violators, tobacco and controversial weapons. There is however some asymmetry between exclusions on longs and shorts. “Heavy polluters are excluded from the long book, but there is some economic rationale for shorting them, though we are of course aware of the debate around providing liquidity to such firms,” says Alentorn.

Internal and external resources for innovation

The innovation has been expedited by both longstanding academic relationships and more recent synergies since joining Jupiter. “Sitting on Jupiter’s investment floor gives us access to all asset classes and geographies, from global value to growth, fixed income and emerging markets in what is a collaborative environment,” says Alentorn. GEAR also leverages Jupiter’s considerable data resources, monitoring 40 million data points per day, including accounting and alternative data: “Jupiter’s data science team are very helpful in accessing more complicated datasets such as flows, and earnings transcripts,” he adds.

Academic advisors at Cambridge University, Cass Business School, Imperial College, London School of Economics and King’s College London are mainly retained on an exclusive basis and some of these relationships date back 18 years to the start of the GEAR strategy. “They inform the team’s conceptual understanding of how markets and academic thinking are evolving in areas such as liquidity and transaction costs,” says Alentorn. 

One of the recent additions to the team of external academics, a Professor of Econophysics at King’s College, Dr. Tiziana Di Matteo, now works very closely with the team. “She uses techniques from physics to help understand how markets behave in areas such as clustering and co-movement,” says Alentorn. Meanwhile, Jupiter is also nurturing the next generation by sponsoring and supervising two full time PhDs. 

GEAR’s former head of Systematic Equities, Ian Heslop, retains input on a panel of senior advisers, distinct from the academic advisors.

Liquidity, capacity and launches

Nearly all the investment universe has some sell side analyst coverage, though a few smaller and less liquid names might not do. However, many small and microcaps are ruled out by liquidity criteria, because the portfolio trades about 2% of its holdings per day and holds positions for an average of 3 months. 

Assets are currently USD 6 billion of which USD 4.5 billion is long only, grouped into geographic mandates for historical reasons, and USD 1.5 billion is in the global market neutral strategy. There is plenty of scalability: peak assets were USD 16 billion, and Alentorn does not believe that assets grew beyond optimal capacity. 

Investors should watch this space for some innovative launches, as the strategy’s alpha engines should soon be applied into a roll out of thematic long only funds.