The following is the first installment of a two-part article reviewing alternative data adoption, perceptions and best practices from Vinesh Jha, CEO of ExtractAlpha, a quantitative independent research firm.
In August 2007, many quantitative equity market neutral managers suffered large losses, which we now know were due to their common exposure to the same alpha factors; when one large manager liquidated their books, others suffered, and liquidated in turn, causing widespread downward pressure on these factors.
In the ensuing decade, quant fund returns were generally good, resulting in inflows and a proliferation of data-driven products including Smart Beta or Risk Premia products. As the costs of data storage and computational power have come down, the popularity of data science disciplines have increased, and more data has become available, one would think that quant managers would spend much of their time looking for alpha in alternative data sets.
Certainly stories in mainstream publications such as the Wall Street Journal would suggest that the hype in alternative data is real, and that most quant managers have taken active steps to protect themselves against another similar crisis – which could be much worse the next time around, due to the crowdedness of the quant space today. Strangely, with some notable exceptions such as WorldQuant and Two Sigma, they barely seem to have done so at all.
Crossing the Chasm
Geoffrey Moore’s classic text Crossing the Chasm details the life cycle of a product from perspective of innovative technology vendors – noting that the toughest part of the adoption cycle is moving from visionary “early adopters” to the more pragmatic “early mainstream,” who are more risk averse in their adoption of new technologies.
The concept is well known among tech startups but hasn’t been widely thought about in the institutional investment landscape – but it applies equally well. It is clear to those of us in the alternative data space that right now we’re at the early stage of adoption, but perhaps at the tail end of the early stage – at the edge of the chasm.
A recent Greenwich Associates survey notes that 80% of buy side respondents would like to adopt alternative data as part of their process. In our experience, relatively few have made more than nominal progress. Most quant managers still rely on the same factors they always have, though they may trade them with more attention to risk, crowding, and liquidity.
News stories about alternative data can be misleading. What proportion of the returns of funds run by a multi trillion dollar AUM manager are really driven by advanced machine learning techniques – and not just used as a marketing gimmick? How many truly AI-based funds are there – enough to know whether such techniques lead to outperformance? How much scalable alpha is there, really, in counting cars in Walmart’s parking lots using satellite images or trading on Twitter sentiment?
Moving Beyond Hype and Hope
The answer to all of these questions is: not much, not yet. Right now there’s a lot of hype and a lot of hope and not a lot more. This is happening at a time of increased attention to systematic strategies generally. Three trends are clear:
- Data is far more abundant than it was ten years ago;
- Funds are becoming more data savvy, resulting in “quantamental” teams – fundamental teams who have started to leverage data and quant techniques for screening, idea generation, risk, and portfolio construction – and the very recent proliferation of Chief Data Officers and data sourcing teams;
- And classic quant strategies like Value, Momentum, and Quality, which quants have known about for years, are becoming mainstream through so-called Risk Premia and Smart Beta ETFs and structured products
All of this means that:
- There is supply of novel data sets – there’s loads of data out there
- There is demand for such quant data and strategies
- Classic quant anomalies are getting crowded – not just the Smart Beta strategies listed above, but other well known anomalies such as mean reversion and estimate revisions
But there are few tools which enable the supply to meet the demand. Nobody seems to know which alt data sets are valuable, how to evaluate them rigorously, and how to integrate them into a traditional quant investment process.
The second and concluding installment will examine crowding in the quant space and best practices for alternative data.