Analysis

ETF Insight: Harnessing technology as industry enters new decade

Scott Longley

text

Technology has been a huge enabler to ETFs. Or rather the other way around; ETFs are the leap in technology which has popularised various new forms of investment.

So it is interesting to see advances being made in the new frontiers of data capture and interpretation – namely artificial intelligence and machine learning – are now the cause of some rethinking on the active management side.

According to Unigestion’s CEO Fiona Frick, the embrace of technology is “essential” in the increasingly competitive landscape of asset management.

“Those who invest in it will undoubtedly outsmart those who do not,” Frick says.

“The development of machine learning and the proliferation of alternative data sources offer enormous opportunities to our industry, but they also present a challenge: as machines continue to evolve, how much control should investment managers relinquish to technology to deliver the best results for investors?”

The terminology here is informative. As James McManus, director for investment management and ETF research at Nutmeg, says, the future of investment is “not just machine-based but technology-leveraged by humans and informed by broader and more intelligent data sets”.

Hector McNeil, co-CEO of HANetf, agrees: “AI is already being used by asset managers to help them to harness data, including the reams of market data they have to grapple with every day and to manage risks more effectively.”

“But it will not necessarily deliver an ‘edge’ to active managers over systematic funds.”

Here is the crux of the discussion, perhaps. Frick says a “systemic approach” to investment decisions offers many advantages, particularly in terms of enabling the processing of vast amounts of data not available.

As she says: “Using technology also reduces subjectivity and the risk of emotional bias, allowing us to develop structured, disciplined and repeatable investment processes.” Such would be the thinking behind passive investments.

However, as the continuing debates around smart beta and the search for factors have demonstrated previously, while models offer huge potential, there are flaws.

“Models tend to overfit by design and are also inherently backward-looking, meaning the impacts of unprecedented events cannot be addressed through a purely quantitative process,” says Frick. “Models can identify regime shifts and identify new risk factors, but they are effectively weak at interpreting data and adapting to new paradigms.”

We arrive back, then, with the problems with back-testing.

The Unigestion answer is human intervention. “We believe systematic analysis should be combined with the forward-looking views of investment managers, who can continually assess the relevance of past data and adapt as markets evolve,” says Frick.

She goes on to say she “strongly believes” that advances in machine learning can help active managers differentiate themselves from a passive form of investment.

“There is huge potential for asset managers to use machine learning to support their investment decision-making and deliver better outcomes to investors, especially if backed up by human experience,” she adds. “Thanks to their ability to process much more complex patterns with better forecasting power, modern machine learning algorithms outperform traditional linear regression.”

It all boils down to that favourite saying of those who deal in data – the signal and the noise. Frick suggests that in any data-driven model, the quality of the inputs has a huge impact. Working out whether the data is revealing a “powerful and undiscovered new connection with strong predictive or explanatory power” or merely represents “spurious statistical noise” or a previously undetected bias in the data is the key.

Frick says the Unigestion answer to what remains a low signal-to-noise ratio is to use machine learning to “extract patterns” in order to assess risk, screen investments and adapt their portfolios to new signals.

In practice, it means that Unigestion is now developing algorithms to enhance stock-specific distribution forecasts and to improve stock selection within the private equity space and soon-to-come mimicking of macro factors to strengthen multi-asset risk models.

“If alpha is generated by skilfully exploiting information, the enormous rise in the volume of data available presents opportunities for asset managers to develop much more advanced indicators than in the past and transform their research into new sources of return for investors,” says Frick.

10 years after

Yet, as McManus from Nutmeg says, such a statement could have been true at any time in the last 10 years and the evidence from that period shows that active managers have delivered “little outperformance for investors over this period in many markets”.

“The issue may not be just one of accessing data but also one of drawing the right insights and whether they give you any advantage,” he adds. “One of the biggest trends of the past two decades is the wider availability of data – there is no longer the information advantage for that once existed for some market participants.

“So the proof will be in the pudding as to whether they are able to deliver outperformance, but the endeavour to explore new data sets in search of an advantage is sensible.”

Unigestion admits as much. In the age of “datafication”, they suggest that the challenge for investment managers will be to “extract meaning and actionable triggers” from big data and in this Frick finds some support from McNeil.

He suggests that with room for both passive and active approaches in any given portfolio, there remains the opportunity for active funds to harness the talents of human managers, “perhaps in conjunction with AI and other means of data processing” to achieve “more expensive investment objectives, and…[be] valued for their performance.”

However, the very technology that Unigestion base their argument on is also, potentially, winnowing away the chances of active managers being successful in just this fashion.

“Markets, on the other hand, are going to become more efficient and while machine learning can help to highlight some of the arbitrage opportunities that still exist, they are few and far between these days and those windows are getting smaller,” says McNeil.

He points out that robo-advisers are now making use of machine learning to provide better solutions and outcomes for investing clients at a much lower price point, making informed portfolio management accessible to the mass market of investors.

“In addition, we are also seeing AI being used in advances in index construction which is being reflected in the next generation of ETFs – i.e. thematic, active and niche ETFs,” he adds.

He cites the HAN-GINS Cloud Technology ETF which uses an index based on the screening of publicly available information such as financial websites, search engines and company publications using a natural language processing algorithm.

“The algorithm identifies companies which have or are expected to have a significant exposure to cloud computing using keywords that describe the index theme, and then ranks the companies according to the frequency with which the company is referenced.”

Can Unigestion’s machine/human hybrid compete? Frick suggests the future of asset management will likely involve a synthesis of human and artificial intelligence that harnesses the power of both. Citing Pablo Picasso’s dictum that computers can only give you answers, she says “we do not believe technology can replace investment managers”.

“Computers excel in responding to well-formulated questions with clear objectives, but humans remain key in asking the right questions and interpreting the results.”

Robotic arms

This is a truism. Yet, the distinct lack of consistency in active performance is surely going to be an issue in the new age of data technology. Frick suggests in an age where super-computers could well cancel each other out – a world where no one has the edge – then asset management could be a realm where “you benefit not just from being smart, but from being smart in a different way to others”.

“Machine learning and AI can provide more insight with less human effort, creating more time for investment managers to think differently,” she concludes. “Embracing new technologies will be one way for active managers to outsmart passive ones.”

But will it? It is, after all, possible to see that the last statement turned on its head. What if the embrace of technology actually allows passive to consistently outperform active? Or more pertinently, perhaps, what if smart beta consistently utilises technology to outperform pure active? What if the machines really are smarter than our best and brightest active managers?

That is a question that, to be fair, will doubtless be taxing the minds of the best and the brightest across the asset management spectrum. We await their answers with interest.

ETF Insight is a new series brought to you by ETF Stream. Each week, we shine a light on the key issues from across the European ETF industry, analysing and interpreting the latest trends in the space. For last week’s insight, click here.

ETFs

No ETFs to show.

RELATED ARTICLES