Macquarie’s $59b number-crunchers embrace AI to beat the market

“We’ve been using these [AI] techniques for over 20 years. But we’re really excited about generative AI and imagination and creativity that we are going to see in the industry,” says Leung.

Scot Thompson, the other co-head of the systematic equities unit, says AI and machine learning allow the team to deepen its analysis of the more than 2000 global listed companies it covers.

“We have a view on everything, and it’s really hard to find the people to do it, and then to be able to galvanise all that information into something that can actually be used to build a portfolio,” Thompson explains.

The fund manager says AI also allows the unit to create multiple universes to train the portfolio and to synthesise instant market information to make sure the team’s trading is informed.

Thompson and Leung have worked in Macquarie Asset Management’s equities team, which oversees $219.5 billion, for about two decades and have been co-heads of systematic equities together since 2014.

In that time, the assets under management of the systematic equities’ strategy have grown from about $40 million to about $59 billion.

Although there’s been a general move of assets from high-cost active funds to low-cost passive funds, Thompson says there’s been a trend away from simply hugging benchmarks towards actually delivering outperformance using cost-effective strategies.

“One of the misnomers is that systematic is passive,” he adds.

In fact, Macquarie’s systematic equities unit has delivered outperformance of 1.68 per cent per annum over five years for Australian shares and about 5.1 per cent for Australian small caps.

Dominant force

Thompson says a well-identified characteristic in the Australian sharemarket is sentiment. That is measured by price appreciation and improving financial metrics, which encourages more buying and further share price appreciation.

“On the smaller end of the [sharemarket] scale, this force is much more dominant, which suggests to us that there is a bigger impact from less sophisticated retail investors,” he says.

What is unusual about the last two years, Leung points out, is that the Australian sharemarket indices have been relatively stable. The index itself hasn’t had many additions, deletions or reconstitutions.

“In our last 20 years of working with the index, the last two years are probably the most quiet,” he says.

Another trend is that superannuation funds, in the aggregate, are more wary of deviating too far away from the benchmarks against which their performance is judged. For Australian equities, that benchmark is typically the S&P/ASX 300.

But super funds are also facing pressure from members to divest from fossil fuels. That in turn creates the risk that they will drift from their benchmark, underperform and face the indignity of informing members of their shortcomings under the Your Future, Your Super regime.

Thompson says the systematic equities unit has been helping institutions solve the issue of reducing the carbon exposure of the portfolio without deviating from the index too much, by “filling the gaps and address[ing] that tracking difference”.

The solution is “elegant” says Leung, and involves finding stocks that have similar characteristics to high carbon-emitting ones. They could be companies exposed to rising inflation, that tend to display high momentum or relate to the performance of certain geographies.

“The systematic process will be able to go ‘well, if you’re underweight a certain carbon-intense name, what other names can we insert [to replicate the exposure]?’” says Leung.

In the US, the dominance of the so-called magnificent seven is a big talking point among investors. The mega-cap tech stocks that include Apple and Nvidia have become even bigger and account for a historically higher portion of the sharemarket.

Concentration limit rules

The market became so top-heavy that one of the Nasdaq’s concentration limit rules led to a change in that index to lower the weighting of certain stocks.

However, this raises a philosophical issue about the purpose of an index. The reality, Thompson says, is these seven companies are, for better or worse, “the market”.

“Yes, there are seven names that are driving the market, but that is actually the market,” says Thompson. “It’s unusual, and it may end up with some risks that we have to be aware of. But the fact that they’re the biggest is because they’ve had the greatest support.”

So the question about whether a benchmark can be too concentrated if it actually represents the true market weight is a difficult philosophical one.

“In Australia, we’ve got 80 per cent of the index in the top 20 names, and we seem to live with that,” says Thompson.

Whatever the benchmark, Macquarie’s unit will use the powers at its disposal to beat it, Leung says, adding that organising the data is their special skill.

“There is probably more value-added or greater opportunity in assembling the data source as opposed to the silver bullet of a super signal that’s going to find the best-performing stock every time,” he says.

“Information and data is so prolific and generally available. Our edge is how we pull that together.”

There are about 60 broad signals that systematic strategy relies on. They range from financial metrics to market-driven information such as share price moves, informed market participant activity such as insider transactions and analyst forecasts, and non-structured or alternative data.

Thompson says there’s an increase in the availability of alternative data sources, and they often source this data to test whether there is a signal.

“This might be natural language processing to assess [earnings call] transcripts, for example,” Thompson says. “We can gather that information, test that and determine whether that is worthwhile pursuing?”

The advances in AI – from machine learning to generative AI – have further encouraged Leung and Thompson to believe their systematic approach can deliver good outcomes for their investors.

But they’re conscious that it will work only for so long. Leung says studies reveal that the intelligence of generative AI tools such as ChatGPT will actually decay over time.

“It feeds on itself and becomes an echo chamber,” he says. “These are interesting computing problems that they have to solve before it becomes true artificial intelligence.”