AMP has evolved its DAA approach to be largely systematic in nature. We speak with Stuart Eliot about the implementation of the systematic DAA model in the investment portfolio.
Starting in May 2023, AMP has revamped its dynamic asset allocation (DAA) model from a qualitative approach to a mostly systematic model as the investment team has been bedding down a broad series of changes to the investment strategy following the sale of AMP Capital and the subsequent transfer of the Multi-Asset Group to AMP.
Stuart Eliot, Head of Portfolio Management at AMP, has been leading the changes to the DAA model and he was keen to implement a more research-driven approach to identify signals that leverage off the company’s ability to implement a nimble and highly diversified derivatives-based active asset allocation program to complement being a long-term investor.
Unless there's some kind of exogenous event that is incredibly material, you don't want to second guess your models because a lot of them are designed around trying to capitalise on a human behaviour that is repeatable
“We investigate an insight that we think might deliver returns and that could be trend following, value or something like that. You test your hypothesis, write a research paper and then that gets peer reviewed and goes through our governance processes,” Eliot says in an interview with [i3] Insights.
If all goes well, then ultimately it gets added to the DAA. This systematic approach to DAA helps the team avoid the pitfalls of behavioural biases, Eliot says.
“Unless there’s some kind of exogenous event that is incredibly material, you don’t want to second guess your models because a lot of them are designed around trying to capitalise on a human behaviour that is repeatable. And so we know that we’ll be able to profit from these behavioural biases over time,” he says.
“It’s extremely rare that we would intervene in the DAA process, but it doesn’t mean we would never intervene. If the model says ‘buy’ and it’s obviously a bad idea, we wouldn’t do that.
“What we might do is be a little bit judicious about the timing of how you might implement something. If I get a buy signal on Monday and the market is crashing because of a weekend tweet, then I might do the buying over three days rather than one just to spread it out a bit.”
Although some elements of the previous DAA process were retained, Eliot and the team largely rebuilt the new process from scratch, adding one signal at a time. It currently has 16 signals in the portfolio across seven asset classes, including equities, bonds, commodities, foreign exchange and even digital assets. This degree of diversification means each trade is relatively small compared to the overall portfolio in order to manage risk well.
“We try to calibrate the active risk from DAA to be approximately the same as the active risk from manager stock selection,” Eliot says.
“We’re targeting a Sharpe ratio of better than 0.5. Our back test says a much higher number than that, but we tend to treat the numbers that come out of a back test conservatively.”
Since the DAA process spans different funds with different risk budgets, the target return for each fund from DAA is slightly different. But on average, Eliot says the program aims for about 25 basis points in added returns.
“We still do a degree of qualitative DAA – and the overweight to US shares last year was a good example of that – but previously that was happening a lot more frequently than it is now, now that we have the systematic DAA process. So maybe we make two or three major qualitative decisions a year,” he says.
Reversing the Overweight to US Equities
The overweight to US equities paid off handsomely last year, as the Magnificent Seven stocks extended their run. But the trade war sparked by US President Donald Trump’s tariffs and the announcement of a lower-cost large language model (LLM) in the form of DeepSeek have made the team rethink the role of the Magnificent Seven and the risk in the US equity market.
As a result, Eliot’s Portfolio Management Team has temporarily taken off the overweight to US stocks.
“We are still positive on US shares and the US economy over the medium term, but we have recently formed the view that the balance of risks, at the current time, skews unfavourably. This is not a forecast that the market will fall, just that the risk of a sell-off has meaningfully increased,” he says.
His view on how the relationship between the Magnificent Seven and the rest of the US stock market will develop over time has changed too with the announcement of the Chinese low-cost LLM. Whereas previously, he expected the Magnificent Seven’s dominance to continue for some time before the benefits of artificial intelligence (AI) spread to other companies, he now thinks this process will be accelerated with the advent of DeepSeek.
We are still positive on US shares and the US economy over the medium term, but we have recently formed the view that the balance of risks, at the current time, skews unfavourably. This is not a forecast that the market will fall, just that the risk of a sell-off has meaningfully increased
“We had a thesis that spending on AI infrastructure would continue to drive markets and economies this year, and this would probably maintain or exacerbate that concentration [in the US market]. But then as companies start to work out how to gain productivity benefits from implementing AI, that would then reverse the gap and you’d have the S&P 493 outperforming the MAG-7,” he says.
“But I think the DeepSeek news has changed that; it has just brought it forward. I don’t want to make a prediction about how it’ll change this year, but I think cheaper LLMs will just improve productivity for virtually every company.”
He is unperturbed by reports about allegations of DeepSeek having distilled portions of ChatGPT for its own model or it underrepresenting the investment cost of the model, even if it means the real cost of developing DeepSeek’s R1 model might be as much as 10 times higher.
“I don’t think it matters how they did it, even if they understated the cost of how they did it. If it’s true that the cost of a query is lower, then that’s the important information. That’s the thing that will lead to improved economy-wide productivity,” he says.
Adding New Signals
AMP Investments manages over $70 billion of assets across a range of multi-asset funds, including for superannuation, retail and institutional clients.
The company has seen many changes in the past couple of years since its decision to break up AMP Capital and sell the real estate business to Dexus, while the global equity and fixed income business went to Macquarie Asset Management.
Since then, the investment team, under the leadership of Chief Investment Officer Anna Shelley, has reshaped the portfolio to add more passive and systematic strategies in the equities portfolios, reduce its exposure to several active strategies, including a reduction to global listed property, and remove hedge funds.
This allowed the team to free up fee budget for more attractive asset classes and it has added direct infrastructure assets and diversified private credit, including high-yield loans, securitised debt and emerging market debt.
“Over the last two years, we’ve gone from being fully active in Australian shares – and fully active can be a wild ride sometimes – to depending upon a blend of index and active. That then frees up the tracking error budget and the fee budget to do things like private debt, direct infrastructure and those sorts of things without making the product more expensive,” Eliot says.
Over the last two years, we've gone from being fully active in Australian shares – and fully active can be a wild ride sometimes – to depending upon a blend of index and active. That then frees up the tracking error budget and the fee budget to do things like private debt, direct infrastructure and those sorts of things without making the product more expensive
Eliot joined AMP in April 2022 from Pendal, where he was a Senior Portfolio Manager in the Multi-asset Investment Team and was also involved in DAA and tactical asset allocation. Although at AMP, he also has responsibility as a portfolio manager for some funds in addition to leadership responsibilities, one of his key focuses is on DAA. And over the years, his approach has changed, he says.
“The way I think about DAA has evolved in my time here and it’s much more pragmatic now when maybe I used to approach it more academically. I used to think that a signal needed to work symmetrically, so it could add value long or short. But we’ve actually got other things that will help the portfolio in that regard and so I have evolved to being quite comfortable with a DAA process that leans long,” he says.
As he has expanded the range of signals, he has implemented a number of trades that lie outside the arena of traditional factor exposures such as value, growth or momentum.
“A lot of the signals are the sort of things you’d expect, but there’s quite a few novel things in there as well. One is based on market microstructures; how various sectors are performing relative to one another can actually give you a pretty good read on the likely direction,” he says.
Another one is Bitcoin. AMP’s funds don’t invest directly in the coin or in any other cryptocurrency for that matter. But he has developed a model for producing returns based on the direction of the Bitcoin price, using futures linked to Bitcoin.
In contrast to Bitcoin, these futures are listed on the Chicago Mercantile Exchange, the main exchange for derivatives, which comes with all the checks and balances of a global market.
“Once the US-listed ETFs (exchange-traded funds) were approved by the SEC (US Securities and Exchange Commission), we started doing research on whether or not our existing models could be applied to the Bitcoin price. And it turns out that they worked very well,” Eliot says.
“Sometimes with small modifications to the way you apply the signal, but we could reuse price trends, general cross-asset sentiment and liquidity to take positions.”
He says it has been one of the most successful trades in the DAA program so far.
“It’s partly because Bitcoin is a less sophisticated market; there is not a lot of institutional participation. And so that allows us to build models that are quite effective in capturing the upside without too much exposure to the downside,” he says.
The exposure across the various portfolios is small though, and while making a positive contribution, it is not a material driver of returns.
__________
[i3] Insights is the official educational bulletin of the Investment Innovation Institute [i3]. It covers major trends and innovations in institutional investing, providing independent and thought-provoking content about pension funds, insurance companies and sovereign wealth funds across the globe.