State Super has established an Academic Oversight Body, to ensure good governance of the fund’s activities in machine learning.
In 2016, State Super embarked on a journey to see if machine learning models could assist in making sense of the vast amount of new data that is produced on a daily basis and making sound decisions in complex financial markets.
Now, four years later, the fund believes it is time to build a robust governance system around their machine learning models and tap into the knowledge of academics who are at the forefront of this discipline.
State Super has appointed Dr Michael Kollo, Professor David Michayluk and Dr Alex Antic to be the inaugural members of the Academic Oversight Body (AOB), which aims to provide oversight of the development of the investment data science capability.
Kollo is the former general manager of quantitative solutions and risk of pension fund HESTA and is currently General Manager at Faethm.ai, a future of work analytics company. He has been appointed as chair of the AOB.
Professor David Michayluk is Head of the Finance Department at UTS Business School and Dr Alex Antic is Head of Data Science at Australian National University.
The Academic Oversight Body gives us academic oversight, because they have the brainpower to do the ongoing research that we don’t have time for. They are the ones that are really plugged into this space
“The AOB gives us academic oversight, because they have the brainpower to do the ongoing research that we don’t have time for. They are the ones that are really plugged into this space,” Charles Wu, Deputy Chief Investment Officer of State Super, one of the initiators of the machine learning program at the fund, told [i3] Insights.
“If you think about it, we get the best of both worlds. Academics generally have access to ongoing research to all the new algorithms. Tapping into the academic network allows us to access this knowledge in a guided manner.” he said.
The establishment of the body doesn’t necessarily mean State Super is planning to expand the machine learning program significantly in the near future. Wu sees it as an exercise to strengthen the governance and ensure the team continues to do the right thing.
“This is the kind of thing which started with two people and a Bloomberg terminal, but now, four years later, things have changed. We need to make sure that it is adding value to our members and that our knowledge is sufficient.” he said.
Wu pointed out that using insights from complex machine learning based models requires a strong governance process to avoid building in biases during the learning process, and to challenge the path of future developments.
The decision to strengthen the governance for machine learning is a sign that this field is maturing, Kollo added.
“If you think about what asset owners had to do in the past, they relied upon asset managers and conferences for insights, but increasingly they are building independent analysis, which helps them to understand markets,” Kollo told [i3] Insights.
“To a large extent, this is a sign of maturity of the industry. We provide an independent technical voice that State Super can rely upon, so they can make better decisions, using better information,” he said.
Kollo sees the role of the oversight body as to provide communication to State Super – on the machine learning developments, oversight of the models produced by State Super and provide academic insights into new developments in this space.
“Alex and Dave have an army of research students and papers at their disposal, so State Super is able to draw upon those resources to create their models and get that advice,” he said.
John Livanas, Chief Executive Officer of State Super said the machine learning program has been successful not just in terms of the return generated, but also in minimising risk.
Charles and his team have developed machine learning algorithms that have been helping us to effectively assess financial conditions and assist our decision-making processes. This has led State Super to maintain superior returns, with significantly lowered risk
“The challenges faced by our funds, of minimising downside risk while pursuing strong returns, in a liquidity constrained manner, are well known,” he said.
“Charles and his team have developed machine learning algorithms that have been helping us to effectively assess financial conditions and assist our decision-making processes. This has led State Super to maintain superior returns, with significantly lowered risk,” he said.
In May this year, State Super also awarded a mandate to Neuberger Berman for its machine learning strategy, led by Michael Recce, Chief Data Scientist at the firm. The amount of the mandate has not been disclosed.
Wu said the firm was among the few managers that has really embraced machine learning.
“If you remember that a couple of years ago this was a very trendy area. But there were a lot of fund managers that put one data scientist in their investment team and said they could do machine learning. But you need deep resources to do it properly,” he said.
[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.