Citywire Selector - For Professional Investors

Register to get unlimited access to Citywire’s fund manager database. Registration is free and only takes a minute.

Union Investment's €29bn funds chief on the rise of robots

Union Investment's €29bn funds chief on the rise of robots

The quants are coming. Headline after headline heralds the AI age and the game-changing impact it signals for fund managers, selectors and even retail traders working on their home computers.

Jörg Schmidt, a senior portfolio manager and head of manager selection within Union Investment’s multi asset division, which oversees about €29 billion in multi asset funds, saw the trend emerging.

Back in 2009 he spoke about how he was both engaging with and getting ahead of the impending quantification of fund selection – light years ago in terms of technological development.

Having seen technology change to become quicker, sharper and more intuitive, at that time Schmidt was drawn into a long discussion with Len Ioffe of Goldman Sachs.

The Frankfurt-based selector and now Citywire A-rated manager, discussed portfolio construction and research-driven quant investing for three hours straight.

When he is reminded of this anecdote on a warm June afternoon, Schmidt laughs. ‘I can’t remember exactly what we discussed back in 2009. But the focus was the same then as it is now: who is making the investment decision?

‘At the heart of it, what we want is diversification and one of the most effective ways to gain that is by blending different asset classes. But it’s the creation alpha – that uncorrelated part of a managers’ performance – where decision making and the way teams are structured comes into the game.

‘From a multi-manager perspective, you can also get diversification through combining different decision-making approaches as not everyone creates the same pattern of performance.

‘We look at fundamental managers but also at quantitative funds because the mixture can be compelling.’

Schmidt is aware of how both his daily and working life are now being increasingly influenced by tech and also algorithmic thinking.

‘In our everyday lives, we are now much more supported by technology than we were five or 10 years ago, so it is not surprising that the same is happening in money management,’ he says.

‘When I look at our internal quantitative investment department and how they have developed over the years, with Big Data, for example, it is much more impressive than it was five or 10 years ago and the same is happening at many asset managers these days.

‘People are hiring data scientists, which is a trend that’s really taking off, but ultimately, we are still not seeing that much pure decision-making based on these kinds of models. Also, you might face model risk if everyone is doing the same thing and using the same quant approach,’ he says.

This is a common theme for Schmidt. While quantitative, or systematic investing, has become more noticeable in recent years, he thinks fund selectors are approaching the question from the wrong perspective if they think this is something active managers should compete with, when it is more the compatibility of the two approaches.

‘As manager selectors, we want to mix different approaches and return patterns but the way that return is generated can be different.

‘It’s about combining man and machine, rather than man versus machine. It is like a pendulum: if it swings too much in one direction then there are opportunities on the other side,’ he says.

Data diving

To keep step with the developments in the industry, Schmidt spearheaded changes of his own. Keen to use all the resources at his disposal, he and his team developed tools to allow a systematic approach to bleed into their portfolio construction – a machine’s eye on human fallibility.

‘We developed internal tools to give us a perspective on drawdowns of managers and how to blend these drawdowns and get insight into overlap between managers, which allows us to explore and compare similar periods of performance.

‘We have spread it across all of our portfolios to find out how to best mix all these managers together to get robust diversification, particularly around drawdowns, over the past two years. That has massively improved the risk/return profile of our multi-manager programme.’

Schmidt says the starting point for this quant-led initiative was not driven by trying to keep in step with tech, but by gaining a deeper understanding of what was really making the company’s portfolios tick.

‘It was a more philosophical debate we had about what risk is and is it just reflected in volatility or are there any alternative measures to assess a strategy’s risk?

‘Drawdown is a good measure for any modern investor to assess longer-term concerns.

‘If you end up with a maximum drawdown of 5% and it continues running over time then this should be reflected in the portfolio. Therefore, we created an alternative view and designed optimisers which can run through seven or eight different algorithms. One of them is purely based on drawdowns and drawdown behaviour,’ he says.

Schmidt likes to present multi-management as three-dimensional, as something which can be adjusted to uncover new angles.

‘Using the example of a cube, we can switch the perspective on the multi-management portfolio, and change the manager mixture or weighting to create something like a drawdown parity portfolio.

‘In this low volatility world, you have to look at different measures of risk, there isn’t just one to fill the needs of every single investor, but drawdown might come close to what “pain” is from an investor perspective.

‘Therefore, it makes sense for us to focus on portfolio construction from our perspective and be aware of something that might force drawdown at the worst possible time. In this way, blending drawdown profiles is an interesting alternative in portfolio construction to avoid cluster risk.’

So, is he preparing for the time when he turns over all his day-to-day working to a machine? Hardly.

‘We have this selection process which also uses quantitative screening but its core element is qualitative analysis, such as the interviews we do. A pure quantitative approach in manager selection would be like cutting off one of the two pillars and focusing on just on one.

'I am convinced there would be poorer returns over the longer term if you went full AI. I cannot imagine how qualitative analysis based on multiple manager interviews could be done by an algorithm,’ he says.

German efficiency

The German and Swiss markets have emerged as an innovative area for AI and quant-focused investing. Allianz Global Investors launched a dedicated strategy to invest in AI-focused companies and Pictet’s Robotics fund has raised $3 billion in assets under management in just 18 months since launch, a clear indicator of investors’ thirst for innovative solutions.

However, Schmidt stresses that several successful quant shops have been in the market for a number of years. He says there can be a tendency among the investment community to focus on what is new and apparently more innovative rather than what is proven.

‘I have sympathy for managers who have been in this space for a long time. Winton Capital, for example, is a dedicated quant manager with a great track record, and they have been using this whole set-up and infrastructure to improve and bring on new strategies and techniques and have implemented them very carefully.

‘We therefore look more closely at how these types of managers have evolved, rather than how new players have emerged. There might be shorter-term developments in the industry, such as robotics funds, but what we are really searching for is a dedicated manager who can deliver.

‘We have followed Winton for 10 years, for example, and we have monitored what they are developing and what kind of models they implement, how they react to Big Data, how they judge machine learning, pattern recognition, etc.

‘In a company like that, where their DNA is designed to work in these changing methodologies, they are much more able to take the next step than someone coming into this for the first time. This is more favourable for us from a manager selection prospective.

‘There is always a case for adding these types of funds, even if the funds themselves have changed quite a bit over time. Such developments are positive as they show an ability to adapt as techniques advance,’ Schmidt says.

Another intriguing development among asset managers, is the ceding of control in the fund manager process to computers. Frankfurt-based ACATIS Investment unveiled its ‘Quantenstein’ project late last year, where it handed a global value equity strategy entirely to a computer system using AI and Deep Learning techniques.

Schmidt, however, is uneasy about the selection community following suit and completely automating investment. If the concept of creating the algorithm and then letting it run without any further human interaction is the future, Schmidt sees problems.

‘We wouldn’t want to invest, if I am honest,’ he says. ‘To use an analogy, if you looked at a football team, and depended on just one characteristic – maybe an entire team of defenders, or left footers or something like that –I don’t think that would be the right way to do it.

‘We get results from an integrated investment process that combines quantitative and qualitative elements.

‘AI-only fund selection could miss what is driving the ‘real’ track record, it may not know if this time period is actually a result of two different track records, with one manager having left a year ago and the next one taking over. Your quant model would see one number and the whole selection process would be flawed.’

Passive problems

When discussing quant-driven funds, the conversation is inevitably drawn to the question of performance fees. If you are plugging into a system rather than paying for a manager’s talent, shouldn’t you pay less?

When asked about fee pressure pushing investors towards passives, Schmidt says this is a false argument. ‘It makes sense to pay higher fees if the end result is better because the only thing that matters for us is net performance after fees.

‘If you drive down fees too much, or create a fee threshold, you could end up with a less efficient portfolio after applying your selection criteria because you miss out on legitimately higher quality performers because of the limits you have imposed on yourself.’

For this reason Schmidt has focused his multi-management team on investing in active areas of the market. ‘We have an open approach which means we are not doing pure fund of funds, but we can invest in single stocks, bonds, funds and also derivatives.

‘We create a very active exposure to active managers in the multi-manager programme at the centre of the book, which is strategic, and we let that run. Then every tactical allocation taken on top of that core positioning is done through futures or options or derivatives.

‘It means we get the best of both worlds, as we can capture the alpha with a very robust multi-manager programme and we can decide to trim down risk without having to de-allocate managers.

‘In a low yield environment, the utility of alpha is much higher than before. If you have a base rate of 3% or even 4%, 1% of additional alpha is fine but maybe not as important as it is now. Today we have negative money market rates, so an additional 1% is much more important. Therefore, I can’t understand why so many people are going passive,’ he says.

Future proofing

For Schmidt the future is about flexibility. Companies such as Winton have demonstrated they can keep pace with changing demands and competition, and fund selectors have to respond in kind. However, he says, the fundamental fund selection process remains intact.

‘Multi-management is always a result of selecting the right managers and putting them together into the right fit. We are all portfolio managers here working in a matrix organisation and every one of us is responsible for managing funds and also covering specific parts of the fund universe.

‘All my colleagues have been doing this for at least six or seven years, so we have built up expertise and specialisation over time. If you combine what these managers know and can do with the technology that is constantly evolving, it shows there is continued improvement – there has to be.

‘We don’t ever want to stop at a buy list, you have to go further and create multi-manager portfolios.

‘Manager selection and portfolio construction is an integrated approach at Union Investment. We always want to see if a manager – be that an active manager or a quant manager – adds something new to our existing multi-manager portfolios.

‘If they do bring something else then the case is there’ he says. ‘We are truly in the era of man with machine, it is not about one competing with the other.’

Schmidt’s enduring interest in systematic investment has drawn him to a number of interesting and innovative firms producing outperformance using quant-driven styles. Here he gives a snapshot of some of his current favourites.

‘We have seen strong performance from Marshall Wace with the TOPS system and also its Liquid Alpha approach. Nordic group Informed Portfolio Management builds up global macro portfolios and is another good performer.

‘Meanwhile, I would name Greenwich-based investment firm AQR when it comes to style and risk premia. We have been invested with them for quite some time. There are also a number of fundamental managers using technology to improve their performance, such as the Velox fund from Melchior/Marble Bar Asset Management with their RAID-Database.’

This article originally appeared in the July-August edition of Citywire Selector magazine.

Leave a comment!

Please sign in or register to comment. It is free to register and only takes a minute or two.
Events
  • Citywire Alternative Ucits Retreat 2017

    Citywire Alternative Ucits Retreat 2017

  • Citywire Milan 2017

    Citywire Milan 2017

  • Citywire Paris 2017

    Citywire Paris 2017

  • Citywire Deutschland 2017

    Citywire Deutschland 2017

  • Citywire DACH 2016

    Citywire DACH 2016

  • Citywire Italy 2016

    Citywire Italy 2016

  • Citywire Milan 2016

    Citywire Milan 2016

  • Citywire Alt Ucits 2016

    Citywire Alt Ucits 2016

  • Citywire Berlin 2016

    Citywire Berlin 2016

  • Citywire Switzerland 2016

    Citywire Switzerland 2016

  • Citywire Amsterdam 2016

    Citywire Amsterdam 2016

  • Citywire Montreux 2016

    Citywire Montreux 2016

  • Citywire Deutschland 2016

    Citywire Deutschland 2016

  • Citywire Latin America 2016

    Citywire Latin America 2016

  • Citywire Milan 2016

    Citywire Milan 2016

  • Citywire Munich 2016

    Citywire Munich 2016

  • Citywire Paris Alt Ucits 2016

    Citywire Paris Alt Ucits 2016

  • Citywire Zurich Alt Ucits 2016

    Citywire Zurich Alt Ucits 2016