You can find original article here WealthManagement. Subscribe to our free daily WealthManagement newsletters.
In a world where asset managers strive to differentiate themselves from the competition and capture the attention of financial advisors, one approach has been systematic investing. Systematic investing involves using a repeatable, rules-based process, often paired with the use of technology, to come up with investment recommendations based on insights gleaned from both traditional economic and alternative data.
According to BlackRock Systematic, a division of global investment giant BlackRock with $336 billion in AUM, its approach to systematic investing is aimed at delivering consistent alpha returns, even through periods of market volatility. Over a five-year period, BlackRock Systematic claims that roughly 90% of its funds have outperformed peer medians. BlackRock’s Systematic investing team, which comprises 230 people globally, has experimented with techniques such as using machine learning for portfolio construction and works on an investment horizon of three to four months, according to Jeff Shen, PhD, co-chief investment officer and co-head of BlackRock Systematic Equities. The number of market signals the team relies on to make its investment decisions has grown from just three when it started in 1985 to over 1,000 today.
WealthManagement.com recently spoke to Shen about the evolution of systematic investing approach, what types of data sets it uses, how it incorporates AI and large language models into the process and why Shen’s team focuses on active equities strategies.
This Q&A has been edited for length, style and clarity.
WM: Can you talk about how BlackRock’s approach to systematic investing is different from competitors?
Jeff Shen: What we do is try to take interesting data—could be traditional data, could be alternative data—and use advanced techniques, such as machine learning, AI and translate that data through modern techniques into forecasts for active portfolios. We are hoping it will generate consistent and differentiated alpha over different market cycles.
Compared to our competitors, what sets us apart is the usage of alternative data through modern techniques, especially learning AI systems. Also, having an investment horizon that’s three to four months is a little bit distinctive. A lot of quant shops can have pretty short investment horizons, sometimes intra-day. For us, it’s an intermediate horizon.
The last thing I want to mention is that we think about this as a team sport. It’s 230 people working together across the globe, across asset classes and trying to bring that together in one platform.
WM: Can you give me some examples of the alternative data you’ve mentioned?
JS: One is a macro example. The labor market is certainly a big variable that the Federal Reserve looks at very carefully. We have been looking at job posting data over the past six to seven years. At any moment in time in the U.S., there are about 30 million job postings populated on different websites—company websites, some of the aggregated websites.
That gives you a bit of a sense of the health of the economy, who’s hiring, the velocity of hiring, wage inflation because some of the postings indicate their wage range. In that way, you can get a bit of a sense of the propensity of the labor market, the health of the labor market, but also forward-looking inflation indications. It covers both private and public companies, so it gives you a pretty good sense of the overall labor market.
A second example has a little more to do with social media information. We are not interested in individual posts, but the aggregated market sentiment we can draw from social media to use a company as a unit of analysis and see what people, whether retail investors or maybe other firms, are saying about different stocks. And then we try to draw a little bit of the retail sentiment through social media on different companies.
The technology underneath it is the use of language processing, and large language models clearly come into play as well.
WM: What has changed over the past several years in terms of the kind of data and tools you might be using? What new techniques are you seeing rapidly developed in this field?
JS: If you go to the last couple of years, one obvious one is the large language model. ChatGPT was released about two years ago. We had a couple of natural language processing insights that we were already using to transform our technology six or seven months before the release of ChatGPT. Nevertheless, we are using a lot of those technologies, generative AI, large language models, to read through a lot of these alternative data sets and social media, financial news, regulatory filings. You can really think about using machines to read through a lot of these texts, to try read between the lines, to try to find sentiment and interesting insights. That stack of technology continues to evolve, and there are a lot more exciting things on the horizon—multi-language, multi-modal. In addition to text, think about voice, video, image.
The less obvious development has to do with thinking about using machine learning for portfolio construction. Mean variance optimization—maximize returns, minimize risk—has been around for a long time, and there have been pretty interesting developments in using machine learning, using neural networks in particular, for portfolio construction. That part may come as a bit differentiating and may be a surprise to people. We don’t really see too much application of that type of technology in portfolio construction, but we’ve been doing quite a bit of work on that over the past couple of years, and it has been showing quite a bit of promise.
WM: When it comes to using AI in your work, can you talk about the main advantages it offers and maybe some of the limitations of AI in the field of systematic investment?
JS: Maybe I define AI in the narrow sense. When people talk about AI it probably has more to do with the generative AI or large language models. But in the broader sense, if you look at any of the AI book, that’s just one part. A very important part, but there are a lot of other things that are the bread and butter of AI.
I’ll focus on generative AI and large language models first. The benefits are that these things are very good at reading texts and finding insights, meanings, and investment theses. So, we’ve applied that in our security selection, some of the macro investments. In that sense, it’s played the role of a financial analyst. With that investment analysis piece, you can use generative AI and large language model to not only provide efficiency, but to provide scalability. You can do this beyond one stock or one company at a time. You can do it on a large scale 24/7 with very timely updates. That creates huge efficiency and productivity, but also there is precision in terms of finding meaning from the text aspect of it.
In terms of the limitations, clearly, there are two things. One is the generic large language model that you can get from a third party—Open AI, or Gemini, or Anthropic. It doesn’t necessarily cater to financial services as a vertical. So, there are limitations on deep understanding of the particular domain.
The second limitation that’s particular to systematic investing is that time is an interesting challenge for large language models. If you were to do a back test or simulation, you need to make sure that the large language model only knows up to that particular time what’s going on in the world. Otherwise, you get this very strong peek-ahead bias by using an off-the-shelf large language model. If you ask, “Is Nvidia a good investment or not?” today, a large language model knows it’s a phenomenal investment. But would it be able to think about Nvidia without that knowledge in the simulation set 10 years ago? So, point in time in a large language model is certainly an important part.
The last part that I want to raise is to zoom it out slightly. I do think there is a lot of excitement about generative AI and large language models, but there is a whole list of additional technologies and systems that we use that I don’t hear people talking about too much. There is reinforcement learning. There is deep learning. There is a lot more depth in AI. The fortunate thing is that a big part of our group is based in San Francisco, so we’ve had the front row seat to AI revolution for the past 15, 16 years. That’s why we are investing heavily into the space.
WM: A lot of the focus in systematic investment is delivering alpha. In the past couple of years, there has been a particular focus on actively managed funds to achieve that. However, from the research we’ve seen from Morningstar, as well as comments we’ve gotten from financial advisors, it’s tough for any given fund to outperform beyond the short term. How do you deal with this dilemma, and where does the systematic investing approach come in?
JS: Active management is definitely not easy. It’s a zero-sum game. From our perspective, the benefit is our history. Our U.S. equity fund was launched in September 1985. So, we have a 40-year track record of trying to beat the S&P 500, and it’s done very much that.
We’ve also expanded our universe internationally, in global markets, emerging markets.
There is definitely the difficulty for active managers to outperform. We come with a certain level of confidence, legacy and history. But on a forward-looking basis, to deliver that consistency of alpha over time, in our mind, it’s about innovation and innovation at scale. You’ve got to think about new insights and what’s going to be driving the market, which is always going to be a little bit different from what was driving the market before.
I do think using AI and machine learning and things we have been talking about to essentially build scale for investment is becoming more important. When I say “scale,” it means “how many data sets do you have?”
We spend millions and millions of dollars every year on data—technology, systems development. We are also using the BlackRock scale and reach. Trying to drive that scale for the benefit of alpha generation to try to deliver that consistency is a differentiator relative to some of the maybe smaller-scale players.
WM: How do you work with financial advisors on all this?
JS: We have three main sets of products that we engage with financial advisors on. There are quite a few benchmark-driven active mutual funds that we help to run to try and deliver returns that are above and beyond the S&P 500.
We do have market-neutral liquid alts funds. We have a Global Equity Market Neutral Fund [BDMIX] that has actually been around for a while and is gaining quite a bit of traction, given that it’s got market-neutral characteristics. But it still delivers that alpha return for advisors. (Over a five-year period, BDMIX delivered a total return of 5.97%. The Morningstar average for the category is 3.61%.)
And we’ve also gotten a few active ETFs that have gained traction. We’ve got a rotation series—it’s rotating between different factors, different themes. And we’ve got some income active ETFs as well. So, active ETFs is another way to engage with the financial advisors.
WM: Among these three types of products, do you find that they appeal to different segments of the advisor ecosystem?
JS: It’s a bit more firm-specific. There are people who certainly prefer an ETF type of vehicle. For model builders, active ETFs can be quite attractive.
For the benchmark-driven mutual funds, clearly, among some of the wirehouses, there is quite a bit of interest in that. It’s consistent alpha with a reasonable fee, and that’s why there is a lot of traction there.
And then the liquid alts market-neutral fund [BlackRock Systematic Multi-Strategy Fund (BIMBX)], from a portfolio construction perspective, the advisors are essentially using it as a fixed-income replacement, as a high return diversifier in a portfolio. We’ve seen a high rate of adoption for that across RIAs and wirehouses. So that appeals across the spectrum. [BIMBX has delivered a total return of 4.94% over a 10-year period compared to a Morningstar category average of 3.02%.]
WM: What do you include in the definition of “liquid alts”?
JS: We have essentially a global equity market-neutral long-short fund. In any given country or sector, we go long on a bunch of names or short a bunch of names to keep it reasonably market-neutral so there is not too much of a net exposure. It’s very similar to a long-short equity hedge fund, but it has all of the liquid alts characteristics associated with it. It’s a daily liquidity fund. But if you take the return we generate in it and correlate it to the S&P 500, you will have a correlation pretty much close to zero.
WM: In terms of building up your capabilities, have you made any outside firm acquisitions in recent years?
JS: Within the BlackRock systematic group, we haven’t made any acquisitions. On the overall talent perspective, we’ve continued to invest. It’s really been an ongoing journey, investing in technology, data science, AI, machine learning, talent.
What we do here, given this three-to-four month investment horizon, is try to get people who have finance/economic background, alongside people who have engineering/computer science/machine learning background and blend the two to solve the problem. From the talent strategy perspective, we’ve been continuously trying to hire top talent.
One thing I want to mention that BlackRock as a firm has an AI lab that has been a firm commitment for the last six, seven years and there are a few Stanford/Berkeley professors we’ve been working with.