Janet Campagna is the CEO of QS Investors, a 100% employee-owned, majority woman-owned investment management company based in New York and founded in 2010. Prior to founding QS Investors, Janet was a managing director at Deutsche Asset Management serving as the global head of quantitative strategies and a member of the global operating committee. She was a principal at Barclays Global Investors from 1994–99 and from 1989–94 an associate at First Quadrant and director of asset allocation research. She is a board member of the Mott Haven Academy in the South Bronx, a charter school for at-risk students in the foster care and child welfare system, and a member of the MFE Steering Committee for the Haas Business School, UC Berkeley. Janet has a BS from Northeastern University, an MS from California Institute of Technology, and a PhD from the University of California, Irvine, in 1990.
In general, the reputation of quant managers took a hammering as a result of the crash. The media pretty much blamed quants as one of the causes of the crash and certainly highlighted the losses of high-profile investment bank quant funds through the downturn. Do you feel that was unfair to mathematical investing?
Although there is no doubt that a number of quantitative funds turned in results through the crash that would not have pleased investors, there is a mistaken impression in the media—and to a lesser extent perhaps among some investors—that all quant funds are the same, and that therefore it makes sense to talk about “quants” in general terms. This paints all quants with the same brush. In reality, there are many different ways of using mathematical techniques to generate alpha, and different funds take very different approaches. Recent academic research shows that the correlation among different quant fund managers is actually substantially less than the correlation between traditional, long-only active investment managers, particularly during market crises. That, by itself, should suggest that generalizations about quants are not going to be particularly helpful, accurate, or even interesting.
What is certainly true, however, is that there is a very strong demand from the investor community for transparency. Not everyone has the background to understand the mathematics, but they expect a sensible layperson’s guide to what it is that we are doing, and we are very happy to provide this. In our case, at QS Investors, we are particularly keen to do so since we believe we are taking an approach that enables our models to be very responsive to changing market conditions and to changing investor attitudes in the market over time. Our results show that incorporating these dynamics generates an enhanced performance over time.
What we have in common with many quant managers is that our approach is a relative returns one, which means that our funds will try to beat the appropriate benchmark for the universe of stocks that we are targeting by a specific amount. There are quant funds that offer an absolute return strategy—one that focuses on capital preservation under all market conditions—but that is not the role of a relative return fund.
We offer various levels of outperformance and construct portfolios that are designed to produce a risk–reward outcome that is commensurate with the target benchmark for that specific portfolio with its risk and return objectives. This gives both our institutional clients and wealth management clients the opportunity to select one or more of our funds to match what they are trying to achieve with whatever portion of their overall funds they assign to us.
This is actually a very important point to get across. We are not in the business of second-guessing the client’s portfolio allocation strategy. They will have already decided how they want to allocate their portfolio across a range of asset classes, and we will get a chunk of the monies that they have decided to invest in equities. Our job is to attempt to fulfill our performance goals while staying within the risk parameters that we have set up for that particular client’s portfolio. If they have decided to invest in, say, our global equity fund for a 3–4% outperformance target, we will have done our job if we beat the global index by 3–4%. If the whole market is down 10%, we will still aim to be around 4% better than that, so our portfolio will be down only 6%. But we won’t suddenly try to outperform the market by 14% to give an absolute return of 4%.
“Doing well” means, in our terms, achieving the goals we set out to achieve. A number of our funds achieved exactly this through the downturn of 2008 and turned in performances that beat their benchmark indexes by 3% to 4%. Investors appreciate and put a high premium on consistency of performance over time.
How do you manage to take “views” of the market into account without compromising the mathematical approach by adding subjective opinions to the mix?
One of the most interesting things about our quantitative strategies is that we use statistical techniques to determine where in the market cycle we are at any time. Note that these are statistical techniques, not opinions. When you have a view on where in the market cycle you are, that helps you to understand which of the insights you have into the market are the most important at any moment. The kinds of things we look at when picking stocks for the portfolio are valuation information, sentiment information, and information on growth, but we determine how important each of these is through another layer in the model which looks at what is going on in the market. Is it, for example, a period of fear, of greed, or of glamor? Is there likely to be a junk rally? We have had this additional dynamic layer in the model for the last five years and it has helped us to navigate through multiple phases in the market cycle, including junk rallies, the liquidity crunch, and the banking crisis. It has also enabled us to show very good results through the financial crisis.
Investors expect consistency, discipline, transparency, daily liquidity, and strong risk management. Our risk management is very mathematically driven, but the management team brings decades of market experience to the process and the model is never simply on cruise control. We monitor the results very carefully. We try to measure risk continuously, and we look across some 15 risk measures to achieve this. You have to monitor risk, note the changes, and understand what is driving those changes. Understanding your risk model and how it works is very different from blindly following a model.
However, the way we bring management team judgment to the table is very different from the way a traditional, long-only management team would set about things. We do not want those judgments coming in the form of emotional reactions to what happened yesterday. Our judgments are all about what goes into the model at the front end, and the judgments are made well ahead of time.
For example, in the run-up to the crash, with markets showing an almost continuous rise, volatility reached very low levels. There were any number of headlines in the press about “the death of volatility.” Quant models, including ours, derived assumptions about low volatility from their analysis of data. However, we absolutely did not believe that markets were on an everlasting upward path with consequent low volatility as an enduring feature. It was clear to us that there would be some sort of crash and that volatility would return with a vengeance. So we built and tested a model that would be much more sensitive to conditions of high volatility than the low-volatility model we were using. As soon as market conditions changed, we made the judgment to take the high-volatility model off the shelf and swap it in for the low-volatility model. We had already fully tested and validated it and it was good to go. That is how we use human judgment to steer the process. It is not about saying “I don’t like IBM today, I think I’ll sell…”
What questions are you asking yourselves as a management team when you monitor the way your portfolios are tracking?
There are three things that go through our minds all the time, irrespective of whether the model strategy seems to be working or not. We watch to see if we are correctly capturing risk and alpha (outperformance), and we look to see if we are well positioned with respect to what the market is doing. We are also constantly doing research both to improve our forecasts and to find ways in which market factors might be shifting and changing. If we have an insight about how alpha could be captured, that insight does not take the form of a buy or sell judgment on a particular stock. It comes as an algorithm that we want to model and test.
As an example, one of our researchers raised a question in a recent meeting as to whether, given the patterns in the currency market, various hedged positions would continue to work. In a traditional, long-only fund that might lead to a decision to intervene directly to change the hedged positions. With us, what we look at is whether there is any forecasting information in that insight. If there is information, then we would incorporate it into a strategy. We would want to develop a modeling factor from it that would track patterns of over- or underperformance across the currency markets. Then we would test the performance of that new formula and decide whether or not to include it in the model on the basis of the test results.
What makes our approach very computationally intensive is that we look at a large number of factors for each stock in a large portfolio of stocks. We’re looking at all the stocks in that universe, and we look at how the price of each of these stocks is moving relative to other stocks. Then there are those 15 dimensions of risk that I mentioned which we monitor on a weekly basis. When any of these look out of whack, we investigate it to see where any of our portfolios are vulnerable.
You have given one example, with the low-volatility/high-volatility mode, of how you might make a judgment call. Is this just a special instance of a constant process of recalibrating your models to achieve a better fit with reality?
You have to evolve in this business, and fine-tuning models is part of evolving. But if you are just reactive because you find that your model is underperforming due to some specific set of events that just happened, you end up chasing what happened yesterday and you cannot possibly perform well tomorrow. Our mantra is to be responsive, not reactive. It is necessary to challenge and question assumptions all the time, including our own “wisdom.” We schedule regular strategy meetings to question and challenge our present strategies. We bring people in to these meetings who do not work on a particular strategy and ask them to try to tear the strategy apart, to find weaknesses in it. It is important to stress that we are investors first and foremost, and quant is just our way of thinking more effectively about the market. We are always asking when we would expect a particular risk model to overestimate or underestimate risk—and “when” here refers to particular states of the market. This is why the ability to incorporate views into the strategy is so important for us.
With a long-only investment strategy, we would be looking constantly to see if we are overweight, underweight, or level relative to the cap index. We want to be in securities that are relatively undervalued and not in securities that are overvalued. However, there are market conditions where stock valuations have little traction with the market. This is why we look at sentiment as well. If fear is driving the market, a valuations-weighted model will not have much bearing on what is happening, so the model will downgrade valuation and increase the weighting given to sentiment. Another dimension might be growth orientation.
We have a lot of respect for fundamental long-only analysts and investment managers and we take advantage of the information available out there from analysts. However, we do this with a very systematic structure and we look to strip the emotion out of it. This is one of the reasons why we do not do the things that fundamental managers do. We do not go and visit companies or pore over their balance sheets. I have never looked a CEO in the eye to see if a company strategy makes sense. Instead we look purely at the numbers, at relative yields, earnings-to-book price ratios, and the speed with which prices are changing. We can do this across a number of factors, over thousands and thousands of stocks, and we can track how these securities are likely to move relative to each other. This is hugely computationally intensive and it simply could not have been done a decade ago. We also work to understand what sort of exposures to systematic risk we have in our portfolios. When we add value to a portfolio, it is this analytical work that generates the value, the alpha. Because this is a mathematical process, it has a discipline and a transparency to it, and these qualities get lost in the talk about all quant funds being alike.
To provide some examples, in our US equities portfolios we run a market-neutral strategy where we are 100% short and 100% long. That portfolio returned 8% above the benchmark in 2008, when quant funds were supposedly doing terribly. With our global strategy we were up 4% versus the benchmark in 2007, and we outperformed again through 2008 purely using quant techniques. When the market is very nervous, and is characterized by oscillations between risk-on and risk-off trades, taking the market environment and investor sentiment into our modeling processes via our views approach helps us to navigate in what would otherwise be treacherous conditions.
However, if the markets move into economic Armageddon territory, then, of course, we are in the same boat as everyone else. Current market conditions throughout the developed world are tough. Growth rates are slow. Countries have ageing populations, with the associated growing public-sector costs, which pushes up deficits. Healthcare and pension costs are on unsustainable trajectories. There are lots of macro issues. From a quant point of view, you just have to figure out how you model these effects on a shorter-term, tactical basis. So far, quant funds seem to me to be doing a pretty good job.