Q: What is the history of the fund?
Quantitative Investment Strategies has been part of Goldman’s Asset Management business since 1989, and I joined the team in 2005. The team started to invest using quantitative techniques that were being pioneered in academic institutions around the world. Behavioral economists and financial researchers were identifying explicit characteristics of stocks (called factors) that were useful in predicting future stock returns.
Examples of those characteristics were factors such as ‘value’: value stocks tend to outperform growth stocks. Our investment business was focused on creating portfolios that invested based on these types of factors.
We started managing large-cap equity strategies in the late 1980s, before getting into managing small-cap strategies using similar data-driven models in the late 1990s (we launched our first U.S. small cap mutual fund in 1997). The Goldman Sachs Small Cap Growth Insights Fund was launched in 2007, driven by client need to have a specific small-cap growth option and a specific way of investing in the small-cap growth space using more quantitative models.
In terms of the broader set of assets, the Quantitative Investment Strategies group manages approximately $100 billion in assets. Of that about $20 billion is managed by the team that I am a part of, our quantitative equity team. We manage about $1 billion in U.S. small-cap strategies and approx. $2.5 BB in ex-US (international) small-cap strategies. The Small Cap Growth Fund has over $300 million in assets.
Q: How do you define your investment philosophy?
The core beliefs that guide our investment thinking, have not changed over 30 years. We are data-driven. We use a very clear, transparent set of criteria to identify stocks and then evaluate whether they are attractive or not.
Until about 10 years ago, quantitative analysis utilized mostly financial-statement-based metrics, such as calculating the price to book ratio of companies, the return on equities, and perhaps the stock price movements. However, in the past 5 to 10 years we have been able to analyze much larger data sets on companies. This allows us to better understand what’s going on inside companies (their operations, intra-quarter sales, etc.) but what is also going on in markets that these companies are in (like themes/trends in the broader market). Our computers can process unstructured or text-based data sets, volumes of information, to identify which companies look attractive. But our philosophy of investing has not changed; it has always been very data-driven.
Conceptually, we identify companies that are cheap, growing and high-quality; as well as ones that are benefiting from positive themes and trends in the market; and that have the right amount of public sentiment around them. But we believe the kinds of data that are available and the way we can use that data objectively has really given us a competitive advantage.
Our goal is to identify attractive stocks that differ from what the rest of the market thinks is obvious and attractive, because we use unique alternative data and a more quantitative approach.
Q: What is unique about investing in small-cap companies?
The number-one factor that makes investing in small-cap companies interesting is that this asset class is clearly a little less efficient than, for example, the U.S. large-cap segment. There’s less information available on small-cap companies, fewer analysts covering them, fewer newsletters being written about them.
But the fact that the small-cap space has a little less traditional research, that it is a little less efficient, offers us greater opportunities to drive alpha from an active management standpoint. It actually works to our advantage; it gives us a bit of an informational edge.
Another important aspect of the small-cap space is the breadth of the universe: there are a lot of small-cap opportunities. This breadth allows us to diversify our portfolios across lots of companies that we find attractive. The broad size of the small-cap universe is often considered a challenge by many traditional investors but for data-driven quantitative investors it is actually an attractive quality.
A third aspect that defines the small-cap segment is that it is a slightly less liquid segment of the market. That matters to us most in deciding upon and quantifying the true net-of-transaction cost benefit of buying or selling each company.
Q: What is your investable market capitalization range?
This is a small-cap growth fund, so the benchmark to beat is the Russell 2000 Growth Index, but we allow ourselves to buy any stock in the Russell 2000 Index. If we think there is an opportunity there, for us it is a good way to drive outperformance. Having said that, we still ensure that our portfolios have characteristics and attributes that are very similar to the Russell 2000 Growth index even though we may invest in some stocks that are outside of the index.
At the high end some companies are almost in the $5 billion range, but the vast majority of the stocks in our investment universe are in the range of $500 million to $4 billion or $5 billion.
Q: Would you describe the pillars of your investment strategy?
We analyze all stocks in the Russell 2000 Index. So, there are about 2,000 stocks in our investment universe, and every single day we re-estimate and re-calculate what we think their expected return is going to be over the next year relative to the market. To do that estimation of expected return we look at a large set of data on those companies, generating an explicit view on every company in the universe. We want to have a very robust and complete set of data on all the companies in order to forecast their returns.
The metrics we look for fall into four broad categories. First, we want to identify whether or not a stock is trading at a discount to its industry peers. So we do a bit of valuation analysis, like using traditional financial statement data to build bottom-up value models, or looking at some different unique alternative data sources that we think are useful in better forecasting company valuation. The end goal here is to identify if the stock is trading at a discount.
The second pillar is an assessment of a company’s quality, its profitability, and its fundamental strength. The goal here is to identify high-quality companies that are stable and well-capitalized but also that are gaining market share, are profitable, and will offer a surprise on the upside when they announce their earnings. Here, too, we use a whole host of traditional financial statement data to run some of these calculations, but we also use other data sources to enhance our return predictions. For example, over the past two years we have been using web traffic data, looking at how many people are on the company website and how much time they are spending there to help us quantify a good or growing business that is going to surprise on the upside from an earnings standpoint.
The third pillar is sentiment, what people think about a company. We do that analysis with proprietary tools and technologies we have built, where our computers read news articles and earnings call transcripts to see what a company’s management is saying about itself, or analyst research reports to see what they are saying about a company. A whole host of machines are reading unstructured data, coupled with analyses of market variables like credit default swap spreads, options data, short interest data, to see whether people are shorting the stock or not. And all that helps us form an opinion on what the sentiment is about the company overall.
And finally, the fourth pillar is an analysis of themes and trends to which a stock is exposed—looking at a company’s own stock price movements and those of its customers, its suppliers and competitors and partners, and companies engaging in similar types of businesses around the world. There is a whole family of metrics that tries to capture whether a company is benefiting from positive themes and trends in the stock market.
Those last two pillars—sentiment, and themes and trends—are more technical in nature, while the first two—valuation, and assessment of a company’s profitability and quality—are more fundamental in nature. Overall we use a very large set of characteristics of companies in those four categories to rank and evaluate companies, and then to forecast their expected returns, every single day.
Q: How does your investment process help in separating companies?
The output is a ranking for each one of the companies. In fact, it is more than a ranking, it’s an expected return. The higher the company rank, the higher we think its expected return is going to be. So the output of our model every day is a company-by-company expected return. Internally we call it the alpha of each company.
Although it doesn’t change very much day to day, we do see changes over time, especially when there is a lot of sentiment or news on a particular company, or new information that could change the expected return estimate of a company, or if a company announces earnings. But the real output of this investment process every single day is a company-by-company expected return. We purposely chose the kinds of data we use to predict expected returns over the next one year.
Part of our decision process is to look at the return we expect to get from a company and then decide whether the cost of getting into it and out of it justifies that expected return. That’s a tradeoff that is much more important in less liquid segments of the market like small cap. We are cautious with getting in and out of securities, and need to have much more conviction in their potential returns when we do trade stocks.
Q: Can you give some examples to illustrate the process?
Airlines would be a good example. In the course of 2015, airlines saw a big move in oil prices and the sentiment around travel. Early in the year, a handful of airlines looked relatively attractive from a fundamental standpoint. Both the discounts relative to industry peers and their profitability along with the quality of their operations were attractive. Yet, on the sentiment side and the quantification of themes and trends there was not much that seemed compelling, and so we didn’t really take any large positions.
Oil prices steadily decreased in mid-2015, and as a direct consequence we saw sentiment around these particular airlines change significantly. The news articles were a lot more bullish and the analysts were much more favorable. The market’s expectation of the segment and the themes and trends they were exposed to were much more bullish. So now coupled with decent fundamentals and the technicals of sentiment and themes and trends looking much more attractive, the expected returns started to look quite a bit higher.
A few more months later, however, Ebola became a large concern, and the net effect of Ebola and lower oil prices led to a more modest expectation of airlines going forward. Travel was going to come under pressure, as fewer and fewer people were going to feel comfortable traveling around the world. So we were a little more tempered in our positioning in airlines, because their expected returns were a little more modest. But the Ebola concern only lasted a few weeks and the lower oil prices were the dominant driver of sentiment, so we finished the year off pretty well, with our set of airlines being a reasonable contributor to the portfolio’s expected return.
That’s an example of how reevaluating our investment options almost daily, deciding whether they look attractive or not, aligning portfolios to what investors care about today, and making sure we are buying companies that are fundamentally attractive and that the investment public is excited about, plays in and helps us.
Q: How do you identify turning points?
One way is that for each data set we decide on the right horizon to look at in terms of aggregating what the data are saying. As a simple example, if a news article about a company today is negative, or is different from what it was yesterday, is that enough for us to act? Or do we want to see a more prolonged, consistent, meaningful change in sentiment around a company?
So each data set or signal we create is somewhat optimized to most effectively predict the one-year return. In some cases we look at very small and quick changes in information, and in some cases we look for much longer pervasive changes in what the data is telling us. On each and every data item, it’s a matter of assessing whether a day-over-day or a month-over-month or a trend over three months is useful in predicting one-year return. Generally we find that on the sentiment side we need to look at a three-month change in a particular sentiment data source before it can be useful in predicting one-year return. That decision is made data source by data source.
At the heart of our portfolio construction process we use an optimizer to help us identify which stocks’ expected returns have fallen enough that we can replace them with different stocks, generally in the same industries or in sectors whose expected returns are quite a bit higher—high enough to justify the cost of the trade. For instance, if we owned stock A and bought it when the expected return was 3% and now it’s 1%, and there is stock B in the same industry and sector whose expected return is 5%, we would analyze if that is a good trade to make.
The decision process would mostly revolve around two questions. What is the cost of selling A and buying B, and what are the risks associated with owning B versus owning A? We generally make a set of tradeoffs in our portfolio construction process to rebalance the portfolio once a week. So while we reevaluate and recalculate expected returns of companies daily, it’s not really until every week that we go in and reevaluate what we own and don’t own, and use an optimizer to help us make these complex trade-offs while being as cost-effective and customized as possible.
In the small-cap space, where cost is such a paramount consideration, being cautious and only trading when there is a substantial improvement in expected return net of cost is part of how we structure our portfolio construction approach. Our proprietary expected return on the stock is driven by data that we think captures whether a company is high quality, cheap, has a positive sentiment, or benefits from themes and trends. But ultimately it’s what we think is or isn’t going to do well.
Q: What do you consider a good expected return?
Our expected return calculation estimates are relative to the market. We think it’s sort of indifferent to our process whether the market is up 30% or 3%, or down 10%. Our expected return estimates tend to fall in the range of +/- 10% versus the market.
Q: What is your portfolio construction process?
In our portfolio construction we take modest active weights on industries and sectors; the majority of the return comes from selecting specific stocks, not sectors or industries. So we were not going to go all-in on airlines, but certainly airlines looked a little more attractive than others, and we took some small overweight positions in them in mid-2015.
We also moderate how much of a stock we buy relative to the Russell 2000 Growth Index. We do lots of analysis on lots of companies to make sure we end up with attractive companies in the portfolio. The key is to diversify the portfolio and limit single stock concentration.
The same goes for the sector / industry side. We don’t want any one sector or industry to have a material overweight or underweight of more than 3% versus the benchmark, so we put guard rails on from the risk standpoint and make sure we target a beta of 1 relative to the Russell 2000 Growth Index.
We put these guard rails on risk, single stock concentration, sector or industry exposure, and beta. Then the optimizer also helps us from a risk and concentration standpoint in terms of names to sell or buy as a replacement, factoring in the risk and cost of buying and selling each stock. We also have daily estimates of the bid-ask spreads, commissions, taxes, fees and things like that.
Typically, we end up owning roughly 20% of the benchmark of the Russell 2000 Growth Index. Since the benchmark has around 1,200 stocks in it, about 20% of that would be 250 stocks. Right now we have closer to 300 stocks.
We monitor our portfolios daily and rebalance our small cap growth strategy approximately once a week. We assess if we should reduce or increase our position in stocks to improve the expected return of the portfolio subject to a set of risk considerations, like tracking error. Our strategy runs at a tracking error target of 3.5% relative to the Russell 2000 Growth Index, so we use risk models to quantify the tracking error of the portfolio and make sure it’s within 3.5%. The turnover of the portfolio on an annualized basis ends up being about 125%.
Q: What is your sell discipline?
It is the same discipline by which we buy stocks, because the expected return does not have to be above a particular threshold for us to like it or sell it. Every time we rebalance the portfolio we compare the expected return for a particular stock with another stock out there that is in the same sector, industry and country that has a better expected return, and similar risk properties and cost characteristics. If so, we would prefer to own at least some of that other stock because it’s got a higher expected return and perhaps sell out of a stock whose expected return has fallen.
Generally when a stock’s expected return starts to fall because the fundamentals or the sentiment and themes deteriorating, we start reducing our position proportionately and replacing it, generally with companies in the same industry or sector with higher expected returns. We make an explicit trade-off among return, risk and cost every time we rebalance the portfolio.
Q: How do you define and manage risk?
At the outset we think about operational risk and data risk. A lot of data comes into our investment process, and we have a whole host of integrity and data checks, with a team of people reviewing all the larger exceptions. It starts first and foremost with an operational and data integrity analysis that helps validate the inputs into our model, and ensures we are comfortable with them. We place a lot of effort on creating a framework for doing this.
At the portfolio level, risk has many different facets. We have our own risk models to measure volatility and tracking errors. We further ensure against adverse market events by putting portfolio constraints and guard rails on our investment process, limiting sector, industry or single-stock exposure, and beta. On top of an estimation of risk we have formal, weekly discussions on risks or events in the portfolio we may be missing or not capturing.
The senior portfolio managers meet with the Chief Investment Officers weekly to discuss portfolios and risks, to get a sense of what the other investors in the division are talking about from a risk standpoint.
During geopolitical events like the U.S. Presidential election and Brexit, we did lots of security analysis and other portfolio analyses. So risk for us involves a broad set of analyses; it’s part and parcel of making investment decisions and being cognizant of the unintended risks. We also have an independent risk management team outside of our portfolio management effort that does its own analysis on our portfolio—but for us it’s just making sure our decision process is as informed as possible.
There are always things that may not be fully captured, and we try to protect ourselves from that unknown.
