Every investment manager needs to hedge their portfolio against possible adverse stock price movements (especially with so much volatility these days!). There are a myriad of possible ways an asset manager could hedge their portfolio (through derivatives, through ETFs, even through individual securities), but they all involve a risk reward tradeoff. A perfect hedge – one that has 100% inverse correlation to the initial position – is the ideal, a security or basket of securities that eliminates all market risk. Of course, with ever shifting capital markets, the perfect hedge is almost impossible to attain. Still, we know that forward thinking asset managers can turn to big data and predictive artificial intelligence to inch closer to that goal.
To that end, we have added new functionality in our cloud-based AI platform for institutional investors – Boosted Insights – called Similar Stocks. Similar Stocks is a tool that allows asset managers to identify and evaluate securities in their universe that move together or in the opposite direction. Using these insights, asset managers can generate hedge, replication and custom portfolios, all from one screen.
We show how using machine learning to hedge stocks can improve results over typical correlation hedging by as much as 25 percentage points.
How it works
The standard approach most investment managers use to hedge takes only historical pricing data into account. Boosted Insights takes all the inputs the user adds to their AI-powered model portfolio and determines factor values. This enables Boosted Insights to better identify factors that contribute to movements in securities and draw parallels to determine similarities and differences between them.
The Similar Stocks screen is divided into a two step process:
Security Selection and Exploration
Here, the investment manager picks a stock (or basket of stocks) for comparison. They see a data-rich view, where they can pick the target factors and view their scores or view the differences in scores from the target stock (or basket of stocks). Additional functionality allows the user to filter and sort to hone in on the exact relationships they want to identify. Price Similarity looks only at price-based relationships between stocks (or a basket of stocks) and Data Similarity takes into account all model variables between stocks (or a basket of stocks). Overall Similarity combines the two to bifurcate the analysis.
Now that the investment manager can see which stocks are similar to their targeted stock (or basket of stocks), they can create a portfolio from these similar stocks. We have added a Basket Trading module to the portfolio settings in Boosted Insights that allows asset managers to customize their AI enhanced portfolios even further. There are selectors for price weight, data weight, factor weight and more, all so the asset manager can create a highly custom basket of stocks that work for their trading mandate.
Case study: Single name stock hedging
Let’s look at a case study, using the Russell 1000 as a baseline. We created two portfolios to hedge out exposure to AAPL. One is built purely on correlation basis and the other uses machine learning derived similarities to construct the portfolio.
Figure 1: Correlation portfolio
Figure 2: ML hedging portfolio
The charts above show the total return, with the black line representing AAPL and the orange line representing the total portfolio – long target security (here, AAPL) + short hedge basket. As shown, both baskets adequately hedge out target exposures. However, the ML hedging portfolio adds additional benefit by actively using ML signals to select stocks that offset the target security while sacrificing less return.
In both cases, portfolio volatility remained the same, both just under 30% standard deviation. However, the ML hedging portfolio had annualized return 15 percentage points higher than the correlation portfolio. Similar Stocks allows users to use big data and AI to find opportunities that are more performant than standard correlation hedging.
Case study: Basket hedging
In this second case study, we created two portfolios to hedge out exposure to FAANG.
Figure 5: FAANG correlation portfolio
Figure 6: FAANG machine learning hedging portfolio
As before, the charts above showcase the total return, but here, the black line represents our FAANG basket and the orange line shows the total portfolio – long FAANG basket + short hedge basket. Again, both baskets adequately hedge out the target exposure. However, the asset manager wins out by using machine learning to hedge their basket, because it selects stocks that offset the target basket while sacrificing less return.
Here, the volatility of both portfolios was just under 20%, but the annualized return of the machine learning hedging basket (using Boosted Insight’s Similar Stocks feature) was 25.32 percentage points higher than the typical correlation hedge.
Similar Stocks is a powerful tool for asset managers to capitalize on AI and big data to hedge their portfolios. It enables users not only to view similar / different securities based on their correlations, data parameters and ML factors, but also to build multi-purpose portfolios. Similar Stocks is a simple use case, both for investment managers looking to implement AI and for institutional investors that are already using AI, to drive success for their portfolios. We have also added the ability to use these tools against an investment manager’s entire portfolio, which we will elaborate on in a future post. To learn more about how Similar Stocks can help you create data-driven smart hedges, please reach out to us here.