Family Resemblance

Around the holidays, we often think of family. According to Wittgenstein, we can understand the concept of family through the following analogy: individual strands that bind together to make a rope. No one strand defines that rope, but together they make a cohesive unit. Families often resemble by looks, mannerism, and habits. At Novus, we’re obsessed with analyzing fund managers, and one way we do this is by understanding their families.

What do we mean by fund families? Well, you often see aspiring fund managers cut their teeth with experienced firms as they are early in their career. (The virtues of this apprenticeship model were recently described by Jennifer Oppold on our podcast.) They learn from methods, approaches, and philosophies at these firms that have a lasting imprint with how they ultimately manage capital on their own. These traits come to define a fund family just as they would one’s parents and siblings. How can we say that for certain? The data comprising ownership of single-name securities.

We have published at length about one particular family tree, the Tiger Family. This is perhaps the most notorious family tree across hedge fund managers, with several legendary investors having their roots tied to Julian Robertson’s eponymous firm. Over the years, we created tooling such as Overlap and Contagion to understand the relationship across fund families. Now we’ve partnered with academic researchers to better understand just how important family is to investment behavior.

‍

Family Resemblance Across Hedge Funds

Researchers Nimesh Patel and Tray Spilker from the University of Hawaii have found that unanimous trades that happen across a family of fund managers can point to significant alpha. In the order of 7% per year, after stripping out most common risk factors. That is a principal finding in their working paper “Alpha by Affiliation”.  Patel and Spilker were able to analyze family trees across fund managers by leveraging the fund-specific tagging that Novus created across all 1,500+ managers in our Hedge Fund Universe (“HFU”).

"Our published research on resemblance has historically focused on Tiger Cubs, but there are dozens of other family trees, several of which are growing in prominence."

Taking a step back, Novus systematically collects this data across every fund manager firm who files a 13-F with the SEC. When we add a fund manager to our HFU, we track some qualitative characteristics about that manager.  First is their principal strategy (are they a long/short fund, or a distressed fund? Quant or Multi-Strat?), and second is their family resemblance. Our published research on resemblance has historically focused on Tiger Cubs, but there are dozens of other family trees, several of which are growing in prominence. Patel and Spilker identify over 100 family trees subsuming 1/5th of our aggregate HFU. Trees are generally small, averaging 3 managers, but they can expand up to 30 funds for one degree of family depth (e.g., Tiger)!  The following table provides some summary statistics about these fund families over time:

‍

What is particularly interesting is the single-security overlap we witness across these funds. Funds average nearly 21 securities of overlap across their family tree. That cross ownership averages over 30% of their reported portfolios! That’s an astounding degree of common ownership, and a point we’ve highlighted for clients in the past. Why does this common ownership occur? One theory is that similarity in training leads to similar ways of analyzing the market and thus similar investment theses. Others talk about idea cross-pollination as a function of familiarity. Perhaps some careful analysis can tease this out.

Peculiar Trades: Unanimous Change

To really drill down on this, Patel and Spilker focus on peculiar trades across family trees: when in a single quarter every manager in a family unanimously buys or sells out of a security, as per their 13F. They call this “unanimous change.”  With a tree of a handful or more of managers, this is a very unlikely occurrence unless they either have very similar ways of thinking about securities, or if they are more than likely sharing information.

By defining unanimous change, these researchers tested how these unanimous change stocks behave in terms of their ability to predict a cross-section of stock returns. Not only do they test the return characteristics of going long on unanimous buys versus short unanimous sells, they also measure a significant price of risk associated with these stocks after known factors (such as style factors, or liquidity) are considered. What they find is a very powerful signal. Over nearly 17 years of analysis, they find a signal that is positive in a market neutral construct in the majority of years. Only recently with some bumpy years for hedge funds (like 2015) do we see negative performance for this signal.

‍

‍

What’s particularly interesting about this signal is how the majority of the abnormal return occurs 2 months after the co-occurrence is registered. While that means that some type of information is being collectively absorbed by these fund managers and thus represented in their portfolios, it unfortunately means that much of it disappears before it becomes available from a 13-F release.  

Nevertheless, this is an important finding for both sides of the GP/LP equation.  For fund managers, they need to be aware where certain pockets of the market are skating toward, and Novus provides this type of custom overlap analysis for every hedge fund in our Hedge Fund Universe the day after 13Fs are due, providing the closest possible advantage to clients who want to absorb this information as fast as possible. For allocators, understanding the family resemblance of their managers or the siblings of their managers is crucial in understanding the competitive edge of their pedigree as well as the true differentiators of their underlying managers. We’re excited to be a part of such compelling research and to continue to provide these valuable insights to our Novus Public Ownership clients.

View "Alpha by Association" here.

‍