I use Google search every day for work and leisure, and it rarely disappoints. The algorithm powering this search engine is amazingly effective at finding relatedness between the searched topics and relevant online content. Almost all digital content is now being classified, analyzed, and surfaced to users based on this concept of relatedness. Enter a movie into your favorite streaming app and it recommends related movies or enter a song into your favorite music app and it recommends related songs or albums. The list goes on and on.
What if thematic investing happened in the same way? While passive investing through transparent, rule-based indices has gone mainstream via ETFs—providing exposure to broad markets, sectors, and investment factors—thematic index creation remains a challenge.
Up until now, most thematic indices have been built by a committee of experts using relatively opaque rules. Based on their knowledge and experience, the committee decides which stocks to include and exclude, often with wide discretion. Seldom would investors be able to find specific, quantifiable selection rules in the methodologies of these thematic indices. Come index rebalancing time (which often lasts no more than two to four weeks), this committee is supposed to sift through thousands of companies to identify which companies to add, remove, or keep. It always begs the question of how complete or objective this selection process really is.
Earlier I mentioned the ubiquity and power of applying “relatedness” to digital content and mused about doing this for thematic investing. The music app, Pandora, stands out as an inspiration. Pandora uses a team of music experts to listen to and classify songs according to a proprietary music classification system; the company describes this as the “Music Genome Project.” The result is a large, dynamic music database that captures each song’s unique set of attributes. A user simply needs to enter the names of a few favorite songs, and Pandora will recommend related songs and albums.
Our idea of Thematic Indexing 2.0 follows this logic, where individual stocks are equivalent to individual songs and thematic stock baskets are equivalent to music albums. All we need is a “Stock Genome” database that captures each company’s unique set of granular business attributes—what products and services they provide and how they are connected in the supply chain.
Let’s run through an example of how to build a thematic basket systematically using this “relatedness” concept. Step one is to brainstorm themes. Ideas could originate from sources such as newspapers, magazines, earnings transcripts, industry whitepapers, and more. Step two is to narrow down to a single investment theme. Step three is to identify several companies that best fit this theme such as Amazon Inc. for eCommerce or Netflix for online streaming. Step four is to generate an expanded universe of related companies based on the initial short list of companies. One way to accomplish this “relatedness” search is to use a supply chain database and find each company’s suppliers and customers. This exercise could grow our initial list of just a few companies to several hundred. The expanded list of companies can then be decomposed into a granular industry classification to arrive at a handful of relevant industry groups. These industry groups will form the basis of our initial thematic selection criteria.
To make things more tangible, let’s go through an actual example. For idea generation, we systematically scanned company transcripts and Google Trends and ranked the popularity of diverse topics. In doing this, Electric Vehicles (or EVs) surfaced as a top-ranked theme. We then identified Tesla, Nvidia, and BYD as core EV-related companies. Entering these three companies into FactSet’s supply chain database generates a list of several hundred related companies, which can be further decomposed into their granular sub-industries such as electric vehicle manufacturing, graphic processing semiconductor, and high-end batteries manufacturing. Using these three sub-industries, we then proceed to construct an EV thematic basket by screening for companies around the world tagged with these classifications. The diagram below provides a visual representation of this process.
The process described above is both disruptive and democratizing. Using this indexing method, companies anywhere in the world could be discovered and subsequently included in a thematic index if they are developing products related to the selected investment theme. These companies don’t need to be followed by large brokers, spend heavily on marketing, or be endorsed by large buy-side institutions. Isn’t this why Google is so widely adopted when it comes to searching? If your website produces the most relevant content for the keywords being searched, you will be discovered no matter who, where, or how big or small you are.
The investment management industry is going through a seismic change and many have characterized it as a shift from active to passive. But underneath, there exists another derivative shift from bottom-up, fundamental active investing driven primarily by human expertise, to top-down, systematic investing driven primarily by big data, smart data, and (increasingly) artificial intelligence. Thematic Indexing 2.0 is a manifestation of this derivative shift.