# Quants – General Financial Models Are the Backbones

The classic financial theory Capital Asset Pricing Model (CAPM) and Arbitrary Pricing Theory (APT) are based upon on the mean-variance analysis, which can be cited as an early and significant example of applying intensive math computation into market analysis. This method is simple. Every student can easily replicate on an Excel spreadsheet, but the concept derived from this analysis is profound. Up to today, ‘diversification’ becomes a deep-embedded principle for almost every institutional investor due to this research approach and conclusion.

the following chart depicts the efficient frontier – best risk/return- formed by a various combination of an equity and T-bond.

CAPM only uses one factor – market, which was challenged by Eugene Fama and Kenneth French. The two introduced and also tested out two additional factors: (i) stocks of small caps and (ii) stocks with a low Price-to-Book ratio.  Note Rmt represents the market return, SMBt stands for the portfolio return composed of longing small cap and shorting big cap groups, HMLt stands for the portfolio return composed of longing high book to market ratio and shorting low book to market ratio groups (B/P, also interpreted as value stocks minus growth stocks). The equation of French Fama’s three-factor model is as below:

In 1997, Carhart added the fourth factor – momentum (MOM factor). Momentum, by its name, means stocks, like any substance in the universe, tends to follow its prior motion as well as the speed, therefore, this factor is also called trend factor or time series factor. In Carhart’s model, he constructs the MOM portfolio by longing the stocks that have last 12-month top-ranked and shorting the stocks that have last 12-month low-ranked.

From empirical research, the sum of these four factors is able to explain over 90% of stock’s return. If we deem the rest term – epsilon as just the idiosyncratic factor, which explains the entire returns, then there is not much for the quants to do, the investment world should be trumped by fundamental guys who are able to scour through every individual company’s profile to find out that idiosyncratic advantage.

But actually, since the emerging in the 1980s, the quants are not only growing but prospering, indicating there are more factors not fully explored, attributing to the part of alpha prefixed in the factor equation.  The more unexplored factors, the bigger alpha part, a.k.a extra returns other than known factors.

Hence, the overarching goal for typical quants is to find out these “more unexplored factors”, so they are always on the look-out for ideas, once they identify economical meaningful ideas, they will do the concept-of-proof testing, in their jargon, back-testing. As is explained, for any potential alpha factors, no matter how many back-testings have been conducted, a final step to run them against the above Fama French Cahart four- factor model is a must-to-do, aiming to ensure the new factor is not subsumed under the four known variables, hence, it’s a true ‘anomaly” for alpha-seeking.

Naturally, according to Efficient Market Theory (EMT), the market, which is made of a vast number of rational and analytical investors tends to figure out any arbitrary trading opportunities, and move the price to its equilibrium level, reflecting the real core value of stocks. Therefore, unless a quant or quant firm can keep the secret sauce to her own forever like Coke Company protects its recipe, other competitors will flock in to emulate, resulting in the dissipation of the “new-found” alpha factor.

Efficient Market Theory is, of course, not entirely true in the real world, meaning smart investors can and is evidenced, that they continuously beat the market. There are two major reasons justifying this phenomenon.

Information advantage can be broken down into three features: timeliness, comprehensiveness, and quality or accuracy of information.

Timeliness is of critical importance as the market is constantly changing; significant news breaking out can move the stock price profoundly in a fashion of seconds. An extreme example on “timeliness” is high-frequency traders, who compete on setting their fiber coil to the closest to the exchange places, so they get a milo-second edge over the rest to “fish” price quote/volume, and their algo basically make nano-profits  on each trade, piling up to a handsome great number, what’s greater, is that it’s near risk-free return. Other examples such as momentum guys trading purely based on price movement, they need the most current intra-day price feed.

The importance of Comprehensiveness can be illustrated with FactSet Revere’s relationship dataset. The dataset itself is straightforward. It maps a network of companies categorized into suppliers, customers, competitors, and partners. The source of data is all public, mainly composed of regulatory filings, press releases, conference scripts, and companies’ websites. The caveat is what we call “reverse engine” in bringing out hidden information, so users have a full picture of all relationships rather than the small tip of an iceberg. For instance, combing through all available public documents of Samsung’s only finds you 8 suppliers, while this relationship data gives 349 company names that are also supplying Samsung. These additional 341 companies disclose Samsung as their customers, along with revenue dependencies in percentage, rendering themselves to be “reverse’ suppliers to Samsung, as relationship, by its nature, is transitive.

Quality or accuracy of data goes beyond just being verbatim loyal to original disclosure because the originally reported data are always disparate, messy and hard to use. So a quality dataset means a normalized or standardized system to organize the data. For instance, Yum Brands reported in its 2014 10k filing that \$6,934 million came from China, \$141 million from India, and the rest from KFC, Pizza Hut, and Taco Bell divisions, verbatim scrub of these data leads to nowhere or even mistaken conclusion. Because in the footnotes, the filing further explains that China only refers to Mainland China, India is composed of Sri Lanka, Bangladesh, Nepal, and India, while a combination of KFC, Pizza Hut and Taco Bell equates the U.S. Imagine every and each company reports in such a disparate way, it’s difficult to compare them apple to apple. A similar example is revenues of products/services or business segment. Companies brand-name their products in whichever way they like, say, Apple Inc. produce iPhones, Samsung manufactures Galaxy, and Google launched Android, only when they are all standardized into one common sector – smartphone, it’s possible for a sound analysis surrounding them. The Revere Georev and SegRev set forth to first create a sophisticated, granular taxonomy, then normalize geographic or business segments revenue data into these taxonomies, resulting in a high-quality, easy-to-use data that help quants build up unique alpha factors.

with respect to analytical advantage, let’s assume that two quant firms were given accessibility of the same set of data, but after a testing period, one comes back with alpha discovery while the other frustrated on nothing being found.  This phenomenon is what I attribute to as “analytical advantage” – the ability to see things through, to discover the unique signal requiring the core intellectual competency of the quants.

It calls for a lot of experience, intuition, and skill set to not only think unique and different but also, in many cases to be able to modify and integrate existing factors for an enhanced alpha.

For instance, with the same data – Revere relationship data at hand, applying a customer momentum strategy is a splendid new idea ten years ago, when Professor Cohen and Frazzini first proposed and examined. In recent years, the academic attention shifts toward computing ‘centrality score’, grandly different than the old ‘customer momentum’ mindset.  However, quite a few quants are still dwelling on this old ‘customer momentum’ strategy, with modified usage, say, adding a layer of news sentiment data on top of supply chain data. They also garnered good excess returns.

This site uses Akismet to reduce spam. Learn how your comment data is processed.