Year: 2020

In Search of Mindful Compounders

taovalue Business, Investment Process , , , , , , , , ,

Note: It’s a republish of the General & Market Commentary section of 2020 Q3 Letter, for easier reference and access.

This pursuit started from studying my own largest omission mistakes (i.e. failed to buy the companies when I could have), namely Amazon, Microsoft (post Satya Nadella), Netflix. But before we dive in, I need to clarify what “mindful compounder” means. I see “compounder” as who persistently create value (not monetary value alone) for all stakeholders across the value chain, and “mindful” is a generalization of many traits I look for Tao & Commander factor, e.g. mission driven, purposeful, rationality, deep thinking & high awareness. (see more in my investment process post[1]). In all cases, I have identified the managers of these companies as mindful leaders yet didn’t buy for the fear of market being already efficient in pricing these mega cap stocks.

But how efficient has Mr. Market been in pricing the value of these companies really? Below I draw the time series of valuation (TTM P/S ratio), and annualized compounding return to date (i.e. if holding form that point of time to 9/30/2020). I excluded annualized return for recent 2 years since they are not representative for the short history.

One clear lesson to be learn here is that market has been far from efficient in evaluating the intrinsic value of these companies. Taking Amazon for example, the worst time to buy it was in January 2000 when it was priced at 17 P/S ratio, but you would still achieve 20+% annual compounding return if you hold till now. Buying it any other time later than 2004 and hold it till now would give well above 30% annual return! Obviously, valuation still matters, but the worst scenario of 20% is far cry from an efficient market return over 20 years. So, the market must have missed something! Luckily, all three companies have long history for me to study patterns that can help identify value earlier.

The “Stumbled-Upon” Quantum Leaps

The first important pattern is that none of them relied on a single product/service to achieve the substantial value compounding. As old businesses mature, all these companies in their lives experienced some pivotal “quantum leaps” expanding to new products/services, which ultimately contribute heavily to its long-term value but was not fully appreciated by the time. With few exceptions, the leaders later would admit that the leap was not by design, or at least that they didn’t foresee the full potential. For example, Netflix had two major pivots from DVD renting, to streaming & to original content. Microsoft also made a major pivot after Nadella took the helm to “cloud & open source first”. Amazon’s AWS was an internal tool built out of its own frustration of its ability to launch new projects/applications, later became a $40 billion annual run rate revenue monster!

Surrogation Bias & First Principle

All companies try to “leap” too, but more often they end with disappointment. I think what makes mentioned 3 compounders more successful in making such leaps is their mindfulness. Case studies of the failed attempts point to a pervasive, yet understudied bias – Surrogation, whereby the measure of a construct of interest evolve to replace the construct itself. Take Wells Fargo’s fake account scandal as example[2], the management initially use cross selling metrics to measure its relationship with clients. However overtime, subsequent executives started to believe the cross selling is the strategical goal (rather than client relationships), leading to the now-infamous mantra – “Eight is great” (to have 8 Wells Fargo products per customer). What made things worse is management started to tie incentives to this single metric for front line bankers, which ultimately led to faking accounts. Such misalignment is detrimental for company value.

On the other hand, all three mentioned companies deeply embed first principle in their operation to overcome surrogation. Amazon provides probably the best example, from Bezos’ first shareholder letter, Amazon has been obsessively following a “customer-centric” philosophy and repeatedly use this qualitative principle to guide quantitative measures & decision making. For example, Amazon started very early to not only use GMV & MAU to define their performance, but also added repeating customer orders because it was a better way to measure the stickiness, or value created for customers. Pinduoduo (one of our current holdings) is another example in putting a qualitative Polaris for Social E-Commerce (To focus on user engagement) around any measurements. Chairman Colin Huang in many occasions dismissed sell side analysts’ question regarding its ARPU trend & guidance saying:

… raising ARPU is not part of our management’s KPI, but I think it will be a natural result as the users’ engagement increases over time…

– Colin Huang, PDD 2020 Q1 Earnings Call

Also, regarding the measurement of users’ engagement, it uses a unique and well-thought-through metric – MAU/Annual Active Buyers. It is like Amazon’s idea of repeating customers orders, but more consistent in a normalized % form.

Radical & Prescient Decision Making

Another common theme I observed from mentioned companies is that on top of the first principle thinking, management tend to make seemingly radical decisions later proven prescient. Microsoft’s Satya Nadella made an impressive example[3], especially considering his soft-speaking personality (e.g. he’s never seen “upset, raising voices or firing off angry email” by colleagues). Yet he demonstrated assertion in many big decisions even early in his tenure as CEO. For example, Nadella decided to write off $7.6 billion from Nokia purchase first year in 2015 and to terminate its Windows division, which was split into Azure & Office divisions in 2016. Additionally, during the agonizing Windows to Azure reorganization (which one executive called “pulling fingernails”), Nadella showed his exceptional ability to make aggressive changes with little drama.

Netflix’s Reed Hastings, a half-mentor of Nadella (as board of Microsoft), is a hallmark of such decision-making ability[4]. Although the decision of pivoting to streaming & original content are both monumental, I think the most radical & prescient decision may be the one he made in earlier year to call off its streaming hardware right before its launch. I was December 2007 and Netflix has been exploring variety of new business models as its legacy DVD renting line matures. A team of about 20 had been working around the clock for years on a project coded “Griffin”. It was so close to the launch that marketing materials had been printed, advertisements were being shot, and Foxconn, the manufacturing partner, was ready to kick off production. Yet to the surprise of all the insiders, Hastings decided to kill it (subsequently it was spun off and became Roku). The magnitude of this decision is now much more clear, had Hastings chosen to pivot to the hardware business, they would compete with all other hardware players (TVs, Apple etc.) and wouldn’t be able to consolidate the demand side which is a conner stone of its later licensing streaming flying wheel. Roku, even at today’s steep valuation, would be worth 1/10 of the current Netflix. According to one inside source, Hastings explained this hard decision in a very simple but vivid way:

“I want to be able to call Steve Jobs and talk to him about putting Netflix on Apple TV, but if I’m making my own hardware, Steve’s not going to take my call.”

[1] Tao Value Investment Process: https://taovalue.net/2017/07/17/sun-tzus-five-factors-business-analysis-framework/

[2] Don’t Let Metrics Undermine Your Business, Harvard Business Review, https://hbr.org/2019/09/dont-let-metrics-undermine-your-business

[3] The Most Valuable Company (for Now) Is Having a Nadellaissance https://www.bloomberg.com/news/features/2019-05-02/satya-nadella-remade-microsoft-as-world-s-most-valuable-company

[4] Inside Netflix’s Project Griffin: The Forgotten History Of Roku Under Reed Hastings https://www.fastcompany.com/3004709/inside-netflixs-project-griffin-forgotten-history-roku-under-reed-hastings

Performance of “Rebalancing” during the GFC

taovalue Quantamental , ,

Note: it’s separate post of same content from my 2020 Q1 letter. I attempted to use quantitative tools to review how different types of hypothetical “rebalancing” strategy would perform during the last market crisis.

It also relevant to revisit it as we may be thinking about whether we are at the same stage of June 2008 or March 2009.

During any crisis, investors like to talk about “rebalance”, value investors think it’s the time to sell names that dropped less and double down on the overreacted ones (let’s call it “reversion” rebalance), whereas growth investors think one should dump low quality which typically plummeted badly, for high quality ones which typically withhold better (let’s call it “momentum” rebalance).

To test which one worked better historically, I decided to backtest a 3-month momentum factor during the Great Financial Crisis (GFC). The basic idea can be explained as: I imagine different investors started to react at various points of time during the GFC and decided to rebalance based on past 3 months return. On Russell 3000 stock universe, we will slice all stocks up into 10 pieces (i.e. deciles) by ranking their past 3-months return, I ranked return ascendingly, so that decile 1 cam be defined as the losers (returned the worst), & decile 10 the winners (returned the best). Below is a snapshot of each decile’s forward 3 months return from various “reaction” time. For example, if an investor decided to buy the worst performing 10% stocks in past 3 months leading to 9/1/2008, he or she will suffer -41.97% in the next 3 months (from 9/1/2008 to 11/30/2008).

I see some noticeable observations:

  1. In the early part of the crisis, (from 200706 to 200712), losers kept losing and winners held up;
  2. During the first hit (from 200712 to 200803), names got hammered similarly across the row;
  3. From 200806 to 200809, the “reversion” rebalance immensely outperformed (on a spread of +38% 3 month forward return). Also note that D1 generated +14% (recover of overacted losers) & D10 suffered -24% (delayed punish for under-reacted winners)
  4. The following 3 month (200809 to 200812) witnessed the entire financial system melting down catalyzed by Lehman Brothers’ bankruptcy. We saw dark red (negative) returns across the row, and past losers were punished much worse than past winners (-42% D1 return vs. -19% D10 return)
  5. Starting from March 2009 (the trough), the market started to recover strongly. Surviving past losers started to outperform the less losers (D1 returned 64% in following 3 months), but not until it flipped in next 3 months.

Furthermore, at each point of time, a “reversion” rebalance could be constructed by buying the losers (D1) & shorting the winners (D10), vice versa for a “momentum” rebalance (buy D10 & short D1). If we calculate the spread between D1 & D10 (D1-D10 for “reversion” rebalance, and D10-D1 for “momentum” rebalance), we can see on cumulative basis, how each strategy may or may not add value for the most important task in a crisis – preserve capital. Below chart tells a very interesting story.

Note that the bar is the spread for “reversion” rebalance (i.e. D1 – D10), thus “momentum” rebalance would be simply the flip signed of such value. “Reversion” style tended to underperform in higher frequency (i.e. more occurrences), but when it reversed it tends to reverse drastically. This can be seen twice during 3 months after June 2008 (the bull trap) & March 2009 (the rebound after trough), which basically wiped out accumulated value from “momentum” rebalance completely. While you might think winning for one is losing for the other, both strategies (remind here they are market neutral) ended in red ($0.8 for “value” & $0.87 for “growth”). I also added the Russell 3000 index returns during same period as a reference of “do nothing” strategy, which ended at $0.8, on par with the “reversion” rebalancing strategy.

This study brought some very important insights for making decision during a crisis:

  • Both rebalancing strategies work, yet in different phases of a crisis. Timing is crucial (as doing either way at the wrong time would heavily impair one’s capital), if you decide to rebalance. So think carefully whether you have good market timing skill before you act.
  • Neither strategy could consistently add value for preserving capital purpose. I think it’s important that buy & sell decisions should still be made on absolute value vs. price basis, rather than on relative price movement basis.
  • “Do nothing” is not as bad as you might think, especially when you know you can’t consistently time the market.

Using Clustering Neutralization to Improve Value Strategy

taovalue Quantamental , ,

Note: This is a republication of a full write-up of the machine learning research to understand and improve systematic value strategy, which was originally shared in my 2019 Q4 letter [Link]

I’m also start trying Medium. It is published on Medium too. [Link]

 

This article is a collection of thoughts and experiments of my personal endeavor from a quantitative perspective to understand why “Value” as an investing style didn’t perform compared to other styles for an extended long period of time. It consists of mainly two parts: 1) review of others’ relevant researches & 2) an experiment of using Clustering to explore reasons for such phenomenon and potential ways to improve systematic value strategies.

“Value” means differently for different investors. It is important to note that “Value” in context of this write-up mainly refers to the “value” factor, e.g. Fama-French HML factor or variations of it. In other words, it is a systematic long-short strategy longing a group of the cheapest stocks and shorting a group of the most expensive stocks.

My main hypothesis is that GAAP/IFRS currently does not do a as well job in representing fairly companies’ value as historically. Thus “Value” may not work on broad stock universe yet may still work in certain types of stocks. To classify stocks into such “types”, I tried to use Clustering, an unsupervised machine learning technique, to test if it works.

You can also download the pdf here:  [Link]

 

Continue reading →