“The plural of anecdote is not data”
I’ve used this quote to discount the validity of a single observation to explain much of anything. That observation is true. Yet the real quote, attributed to Stanford researcher Ray Wolfinger, is the following:
“the plural of anecdote is data”. Ray Wolfinger
Every data point has a story and sometimes that story can illuminate a larger truth.
I think the anecdote of Tesla’s recent stock surge this year gives us some insight into the Momentum Factor, namely that stocks like Tesla that have already surged in price are emotionally impossible to own for those of us not day trading on Robinhood. And for good reason: they can fall out of favor fast! The potential for these massive price corrections and the psychological toil of owning them helps explain the return premium associated with the Momentum Factor. Tesla’s recent stock volatility and the emotions it triggers not only makes it a worthy poster child for momentum stocks but also provides some insight into the differences across both momentum factors and momentum funds. And dare I add when looking at the price action of similar stocks on November 9th, it is a warning on what may yet still be on the horizon.
November 9th: A Unique Day for Momentum Investors
The traditional academic “momentum” moniker most simply applies to stocks that have performed the best over the past year. These stocks over various time periods and across markets, both in and out of sample, have exhibited excess risk-adjusted returns as well documented on this blog (here is an example) as well as many others(1). Equally well documented is the ability of momentum stocks to drop like a stone. In fact, on November 9th of this year, one of the sharpest price reversals for these stocks in the last couple of decades occurred. Yet the event passed largely unnoticed by the general public (outside of some by the #fintwit community and modest coverage in the Wall Street Journal).
The scatter plot above captures this extreme relative move in the momentum factor on November 9th. The -14.5% return of the Dow Jones U.S. Thematic Market Neutral Momentum Index was almost 15 standard deviations from the norm (2). Bloomberg’s US Pure Momentum Portfolio factor was down a more modest 4% or 4 standard deviations from the norm. Since 1999, the most significant daily drop in price was a meager 7.5% for the Dow’s on March 23rd, 2000, and a 1.7% drop for Bloomberg’s factor on 1/09/09. Hmmm, coming out of the Internet Bubble and Great Financial Crisis the Momentum Factor experienced much more than just a daily reversal. This is a well-documented period of gut-wrenching drawdowns for the factor(3). French’s data goes back much further than the 01/01/2000 start of the other two factor indices and records 5 days exceeding a 10% drop, mostly in the 30s during the Great Depression(4). Again, the largest drawn down in the Momentum Factor’s history also occurred over this time period and the proximity of the large daily reversals should give us pause.
Differences Across Momentum Factors and Funds
And what about the large performance differences across the factors for this date? What first appears to be minor differences in both how and when to calculate the factor has major repercussions for returns. The Momentum Factor attempts to neutralize swings in the market by going long these momentum stocks and short out-of-favor stocks that have done poorly. But different quants calculate that factor differently. Ken French of Dartmouth sorts daily to calculate the Fama-French Daily Momentum Factor and combines both big company and small company portfolios to go long and short momentum. Bloomberg uses an implicit factor model using multivariate cross-sectional regression analysis designed to zero out exposure to other factors, including the market. Dow Jones sorts stocks based on 12 months returns within various sectors in attempt to minimize industry bets. French uses market value weights. To avoid idiosyncratic risk, Bloomberg uses the square root of market value when determining weights. Dow Jones uses equal weighting and rebalances quarterly.
Even momentum definitions can differ. Although momentum enjoys considerable consensus on what it means, namely 6 to 12 months of relative strong performance with an adjustment for well documented short term (up to a month) reversals, this also can lead to differences in returns. French-Fama’s factor uses returns starting 251 days back (roughly 12 months) and stopping 21 days back (roughly 1 month) to avoid short-term reversals. Bloomberg goes back 54 weeks and stops 2 weeks back. Dow Jones just goes back 52 weeks with no adjustment for short-term reversals.
Definitions aside, the differences in returns of the long-short measurements of the Momentum Factor also translates into differences in returns to long-only Exchange Traded Funds (i.e., ETFs) and mutual funds trying to capture the momentum strategy’s excess returns. As shown in the table below, on November 9th most, but not all, of the listed funds dropped with the factor indexes while the U.S. Market, as measured by Vanguard’s Total Stock Market ETF <VTI> was up 1.25%.
Performance differences across funds have been especially stark recently. For example, from when the Momentum Factor began to suffer recently on September 2nd through December 15th, the largest Momentum ETF, iShares MSCI USA Momentum Factor <MTUM> was up only 1.9% whereas the Vanguard U.S. Momentum Factor ETF was down 3.5% while the US market, as measured by Vanguard’s Total Stock Market ETF <VTI> was up 6.5%. These performance differences, although not as glaring, also appear in other periods of poor momentum strategy returns. Details matter.
Tesla and its Ilk
Note in my table, I include the returns of ARK Innovation ETF <ARKK> as well as Tesla and “Jon’s Bubble” stock portfolio. Let me admit the sad back story about why I did.
Back on September 9th I proudly wrote to my clients that Tesla’s then drop in price may reflect a long-overdue correction.
In my article on “Price” from earlier this year, I shared this graph constructed by Morgan Stanley analyst Adam Jonas as a humble admission that the then-current price of Tesla may prove to be way off in hindsight but, like all stock prices, was likely the best guess of value at the time. The future movement of stocks are just too hard to predict. Even after dropping 33% since the end of the month, Tesla’s stock price is still at a split-adjusted $1,651, up nearly 400% from the date of the article and over 3X’s his bull case.
Note my emphasis on past performance as a set-up in explaining why Tesla was overpriced. Ugh. I went on to reassure them that although I was not a stock picker, I stay clear of stocks like Tesla that trade at “crazy” valuations as part of my value bias when constructing stock portfolios. (5) I named Zoom, DocuSign, Peloton, and Square as similarly overpriced stocks despite booming businesses and great products. I then gloated that as a group, those 5 stocks were on average down 17.5% in the previous 3 days and we owned none of them.
Yep, I got that one wrong.
Tesla closed December 18th at $695 or a split-adjusted $3,475, more than double from when wrote my letter and up over 10 times since Adam Jones wrote his article declaring a potential upside of only 50% (and downside of 97%!). The group of 5 stocks which I now call my Bubble Portfolio together were up nearly 20% since my letter, well outperforming the market’s (6) otherwise stellar returns of 6.5% over the same time period ending December 15th.
I picked the 4 additional stocks as examples along with Tesla of well-known success stories that also had little actual earnings and “crazy” valuations. What I didn’t factor in was that I was also creating the perfect (with hindsight!) Momentum Fund. Over the 11 months through October (remember, most measurements of momentum drop the latest month as stocks tend to revert from shorter-term movements(7). My basket of 5 stocks was up 304%(8). Tesla stock itself was up 488%.
As the table above shows, other Momentum Funds have had much smaller percentages of these companies but do have large exposures to Apple, a worthy momentum stock but hardly one of the market’s best-performing stocks. It is the biggest and when the fund also uses the market value as a criterion for position size, you see the effect. (9)
Next, take a look at ARK Innovation ETF <ARKK>, the hot fund for 2020. It first appears to be a momentum fund in disguise. I assume they construct their portfolios more thoughtfully than I did when creating my Bubble Portfolio, but the recent performance patterns are uncanny.
This begs the question, are Momentum Funds innovator funds in disguise? Certainly, not all momentum funds qualify if my 5 stocks are a proxy of any kind. Methodology and definitional differences explain much of the differences between Momentum Factors and Funds, especially the role market cap plays in weightings, but the idiosyncratic timing of rebalancing seems to have an even larger impact, at least recently. Granted, the most popular momentum fund holdings today look like a who’s who list of stay-at-home, order in, and the “hope for a vaccine” stocks: Apple, Amazon, Mirati Therapeutics, and Nvidia. But these stocks will eventually roll out to be replaced by other market favorites. For instance, last September when momentum ETFs took it on the chin as interest rates rose, their portfolios were dominated by REITs and Utilities.
Momentum: An Efficiently Priced Risk Factor or a Behavioral Mispricing?
Again, I want to use Tesla as an antidote to help dissect the long- and short-term success of momentum strategies. The Momentum Factor I believe acts as both efficient compensation for risk and as a tool to profit from common mispricing (gasp, market inefficiencies) due to well-established behavioral heuristics. A better understanding of this dynamic keeps me invested in the strategy even though I believe another correction is coming. My disgust with Tesla’s continued escalation solidifies my awareness that my biases to shy away from these types of stocks are deep and the pain of owning them, even via an ETF, is real.
Although I’m still climbing, the initial steps out of this deep-rooted bias against momentum were also arduous. I’m a product of Gene Fama’s Financial Theory course, TA’d at one time by both Cliff Asness of AQR fame, a firm built with a momentum foundation, and Wes Gray of Alpha Architect, the author of The Book on momentum investing who also hosts this blog. At the time, 1992, Fama was co-chairing Asness’s dissertation on momentum’s unexplained persistence to generate excess risk-adjusted returns(10). He undoubtedly thus wasn’t naïve to the factor. By then, the empirical evidence supporting the outperformance of those stocks that have performed the best in the last 12 months (not including the first month) had been published by Narasimhan Jedadeesh (11). Yet, it didn’t make our reading list. In fact, Jegadeesh (1990) was only barely mentioned in Fama’s 1991 seminal survey of attacks on market efficiency, Efficient Capital Markets: II, which concludes, “…the new research on the predictability of long‐horizon stock returns from past returns is high on drama but short on precision.”
It seems Fama had his biases, too.
But like the factor return itself, the evidence for employing momentum strategies kept compounding. Corey Hoffstein of Newfound Research provides a nice summary of the path of acceptance through both time periods and markets here. Eventually, even Fama and his prolific counterpart Ken French (2017) (12) “somewhat reluctantly” accepted momentum as a worthy factor for explaining the cross-section of stock returns. Although trained to recognized that supposed patterns in stock prices were illusions to those weaker minds seeking order in the reality of randomness (13), eventually (with the help of Wes/Jack…thank you!), I, too, succumbed to the evidence. But it was hard, especially when looking under the hood to see the stocks these momentum ETFs owned. But general acceptance of the data hasn’t translated into an acceptance of the theory behind the data.
Modern Portfolio Theory has evolved in its broader efforts to explain differences in returns across stocks. From the vantage point of an efficient market assumption, Bill Sharpe constructed a world where investors look only for compensation for risk tied to their current portfolios (e.g., CAPM). Robert Merton’s world saw investors willing to hedge their current and future consumption with a variety of factors (e.g., Intertemporal CAPM or ICAPM). And Lu Zhang more recently argued for an investment world where differences in expected return are driven by firms’ decisions to invest (Investment CAPM) instead of investors’ desire to hedge. Momentum’s tendency to crash like it did on November 9th and other periods favor viewing the excess returns generated by the Momentum Factor as fair compensation to disruptions in future consumption a la’ ICAPM. But my experience indicates it is also compensation for my stomach aches and sleepless nights. I’ll wait for someone else to build that fact into a workable pricing model.
Investing in a fund that owns stocks like Tesla let alone owning the stocks outright isn’t easy but why doesn’t the knowledge that these companies provide excess returns act as a tonic? Or why not alleviate the side-effects of momentum investing with Pepto Bismol and sleeping pills? In other words, why do the premium and my stomach aches persist? Behavioral heuristics also seems to have a role.
Anchoring is the tendency of mere mortals to place too much emphasis on the first piece of information we receive. Although I had an unbiased view of Tesla at the start of the year, I saw the analyst Adam Jonas give the stock an upside for 2020 of $500 as reasonable. No surprise that when Tesla reached a split-adjusted $2,000, I was incredulous.
We also place a greater value on avoiding a big loss over capturing a big gain. I can’t imagine buying any of my bubble stocks, including Tesla, right now (14). Our tendency towards loss aversion makes us overly fear the second half of the idiom “pigs get fat and hogs get slaughtered” so much that we never fatten our portfolios.
But hogs many of these stocks become and a subsection of the investor community (again, see Robinhood) seem morbidly obese. “One more thin mint?” The momentum factor also seems to capture the tendency to believe that success is based on skill versus luck and breeds overconfidence and rationalizations like claiming Tesla could be the next Amazon.
Yes, Amazon’s similar valuation was once scoffed at, too. But rightfully so even with hindsight. Amazon shares dropped 94% between December 1999 and September 2001 even though its sales nearly tripled in the ensuing two years. It wasn’t until 2007 and sales had increased over 600% before it regained its stock price from the end of 1999.
In short, Tesla’s stock performance makes me squirm. Back in September I let my clients know that I thought the valuation was crazy. But TSLA doesn’t care what I think and doubled yet again. It takes all the strength than I can muster to invest in stocks like Tesla that have already skyrocketed in price, and even then I outsource the task to momentum ETFs. It also takes a higher expected return to attract investors. So far this year, many, but certainly not all, momentum funds exceeded those high expectations. The devil is in the details on why some versus others have performed well, but the results show up in the holdings: those weighted most toward stocks of companies experience product innovation and growth even in the face of COVID-19 have smoked the market.
Will these stocks likely tumble back to earth in a flame at some point. History tells us yes. And history tells us that November 9th was an almost unprecedented extreme rotation out of momentum and major moves are typically clustered. Three antidotes certainly don’t translate into data, but given that the other large reversals occurred coming out of The Great Depression, The Internet Bubble and The Great Financial Crisis when the momentum factor suffered some of its largest drawdowns does make one pause. Having already feasted plenty from the trough, let’s hope momentum factors and funds will rotate out of the current crop of market darlings this time and into the new hot stocks that I seem equally mentally programmed to want to avoid.
|↑1||see especially the blogs at Newfound Research and AQR|
|↑2||implying this distribution is anything but normal!|
|↑3||See Table 8.5 of the book Quantitative Momentum|
|↑4||As of publishing, French hasn’t updated his data for November|
|↑5||Although direct indexing more easily allows for tax-loss write-offs, I more typically use ETFs and only use individual stocks when trying to manage around legacy assets or professional risk.|
|↑6||well, Vanguard’s definition of the market, the ETF (VTI)|
|↑7||see Asness, C. (1994), “Variables that Explain Stock Returns,” Ph.D. Dissertation, University of Chicago. An adapted and extended version of this paper can be found at AQR here.|
|↑8||rebalanced on 12/31/2019 and 09/02/2020|
|↑9||Note that SPDRs S&P 1500 Momentum Tilt ETF as S&P requires a certain amount of profitability prior to even entering the universe for consideration. Tesla only just qualified.|
|↑10||Asness, C. (1994), “Variables that Explain Stock Returns,” Ph.D. Dissertation, University of Chicago. An adapted and extended version of this paper can be found at AQR here.|
|↑11||Jegadeesh, Narasimhan, 1990, Evidence of Predictable Behavior of Security Returns, The Journal of Finance, 45 (July), 881-898|
|↑12||Fama, Eugene F., and Kenneth R. French, 2017, Choosing Factors, forthcoming, Journal of Financial Economics.|
|↑13||see A Random Walk Down Wall Street, written by Burton Gordon Malkiel|
|↑14||In fact I did when I outsourced the task to my Momentum ETFs|