Woof! for descriptive statistics

Gary Smith points to this news article in the Economist, “WOOF, CAKE, BOOM: stocks with catchy tickers beat the market,” which reports:

In a study published in 2009 by Gary Smith, Alex Head and Julia Wilson of Pomona College in California, a group of people were asked to pick American public companies with “clever” tickers. The resulting list of 82 shares included MOO for United Stockyards, a livestock company, GEEK for Internet America, a service provider, and SPUD for 1 Potato 2, a restaurant chain.

The authors found that a dollar invested in their portfolio of cleverly named ticker symbols in 1984 would have been worth $104 in 2006, an annual return of 23.5%. The return for all shares in the New York Stock Exchange and Nasdaq, the United States’ two largest exchanges, during this period was a mere 12% per year. Traders, the authors argued, are unwittingly drawn to companies with memorable tickers, which increases the demand for their shares and drives up their value.

Unless I’m missing something, I don’t think that last statement is an accurate description of the paper. I didn’t notice anything in the Smith/Head/Wilson article saying anything about traders doing this unwittingly.

One thing I really like about their paper is that it’s unabashedly descriptive:

– From the abstract:

This paper investigates the performance of stocks with memorable ticker symbols during the years 1984–2005 and finds that, on average, their daily returns are higher than for the overall market.

– From the conclusion:

In recent years, a substantial number of companies have chosen clever ticker symbols. On average, these stocks have outperformed the market by a substantial and statistically persuasive margin. We do not know why these stocks have done so well. Perhaps a clever ticker symbol has been a useful indicator of the managers’ ability—ability that revealed itself over time as the company repeatedly exceeded investors’ expectations. Or perhaps a clever ticker matters because it is memorable and has a subtle, but persistent, influence on investors who buy the stock and on those who are considering a merger or acquisition. . . .

Openly descriptive, and without strong causal claims.

The Economist article continues:

But might the results of the study simply have been a fluke? A decade on, Mr Smith and two new co-authors, Naomi Baer and Erica Barry, have repeated the analysis with 22 of the original 82 companies that were still trading in 2006. (Only 12 of the 82 firms went bust; most ceased trading for other reasons, such as buyouts, mergers, or delisting.)

Once again, the portfolio containing tickers such as CAKE (for Cheesecake Factory, a dessert chain) and WOOF (for VCA Antech, which offers veterinary services) outperformed. This group of companies earned 13.2% per year between 2006 and 2018, compared with an annual return of 4.9% for the broader market. Mr Smith also reviewed the performance of a second portfolio composed of newer firms with clever tickers such as PZZA (Papa John’s Pizza) and WIFI (Boingo Wireless). This group managed an annual return of 11.3%. . . .

It’s good to see people follow up on their research.

Look—I don’t want to make too big a deal of this. It’s just one little study. But we spend enough time having to deal with scientific hype of various sorts, that it’s good to see some clean description and replication.

36 thoughts on “Woof! for descriptive statistics

  1. Should we even expect a study to replicate if it presents evidence that stocks with some characteristic do better than average? If it was a fluke, it won’t replicate. But if traders believe it’s a real association, it should be priced in and vanish.

  2. Re the “unwitting” part. Surely the investors are not unwitting here. But there is a related phenomenon in which stocks of companies with ticker symbols which can be plausibly confused with the names of other companies rise and fall with news about the companies that they are not actually related to. This happens quite a bit, most recently with Zoom Video Communications Inc (ticker symbol ZM) which has caused major moves in the unrelated stock Zoom Technologies (ticker symbol ZOOM) https://seekingalpha.com/news/3546608-zoom-technologiesplus-52-in-ticker-confusion

    • “Surely the investors are not unwitting here.”

      Yes I agree, the “unwitting” explanation is almost certainly wrong for periods of 20+ years! Strong movement over that time has to be related to the performance of the company.

  3. Here’s the likely explanation of the phenomena:

    “In recent years, a substantial number of companies have chosen clever ticker symbols. ” (Head et al 2009, first sentence of conclusion)

    The data set probably has a very strong bias toward new companies, which are created to exploit new market niches, which is the explanation of why the market rises over time in the first place. What would be really interesting is to see the effect of company age on stock price.

    Also I suspect there’s an acquisition effect related to two market peaks (1999, 2007). Acquisitions occurring in 1998 and 2006 would in effect “fix” the peak market price for the company even though if it had continued trading on its own, the price would have plummeted. (I browsed the paper but I didn’t see how they handle acquisitions, maybe they have some method of accounting for that or just ignore acquisitions?)

  4. The portfolio has sector exposure. Funnily enough that’s the one explanation they don’t investigate quantitatively, they just count industries and claim the portfolio is ‘diverse’. Hmmm…….

    Their portfolio is also equally weighted, which introduces further bias, and casts doubt on their regression analysis.

    The indices are rebalanced every month. It’s not clear what their proposed portfolio does. Does it start equally weighted and evolve over time, or have they just compounded up daily returns and averaged, so that there’s an implicit daily rebalancing?

    The portfolio has a higher volatility than the market.

    There’s a million things wrong with this analysis.

  5. I agree replication is great, but this study is probably not the best object for replication. The selection of tickers (i.e., stocks) was done including too much information: “ticker symbols, the company names, a brief description of each company’s business”. This procedure probably favored well-known companies. Among public companies, being well-known is connected with being large and valuable. Thus the list of 82 out of 358 stocks selected is biased toward firms with high values at end of sample. Guess what? But being valuable means having had high prior returns, so we have selection on the dependent variable.

    A better method would be using forward-looking returns (those measured *after* the selection was done) and selection based on tickers alone. There are more potential problems here – just to mention one, the replication excludes the twelve companies which went bust, i.e. the lowest returns were excluded from the sample, making the finding of positive average returns less impressive.

    These are specific issues at a glance, but I think there were also many “forking path” opportunities in the research design.

    So, by all means replication, but maybe best to do it on good methodologies. Obviously, “good” is domain-specific. In the field of finance, the timing of the selection of securities for this kind of study is a well-known issue, creating huge potential bias yet easy to address.

  6. Cam Harvey actually discusses this paper in his presidential address:
    https://onlinelibrary.wiley.com/doi/full/10.1111/jofi.12530?casa_token=fcnxY1Vuwt8AAAAA%3Ag44LAC0e-uDvU-KdAxadsg-GDHyG37l_iNUMmRyjTx2ESgXP7ene1rHHLXMYt8jQtdabBL3N3kbwcw

    I’ll quote:

    Many would argue that we should increase the choice space further because there are other possible choices that I did not give to my research assistant. Suppose, for instance, there are three ways to handle delisting returns. Ex ante, one was chosen. The argument is that we should consider the fact that, hypothetically, we could have had three choices, not just the one chosen (see Gelman and Loken (2013)).

    It is not surprising that, under a large enough choice set, the long‐short strategy has a “significant” t‐statistic—indeed, dozens of strategies have “significant” t‐statistics. This is an egregious example of what is known as p‐hacking.

    One might think this is a silly example. But it is not. A paper referenced in the HLZ (2016) factor list shows that a group of companies with meaningful ticker symbols, like Southwest’s LUV, outperform (Head, Smith, and Watson (2009)). Another study, this time in psychology, argues that tickers that are easy to pronounce, like BAL as opposed to BDL, outperform in IPOs (Alter and Oppenheimer (2006)). Yet another study, in marketing, suggests that tickers that are congruent with the company’s name outperform (Srinivasan and Umashankar (2014)). Indeed, some have quipped that ETF providers such as Vanguard might introduce a new family of ETFs called “AlphaBet” with each ETF investing in stocks with the same first letter of a ticker symbol.

    • This makes me wonder if there are similar effects of political candidates names. We know that Hollywood actors often change their names to stage-like names. I’ll be someone (Andrew?) has done research regarding the relative success of easy-to-pronounce names (Trump, Sanders, Clinton, Bush,…) compared with harder names (Mayor Pete?).

      • Dale:

        I’ve not done such research. Based on other research I’ve seen and my general understanding of voting behavior, I’d expect names of political candidates to matter a bit in low-information, low-salience elections, but not so much in elections for president, where so much more information is available.

        • When was the last presidential candidate that had a hard to pronounce name? The closest I get is Pinckney in the early 1800s, LaFollette in 1924, Dukakis in 1988 (all of which are questionable, as to difficulty in pronunciation). Of course, to get that far, they need to survive multiple smaller elections, so perhaps those low-information, low-salience elections weed out those that might have survived in higher-information/salience elections.

        • Isn’t there a difference between non-clever ticker symbols, which may simply register as nondescript abbreviations, and non-Anglo surnames in the US? The definition of “easy to pronounce” will vary with the person who is trying to pronounce it, after all. I wonder whether it’s not more likely that those multiple smaller elections serve to weed out a lot of candidates whose unfamiliar names go hand-in-hand with unfamiliar beliefs, languages, etc.

        • John Barrasso
          Debbie Stabenow
          Amy Klobuchar
          Catherine Cortez Masto
          Lisa Murkowski
          Kyrsten Sinema
          Dianne Feinstein
          Richard Blumenthal

        • With politicians – especially former presidents – there’s a very strong familiarity bias. “Washington” seems like an easy name because we’ve all been saying it since we were three years old. But compared to “Taft” it’s hard to pronounce. There have been many presidents with potentially “hard” names, and there are lots of current senators with “hard” names.

          But I think there are more women with hard names than men. So are women with hard names more likely to get elected than men with hard names? Sounds like a job for p-hack-man!!!! :)

          Former Presidents with “hard” names:
          Washington
          Coolidge
          Van Buren
          Lincoln
          McKinley
          Eisenhower
          Roosevelt

          Current Senators with hard names:
          Debbie Stabenow
          Amy Klobuchar
          Catherine Cortez Masto
          Lisa Murkowski
          Kyrsten Sinema
          Tammy Duckworth
          Dianne Feinstein
          Richard Blumenthal
          Kelly Loeffler
          Mazie Hirono
          Cindy Hyde-Smith
          Jeanne Shaheen
          Bob Menendez
          Martin Heinrich
          John Hoeven
          Jim Inhofe
          Shelley Moore Capito

        • What’s your definition of a “hard” name? The only names that I might consider “hard” In your list of “Current Senators with hard names” are Kyrsten Sinema and maybe Jim Inhofe. Maybe what someone considers “hard names” depends on their background? To me, a “hard name”. is something like Piwowarski, Jauregui, or Przytulski.

        • “Maybe what someone considers “hard names” depends on their background?”

          Yes, that’s partly true.

          You don’t think Klobuchar is a weird or unusual name? I do. I know how to say it, so it’s not “hard” to say. Same with Feinstein and Murkowski. But it’s definitely a mouthful. Once you know how to say it, no name is “hard”.

          I’ve never met or heard of anyone named Roosevelt outside the well known Roosevelt family. It’s obviously familiar to us now, but it was probably unusual when Teddy first became a politician.

          Let’s face it: there’s no show biz personalities renaming themselves “Lady Murkowski”.

      • I think his point was quite the opposite–indeed we should not trust these types of tests because there are so many researcher degrees of freedom in the construction of the portfolios.

  7. Here’s a hypothesis I haven’t seen: “clever” tickers are created by “clever” companies–those astute to maintaining a good public image, with attention even to smaller details, attributes that (all else being equal) tend to favor success. With this being publicized, you could see a bandwagon effect, so that it no longer serves as an indicator of cleverness.

  8. Something doesn’t feel right about this. Why would this persist over time? This makes sense as an election effect for obscure offices, but not really for stocks over time, because this would suggest that nobody is paying attention to company fundamentals.

    An example might help. Suppose we start in the year 1984 with two companies, KOOL and XZTQ, with identical characteristics and over the years identical rates of sales growth and profit growth and a stock price of $10 per share.

    But if the stock price of KOOL compounds at almost twice the rate of XZTQ (“an annual return of 23.5%. The return for all shares in the New York Stock Exchange and Nasdaq, the United States’ two largest exchanges, during this period was a mere 12% per year.”) then by 2008 KOOL is worth $1585 per share and XZTQ only $152, which is far too much a difference for companies with my assumption of identical fundamentals.

    So my Occam’s razor tells me that something else is afoot, most likely that the KOOL companies are in different sectors, or they have some gigantic outliers, or something.

  9. Bob and zbicyclist are right.

    This is a flimsy paper. Ignore it.

    The reason it is so “unabashedly descriptive” is that it has very little analytical depth. It is simply a description of some simple calculations the authors did.

    • Seems like a perfectly valid observation, even if the explanation leaves much to be desired.

      I don’t understand how it’s different than saying, for example “the (tech industry, Internet 2.0 Corps, Social Media, MSFT, SBUX, San Fran Startups….) has beat the market by X over some time period”, things which are probably true, readily apparent from plotting prices, and things people want to understand or know about.

      • I’m talking more about how the paper doesn’t do the work necessary to convince the reader that what they find is real. To take just one example of how the results could be spurious, if a stock falls to $.10 per share and bounces from, say $.10 to $.50 per share for a while, that stock would seem to show very high average returns because returns some months would be -80% and 500% on other months.

        • They did actually track the total returns for the period in the 2009 paper. I presume they did also in the more recent paper although I don’t have access to it.

          “Overall, the compounded annual returns were 23.5% for the clever-ticker portfolio and 12.0% for the NASDAQ/NYSE portfolio.”

          I don’t doubt that it’s a real phenomenon; the question is why? And I think the answer is probably pretty straight forward: these are mostly new companies, and new companies are born to tap new and growing markets. Some fail, but the failure of a $100M company has a small impact compared to the decline of a once-multi-billion giant like Kodak. The broader market also has lots of mature cyclical – but very slow growing – businesses like railroads and automobiles.

          so I guess the lesson is keep your eye on the IPOs.

        • It may be a real phenomenon, but they haven’t shown that. Essentially they need to be neutral to all other risk factors, including the overall market. The fact that they don’t know this is poor, but unsurprising. It also means their quantitative analysis, including their t-test, is incorrect.

        • “It may be a real phenomenon, but they haven’t shown that.”

          OH! Now I see what you’re saying. You’re saying that to *isolate* the ticker name from all other risk factors and determine unequivocally that the ticker name is the precise factor solely responsible for the variation, you need to isolate other risk factors. Right?

          OK, I’ll buy that.

          But that’s not what they’re saying (which is why I was confused). They’re making a much simpler claim: that a basket of “clever ticker” stocks outperformed (some index). In fact, they specifically avoid claiming that the name of the ticker symbol was alone responsible:

          “This paper investigates the performance of stocks with memorable ticker symbols during the years 1984–2005 and finds that, on average, their daily returns are higher than for the overall market.”

          It goes on to postulate explanations, most of which seem implausible.

          The t-test doesn’t interest me at all, which is funny because they probably had to do that to get the paper published. Who cares about the average daily return? Investors care about basis and proceeds. Presumably the chart in Fig 1 “(2009 paper (PDF)) represents an equal “investment” in each group at t=0. that might be wrong but it seems like a fairly safe bet.

          But there are caveats even to the more basic claim:

          1) how acquisitions were handled. If a company was acquired, how is it’s value accounted for? In shares of the acquiring company as though it were an all-stock transaction? Or in cash equally distributed across the remaining “clever” companies, acquiring those shares on the closing date? IMO the former would be the appropriate method.

          2) is the market “index” treated as an average of indices, or is it treated as an index fund (which has expenses)?

          3) how are dividends accounted for (in both groups)?

        • Yes, but the other claim is not meaningful or useful. It’s like saying tickers that sound like baby talk outperform the market because you bought GOOG at IPO.

          The key thing to understand in this kind of analysis is what’s driving your performance. Standard industry approaches leave a lot to be desired, they’re just a linear regression or ‘factor model’, but if you were to show this to a hedge fund manager they’d see that you’re overweight sector x and underweight sector y, no secret sauce.

          The stuff about acquisitions increases the complexity consderably, you could get any answer you like by backfitting your approach. Same for dividend, messy stuff.

        • “It’s like saying tickers that sound like baby talk outperform the market because you bought GOOG at IPO. ”

          Perhaps this is a better example of what you mean:

          Suppose we took a set like “companies with ‘N’ as the third letter of their ticker”. Is there some set of fundamental properties about “Third Letter N” companies that could possibly relate to market performance? Highly doubtful. I certainly can’t think of one.

          But is there a set of fundamental properties about “clever ticker” companies that could possibly relate to market performance? Yeah, I think there is: the age of the companies. Prior to the 1980s, companies didn’t pay that much attention to tickers. Now they do. Younger companies grow faster – that’s the whole reason the market goes upward, because growing young companies replace older, declining companies. SW airlines (LUV) in. PanAm out.

          So in selecting “clever tickers”, the authors are (apparently unwittingly) selecting a set that’s systematically overweight in young, fast growing companies. Hence it outperforms the broader market.

          There is value in the knowledge of how markets work. I’m guessing you think this is stupid because you see “clever tickers” as a random set, like the “third letter N set”. But it’s not a random set. It’s a set that *likely* has fundamental properties that relate to market performance. As Andrew so often emphasizes, there *is* a theoretical reason to believe that a correlation exists, and possibly a very strong one. The authors, not knowing anything about markets, miss the likely reason and instead focus on problematic claims like trader name recognition and public stupidity about markets.

        • So I guess to me the real phenomenon is that two baskets of stocks performed differently. One basket was composed of “clever ticker” stocks. That doesn’t imply that the name caused the performance difference. In fact that claim would be preposterous.

          Rather, it’s some factor correlated with the “clever” tickers: probably that the “clever ticker” basket is composed of new companies following new trends in the economy, and therefore growing faster than the old guard companies that comprise the major indices.

        • ’Im guessing you think this is stupid because you see “clever tickers” as a random set, like the “third letter N set”.

          You’re guessing wrong. I’ve explained pretty clearly why it’s stupid.

  10. If they picked the stocks in 2006 and then compared market returns from 1984, there is a survivorship bias. This kind of study only works if:
    a. They looked for clever tickers among stock traded in 1984 and then compared
    b. They compared returns with all stock trading in 2006, not those trading in 1984 (so the comparison stocks also exhibit the same survivorship bias).

    Study seems to be gated so I could not check this point.

  11. You can get a copy of the paper here:

    http://economics-files.pomona.edu/GarySmith/Econ190/tickers.pdf

    It is tissue-thin.

    The clever portfolio is equally weighted and rebalanced daily. This is a red flag for me.

    For each trading day between the beginning of 1984 to the end of 2004, we calculated the daily
    return for an equally weighted portfolio consisting of those clever-ticker stocks with daily returns
    in the CRSP data base. As time passed, some clever-ticker stocks stopped trading for a variety of
    reasons (including bankruptcy, merger, buyout) and other clever-ticker stocks entered the CRSP
    data base. The portfolio adapted to these changes with the equal weighting of those stocks
    currently in the portfolio. The clever-ticker portfolio began in 1984 with 17 stocks and averaged
    24.3 stocks over this period, with a low of 17 and high of 33 stocks.

      • Just looking at how thin the paper is tells us to ignore the paper.

        To actually prove they found something would require all sorts of additional analyses that show it is due to x rather than to a, b, c, …, or m. The authors haven’t done that, so I see no reason to take the results seriously.

        So, I agree with everything you have said.

  12. The “clever ticker” explanation also doesn’t make sense theoretically.

    If it were true, we would expect a constant price premium over time, it would not imply a constant higher return over time (which is what they claim to find).

    A constant premium implies a higher, but constant P/E over time. This is not absurd — there would always be a little more demand for the stock because of the clever ticker so the price would always be a little inflated.

    But, a constant higher return over time implies a steadily increasing P/E with no upper limit. This makes no sense at all.

    • “But, a constant higher return over time implies a steadily increasing P/E with no upper limit. This makes no sense at all.”

      Your missing one factor. What are the two factors that control PE? Price and earnings. As long as earnings are rising price can rise without changing PE.

      “Clever Ticker” is a proxy for earnings.

      Rising earnings are correlated with “clever tickers” because the “clever tickers” basket is overweight in young companies, which are likely to grow faster than the overall market.

      What would be interesting is to select several baskets of stocks based on, say, 2- or 5-yr ranges of IPO dates, then reinvest all dividends and treat acquisitions like all-stock transfers, so the basket acquires shares in the acquiring company. I would definitely expect these to outperform the major indices.

      • We agree.

        If the clever portfolio outperforms because they have higher E growth, then it isn’t because of their clever tickers, it is because of something else. This would contradict the paper’s hypothesis.

    • That’s right. They might have a constant price premium over time, for example they might maintain a P/E (price to earnings) ratio of 14 whereas comparable stocks in their section might average 13.

      I vaguely remember a study from decades ago that companies that did more active investor PR (more media interviews, more analyst briefings) had higher P/E ratios. That makes sense; analysts may be more likely to recommend a stock if they feel they know the upper management. [Of course, there’s an obvious confound; CEOs with stocks that are performing below the market may be less likely to want to meet with analysts.]

Leave a Reply to jim Cancel reply

Your email address will not be published. Required fields are marked *