Mr. Buffet Sends His Regrets

In social data as in so many other arenas, Pareto’s principle of an approximate “80-20” dynamic holds true: in 2014, 66% of Twitter "cashtagged" ¹ traffic related to the most popular 20% of S&P 500 stocks.  $AAPL alone accounted for 11% of the tweets – more than the least popular 200 stocks combined.²  On a distribution graph that shows the % of traffic associated with each stock, the head of the curve is so narrow and the tail is so long it can be hard to tell them apart from the axes:


Everyone will know the names at the top – Apple, Facebook, Amazon – just as few will be familiar with the companies down at the other end: Hudson City Bancorp, SCANA, Stericycle.

But failing to be well-known is not the same thing as being antisocial – that designation is saved for those who are less discussed than they should be.  So rather than using a straight count of tweets, we looked at it against daily trading volume, expecting that stocks traded more often should be mentioned more often:


Apple and Facebook still stand out, and we have a new outlier in the other direction: Bank of America.  But are these stocks really that different than all the others?  We felt we were still missing something.  Since lower priced stocks are typically traded in higher volumes per trade (at $15 BAC is well below the median for S&P 500 stocks), we plotted daily tweets vs. average daily dollar volume to see if a pattern emerged:


Pretty, isn't it?  At least as far as graphs go.  (If you’re interested in that sort of thing, the r² here is 0.8. Translation: this probably is a reasonable way to approach what we're trying to do.³)  BAC still looks a bit antisocial, and it is, but it stands out primarily because it has a lot of dollar trading volume (similarly, FB is only slightly more social than average).   In terms of finding the most antisocial ratio, it helps to zoom in:


And the clear winner (loser?) is . . . Warren Buffet’s Berkshire-Hathaway (B class), which on average had more than $4MM in trading activity for each tweet generated in 2014.  That was twice the antisociability of runner-up Actavis, which came in at $2MM per tweet, and more than 13 TIMES the average value for S&P 500 stocks.  (LyondellBasell, Schlumberger and Phillips 66 followed next as three more of only 9 stocks that saw fewer than 1 tweet for each $1MM in trading volume.)  

As to what such performance might indicate about a stock, well . . . we'll let you develop your own theories.  Or become a client and you can borrow some of ours. 

We’ll talk about the most sociable stocks in another post – it has an extra complicating factor we’ll get into then.  In the meantime, we’ll see all of you (well, maybe not Warren Buffet) in the social data.


1.      A “cashtag” is popular way to specify a particular financial instrument; it is generally understood, for example, that "$IBM" refers to IBM publicly traded stock, whereas “IBM” references the company itself.  The practice originated on StockTwits and is now common on Twitter.

2.      For a variety of reasons, we restricted this analysis to tweets with only one cashtag, although most of the analysis does not vary much if you include tweets with multiple cashtags.  And we did exclude a few tickers that changed during the course of the year.

3.      An even better normalizing variable might be number of trade transactions rather than dollar value of shares traded, but that is difficult to get on a broad scale with any accuracy.