Via Alea, a very interesting econometric study on the impact of false news on stock prices.
In September 2008, an article on United Airlines’ 2002 bankruptcy resurfaced and was distributed as if it were new information. The company’s stock plummeted immediately, but bounced back and by the end of the day was off only 11% - which is still surprising, given that the news was in no way relevant. The paper uses a factor model to examine how the stock returned to its “correct” level following the false idiosyncratic shock – a process which took almost a full week, as illustrated here:
The factor model only has an R2 of 40%, but it appears to fit the data quite well at a glance – I’ll hazard the guess that with a stock like this, enough outlying idiosyncratic moves exist to distort the R2 even as the model preserves most of the behavior. This is an interesting opportunity to examine the interplay of idiosyncratic and systemic information – the alpha and beta moves are quite apparant (assuming we accept the factor model, but given its out-of-sample performance I’m inclined to do so in this case). Following the initial shock, the stock tracks the lower error band, though I’m not convinced there is significance in it tracking that particular line. Over the next few days, a deliberately positive alpha move carries it back to the factor model baseline, but the beta sensitivities remain apparant. Finally, alpha dissipates and the stock resumes its factor-predicted path.
Unfortunately, such cases are so hard to observe and one example does not data make. I’d be curious to see a similar analysis of Apple’s stock following the false announcement of Steve Jobs having a heart attack.
In a post called “So Long and Thanks for All the F-Tests“, Freakonomics writes about a new book called Mostly Harmless Econometrics: An Empiricist’s Companion, which they describe as:
…the rare book that captures the feeling of how to go about trying to attack an empirical question; and it does this by working through two or three dozen of the neatest empirical papers of the last decade…. It is also peppered with references to Douglas Adams’s writing — so what’s not to like?
Now, wait a second. TGR is peppered with Douglas Adams references. TGR has a post titled “F-tests begone!” How did I miss this?
(Freakonomics’ post title and the book’s title are themselves HHG2G references. I have refrained… for the moment.)
Seven years ago, the Mets were among the first teams in MLB to adopt a tiered pricing system in which it costs more to see a game against a good opponent than a bad one. At the time, it was the most sophisticated such plan in baseball. Others included simple methods like charging more for weekend games or prime summer games.
The act was a major step toward recognizing that while the supply of tickets was more or less constant for every game, demand could vary wildly. Nonetheless, ticket prices were still set for every game before the season started, and were locked in no matter how attrative (or otherwise) the game ultimately appeared. Rain, cold, chances to see historial achievements, unexpected opponent records (good or bad) – all of these affect demand in ways which can not be anticipated in March.
Now, the Giants have announced a new initiative which will adjust ticket prices up until the first pitch. A software program will estimate demand based on dynamic variables and adjust ticket prices according. The factors include both teams’ records, the stats of individual players (including the starting pitchers), the weather, the number of seats left to sell, proximity to gametime, promotional nights, and other similar metrics. Fans showing up on a rainy weekday night just before the game might be able to snag tickets for $5 that were $30 the week before.
It is likely that the model is not nearly as complex as one might think – in fact I imagine this is a perfect example of a time when a simple model can account for a surprising amount of variance. Not that this approach is revolutionary outside MLB – airlines and hotels have been doing something similar for years – but I think its an exciting and approrpriate use of data and would not be surprised if next season the metric is rolled out in a more expansive manner.
To think it was just a decade ago that the concept of statisticians on the field was inconceivable (but taken for granted today) – and now we have teams employing econometric methods to estimate demand. Quelle journee!