Jason Roos sends along this article:
On election days many of us see a colorful map of the U.S. where each tiny county has a color on the continuum between red and blue. So far we have not used such data to improve the effectiveness of marketing models. In this study, we show that we should.
We demonstrate the usefulness of political data via an interesting application–the demand for movies. Using boxoffice data from 25 counties in the U.S. Midwest (21 quarters between 2000 and 2005) we show that by including political data one can improve out-of-sample predictions significantly. Specifically, we estimate the improvement in forecasts due to the addition of political data to be around $43 million per year for the entire U.S. theatrical market.
Furthermore, when it comes to movies we depart from previous work in another way. While previous studies have relied on pre-determined movie genres, we estimate perceived movie attributes in a latent space and formulate viewers’ tastes as ideal points. Using perceived attributes improves the out-of-sample predictions even further (by around $93 million per year). Furthermore, the latent dimensions that we identify are not only effective in improving predictions, they are also quite insightful about the nature of movies.
My reaction:
I’m too busy to actually read this one, but it’s a great title! But–hey!–whassup with all those tables? Let’s start by replacing Tables 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 by graphs. (See here for some tips.) Or are you gonna tell me that the reader really needs to know that the standard error for the “Turnout – Midterm” variable in Dimension 4 is “0.108”? I mean, why not just print out a core dump in hex and cut out the middleman?
In all seriousness, the paper looks interesting and I’m sure would hugely benefit by some plots of data and fitted models.
On a more specific note, I wonder if the authors can shed any light on the controversial question of the Brokeback Mountain’s popularity in Republican-voting areas in “red states” (search Kaus Brokeback Mountain for more than you could possibly want to read on this topic). I never saw the movie, nor have I followed the debates about its box office, but I recall there being some controversy on the topic.
Does the following strike anyone else as odd?
"Furthermore, while the sample used by the Census is quite large, it is still a sample. The results of the election are not a sample in the sense that even abstaining from voting is informative."
Are the authors concerned with selection bias in the Census? (undercounting the homeless comes to mind) If not, what's their problem?
And as for abstaining being informative, isn't this only true when combined with some kind of polling to determine the make-up of the abstainers? (There are, after all, a lot of reasons not to vote.) If you're not happy with Census data I can't imagine that post-election polling will leave you feeling warm and fuzzy.
I'm not saying that voting results aren't useful for marketing. I spent years at a big financial services company arguing that we were leaving all sorts of useful geo-demographic data on the table. I'm just a bit perplexed by some aspects of the authors' approach.