« Newspapers are obsolete—Part 2: Understanding the economy | Main | Obama’s economic, energy, environmental, and national security policy »
Thursday
Oct302008

Newspapers are obsolete—Part 1: Public Opinion Polling

Newspapers report lots of new data points, but too rarely provide useful patterns, trends and context, especially in their print editions. Take, for example, public opinion polls about the Presidential race. The lead story in the Los Angeles Times yesterday morning was this:

Barack Obama is leading Republican presidential rival John McCain in two battleground states, Florida and Ohio, where voters have more confidence in his ability to handle the troubled economy, a new Los Angeles Times/Bloomberg poll has found.

The story usefully reports on which issues seem to be driving voters, but if you want to know—and you probably do—who's going to win the election, that's not in the fish wrap. You have to go to the online version of the Times where you will find an interactive graphic showing Obama projected to win 318 electoral votes, McCain 174, and 46 up for grabs. Even online it is not explained why the graphic says Florida (50% to 43% for Obama in the poll) "leans" toward Obama and Ohio (49% to 40% for Obama in the poll) is "up for grabs." Nor would you get any information about trends over time in Ohio, Florida, or nationwide to help you judge if the latest poll results confirm others or are perhaps outliers.

To get the most complete and current analyses, aggregating all the polls and showing trends over time, you have to go to the blogs. Fivethirtyeight.com is the one I have followed daily for a month or so. There you will see that Obama is projected to win nationwide by 6-7% of the popular vote and 159 electoral votes. Not only are the LATimes/Bloomberg results reported here but all other surveys for Ohio and Florida are also reported. Scroll down a little and you see that the national popular vote margin has been essentially unchanged for several weeks.

Dig deeper and you will find how much weight this blogger (Nate Silver) gives to each poll depending on his analysis of how each determines who are "likely voters," how it allocates "undecideds," any consistent biases toward one party/candidate or the other, accuracy in 2004, etc. You can scroll down and read each daily update analysis since October 9. You can click on Ohio (74 poll results) or Florida (70 poll results) and instantly pull up everything this blog has ever said about polling in those States, including chronological tables of all polls for each State.  (Corrected 11/3/08: Nate Silver says here that he allocates the undecideds himself.  Hmm.  That would introduce a consistent bias into all polls.)

Another blog that aggregates and analyzes all polls is Pollster.com. I haven't been following it (thanks for the tip, Ray), so I know less about the detail that is available or how to find it, but it seems to have very similar information and better, interactive, graphics. It shows Obama leading McCain by 169 electoral votes with 85 in the toss-up category. It projects that 272 EVs are "strong" for Obama and 39 "leaning." (Only 270 are needed to be elected.) By clicking around, I was able to see trend lines nationally and by State and learn that Pollster uses different weightings and projects slightly different margins than Fivethirtyeight.com.

Fivethirtyeight.com has an acknowledged Democratic sympathy, and Pollster.com is said to lean Republican. If I cock my head to one side and squint, I think I might see both as being influenced slightly by their sympathies. A third website that seems to have even richer data is RealClearPolitics.com. (Thanks to Ray for this one too.) I haven't dug into it much, but its analyses and projections seem very similar to the other 2 sites. Like Pollster.com, it shows Obama leading by 169 EVs, but it is more cautious about "calling" States for either candidate and classifies more as "leaning."

Finally, I love this graphic from Brad DeLong's blog because it scales the area of each State to the numbers of electoral votes.

Path Finder

Bottom line, in less time than it took to read the LATimes piece in print, I could have checked all 3 websites and had a very good idea where the race stands and which States are close and which are not. In fact, I did check Fivethirtyeight.com about 12 hours before the fish wrap hit my driveway and it had already incorporated the results of the LATimes/Bloomberg poll. To be fair to fish wraps, Fivethirtyeight.com and Pollster.com don't report the poll internals showing which issues are driving voters (but that information was in the free online version of the Times and probably elsewhere online as well).

PrintView Printer Friendly Version

EmailEmail Article to Friend

Reader Comments (4)

So does that map say that Colorado is safe for electoral votes, or projected to go Obama?

I just checked out a terrific book called The Black Swan. The author, Taleb is the Dean's Professor in the Science of Uncertainties at the U.of Mass, Amherst!
I want to post the line about academicians on my bulletin board outside my office!
October 30, 2008 | Unregistered Commenterchristine
538 scores Colorada as solid Obama, with most likely margin 6.9% and Obama winning 93% of the simulations. RPC and Pollster score CO as leaning, and both project a 6.5% margin. LAT says CO leans to Obama.

Which line about academicians?
October 30, 2008 | Registered CommenterSkeptic
Lol, who says pollster.com leans republican? It was founded and is run by Mark Blumenthaul who is a former Democratic pollster. 538 and pollster are both run by those with a clear slant to the left. http://www.pollster.com/bio/mark-blumenthal.php
October 30, 2008 | Unregistered CommenterSkeptic2
Thanks for the correction, Skeptic2.
October 30, 2008 | Registered CommenterSkeptic

PostPost a New Comment

Enter your information below to add a new comment.

My response is on my own website »
Author Email (optional):
Author URL (optional):
Post:
 
Some HTML allowed: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <code> <em> <i> <strike> <strong>