Thursday, November 6, 2008

Grading the Pollsters - Part 1: Popular Vote

Hello World! Say Hi to a New America! Hearty Congratulations to President Elect Barack Obama. As you may have realized from my earlier posts, I am very excited that my candidate won and won quite handily, contrary to what some of the polls were indicating towards the end.

Speaking of which, I decided to put my limited Excel skills to use and try to come up with a Report Card for the 12341521 pollsters (seemingly) tha
t covered this election. In this post, I have compiled a list of all the final National Polls (courtesy and have tried to rank the pollsters based upon the proximity of their numbers to the actual results.

First, here's the table that I use (Click on the picture for a larger view). The tan row is the RCP average. The green row is the actuals. The yellow highlights indicate the numbers/pollsters who are closest to the actuals.

(Click on the Picture for a larger view).

As you can see, when it came to estimating Obama's support, the top 3 were CNN/Opinion Research, Ipsos/McClatchy and ABC News/Washington Post. They were all only 0.17 percentage points off the actual number (53.17%).

When it ccame to estimating Sen. McCain's support, the top spot goes to Battleground (Lake) who were off by 0.17 percentage points and the second spot is shared by 4 pollsters - Rasmussen Reports, Ipsos/McClatchy, CNN/Opinion Research and Pew Research, who were all off by 0.83 percentage points.

Finally, when it came to estimating the spread, the top spot is shared by Rasmussen Reports and Pew Research who were off just by 0.34 percentage points and the third place is tied between Fox News, Ipsos/McClatchy and CNN/Opinion Research, who were all off by 0.66 percentage points.

So, if we were to grade these pollsters using some basic ranking logic, here's how it would look like:

First Place: Tied between CNN/Opinion Research AND Ipsos/McClatchy
Third Place: Tied between Rasmussen Reports AND Pew Research

Occupying the bottom rungs would be Battleground (Tarrance), Reuters/C-Span/Zogby (no surprise here, Zogby is a really bad pollster by any measure), Diageo/Hotline and Gallup (that's a surprise!).

Some additional thoughts:
1. The narrow margins by which these polls were off tells us how far along election polling has come in this country.
2. None of the polls were incorrect when it came to predicting the winner.
3. We can officially put the Bradley Effect to rest.

In Part 2, I'll look into how the pollsters did in the so-called "battleground" states (mental video of Sen. McCain drawing air-quotes).
Sphere: Related Content

No comments: