MyDD :: Another look at polling outfits

Another look at polling outfits

by poserM, Sat Apr 12, 2008 at 02:08:12 PM EST

Not to pick on SurvayUSA, But I generally don't put too much stock in these polls. So, considering that SurveyUSA is considered the most accurate, I thought I'd look at how they've performed so far. Basically, I've listed the difference between SurveyUSA's poll and the actual result. -4.5 in Texas means SurveyUSA predicted a +1 Obama win and Hillary won by +3.5. 0 in Ohio mean SUSA was dead on (or close to it since I do some rounding). I general, I think they are doing a great job--although I won't think twice before betting.

Texas: -4.5
Ohio: 0
Virginia: -6
Maryland: 0
California: 0
New York: 0
Illinois: +3.5
New Jersey: +1
Massachusetts: +1.5
Missouri: -12
Alabama: -12
Connecticut: -1
Oklahoma: +3
Florida: +7
South Carolina: -16

For comparison, here's Zogby
Texas: 0
Ohio: -10
California: -23
New Jersey: -5
Georgia: -15
Missouri: +2
South Carolina: -14
Nevada: 0
New Hampshire: -15
Iowa: -4

Looking at where Zogby and SurveyUSA have data:
TX: S: -4.5, Z: 0 (Zogby more accurate)
OH: S: 0, Z: -10 (SUSA more accurate)
CA: S: 0, Z: -23 (SUSA more accurate)
NJ: S: +1, Z: -5 (SUSA more accurate)
MO: S: -12, Z: +2 (Zogby more accurate)
SC: S: -16, Z: -14 (Zogby more accurate)

Looks like 3 for 3. But, that could be misleading because if you look at how far off they've been and average the numbers of points off, SUSA is off on average 5.6 points compared 8.7 for Zogby.

Of courts, I'm not a mathematician, scientist, statistician, pollster, etc. So I can't vouch for my methods as being appropriate. But this is my lay person view of the numbers. Perhaps others can chime in.

As a side note, SUSA has polled many more states than Zogby. I wonder how this factors into people's impression of SUSA's accuracy.

Tags: polls, SUSA, Zogby, election 2008, election (all tags)

Permalink :: 2 Comments :: Post a Comment

Re: Another look at polling outfits (none / 0)

I've noticed a major factor separating these polls is how they report Obama's support in major cities and their surrounding suburban tier cities.

For example, SUSA has Clinton ahead of Obama in SE Pennsylvania (Philly + 'burbs) while the polls that place Obama within single digits show him ahead in that region, typically be a considerable margin (10+ points). These polls are more or less in agreement with how he'll perform in the rest of the state.

Ultimately on the 22nd the suburbs of Philadelphia will decide whether its a Clinton victory or a Clinton blowout.

by Homebrewer on Sat Apr 12, 2008 at 02:16:32 PM EST

Re: Another look at polling outfits (none / 0)

This is a good point. It's difficult to tell simply by looking at polls that one is more accurate than the other. You have to look below the hood . . . check the engine, kick the tires. And if you find something that raises a red flag, then you may be on to something.

Because ultimately the poll numbers depend on (1) questions asked, (2) how hard undecideds are pushed, and (3) the cross tabs, any one of these can affect one company's projections for a particular part of a state versus another.

I guess, as you say, we'll find out on 4/22.

by poserM on Sat Apr 12, 2008 at 02:45:55 PM EST
[ Parent ]

You are not logged in.

In order to post a comment, you must be logged in. If you have a member account, please log in to comment.

If not, you can make an account right here. It's quick and free.