Edison - Mitofsky Exit Polls Method Statment

[sabra]

<<SNIP>>
http://www.exit-poll.net/election-night/MethodsStatementStateGeneric.pdf

Edison Media Research and Mitofsky International conducted exit polls in each state and
nationally for the National Election Pool (ABC, AP, CBS, CNN, FOX, NBC). The polls should
be referred to as a National Election Pool (or NEP) Exit Poll, conducted by Edison/Mitofsky.
All questionnaires were prepared by NEP.

...

The polling places are a stratified probability sample of a state. Within each polling place an
interviewer approached every nth voter as he or she exited the polling place. Approximately 100
voters completed a questionnaire at each polling place. The exact number depends on voter
turnout and their cooperation.

...

The table below lists typical sampling errors for given size subgroups for a 95% confidence
interval. The values in the table should be added and subtracted from the characteristic’s
percentage in order to construct an interval. 95% of the intervals created this way will contain the
value that would be obtained if all voters were interviewed using the same procedures. Other
non-sampling factors, including nonresponse, are likely to increase the total error.
%Error Due to Sampling (+/-) for 95% Confidence Interval
Number of Voters in Base of Percentage
% Voters with
Characteristic
100 101-200 201-500 501-950 951-23502351-5250
5% or 95% 6 5 3 2 2 1
15% or 85% 11 7 5 4 3 2
25% or 75% 13 9 6 5 3 2
50% 15 10 7 5 4 3
STATE # of polling locations Total # of interviews

AL 14 740 740
AK 29 1194 1194
AZ 35 1881 1500 381
AR 35 1483 1483
CA 35 2015 1541 474
CO 40 2501 2024 477
CT 15 879 879
DE 15 778 778
DC 15 796 796
FL 55 2760 2384 376
GA 35 1650 1650
HI 15 633 633
ID 15 804 804
IL 30 1443 1443
IN 25 946 946
IA 45 2491 2146 345
KS 15 671 671
KY 20 1075 1075
LA 39 1694 1694
ME 35 1992 1992
MD 20 1071 1071
MA 15 893 893
MI 50 2422 2198 324
MN 45 2192 2192
MS 15 806 806
MO 50 2276 2276
MT 15 664 664
NE 15 788 788
NV 39 2181 1716 465
NH 40 1896 1896
NJ 30 1524 1524
NM 35 1980 1609 371
NY 30 1476 1476
NC 40 2099 1800 299
ND 15 694 694
OH 49 2042 2042
OK 30 1604 1604
OR N.A. 951 N.A. 951
PA 50 2111 2111
RI 15 813 813
SC 34 1803 1803
SD 38 1567 1567
TN 25 1685 1352 333
TX 20 1583 1237 346
UT 15 828 828
VT 15 709 709
VA 25 1432 1432
WA 34 2063 1387 676
WV 43 1751 1751
WI 45 2340 2340
WY 15 767 767

<</SNIP>>

49 Precincts polled in Ohio, seems pretty wide spread to me...

Edit: OH bold

[sabra]

1. Here's the questionnaire used:

http://www.exit-poll.net/election-night/Final_State_Questionnaires.pdf

[Tace]

2. self delete

[IndyOp]

4. I'm no expert either

...and samples *can* predict a lot about what is true for populations. If I flip a coin 10 times and count the number of times the coin lands heads up I can make a useful prediction about the number of heads I would expect if I were to flip the coin 100 times or 1000 times or 10,000 times.

Also, note that the number of people polled in swing states (about 2000) is about twice as many as in non-swing states (about 1000 people). National opinion polls (Harris, Zogby, NYT) usually have a few hundred randomly selected people -- these have thousand(s).

Just a thought --

:eyes:

[TruthIsAll]

5. Indy, National polls usually sample 900-1200.

and have a MOE of around 3%

Exit polls in key states sampled over 2000. The exit poll MOE is much smaller, not only because of the larger sample size, but also because no voters are undecided. The MOE is less than 2%.

[tritsofme]

10. I understand why MoE should be around 2%, but why does their site say 4%?

The margin of error for a 95% confidence interval is about +/- 3% for a typical characteristic from the national exit poll and +/-4% for a typical state exit poll. Characteristics that are more concentrated in a few polling places, such as race, have larger sampling errors. Other nonsampling factors may increase the total error

http://www.exit-poll.net/faq.html#a15

And why do they say their national MoE should be 3%, when a sample of over 13,000 with a 95% confidence interval should clearly be under 1%?

[reality_bites]

9. Huh...

"If I flip a coin 10 times and count the number of times the coin lands heads up I can make a useful prediction about the number of heads I would expect if I were to flip the coin 100 times or 1000 times or 10,000 times."

No you can't. Well, you said your not an expert. I'm not either, but I've taken my share of stat. classes in my other life. It is true that "samples *can* predict a lot about what is true for populations," but coin flips don't represent populations well. Each coin flip is a independent action. If I flip a coin 7 times and get 7 heads, that says nothing about what I will get on the 8th flip of the coin.

If I flip a coin 10 times and get 7 heads, it doesn't hold that I should get around 700 heads if I flip it 1000 times.

[IndyOp]

3. Wow - look at the number of polling places...

More than 40 polling places in CO, IA, MI, MN, MO, NH, NC, OH, PA, WV, WI.
In comparison California had 35 and New York only 30.

Ohio (with 88 counties) had 49 polling places -- and the exit polls are wrong?

It looks as if there is an inverse relationship between the number of polling places and the accuracy of the exit polls -- more polls in swing states, most swing states polled for Kerry, though the tabulated votes went to Bush.

:eyes:

[skids]

6. Good find. Positive proof "adjustments" made later are wrong.

Now that I know that early voting was included in the exit polls,
I can say with confidence that the "adjusted" exit polls on CNN's
site are completely impossible. Why?

Because (repeat after me)

121,333,814 * 0.43 > 50456000


(And the numbers get even more wrong when you take away all the dead Bush2000 voters.)

So the next time Mitofski says "those numbers shouldn't be used, you should use the final numbers" tell him to put that in his crack pipe and smoke it.

[dargondogon]

8. Write to Warren Mitofsky

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=125547&mesg_id=126282

I can't, because of personal reasons. But perhaps others can.

[newyawker99]

11. Hi dargondogon!!

Welcome to DU!! :toast:

[dargondogon]

13. Thanks

It's nice to be here

[sabra]

12. Thanks, and good point! n/t

[ClintCooper2003]

7. Check out Florida - 55 polling places!

[shraby]

14. Current Subscribers to the poll

Current subscribers include:

Newsweek The New York Times

Los Angeles Times USA Today

The Wall Street Journal
The Washington Post
Chicago Tribune
National Public Radio
Atlanta Journal Constitution
Baltimore Sun
Boston Globe
Cleveland Plain Dealer
Dallas Morning News
KCRG-TV Cedar Rapids
Milwaukee Journal-Sentinel
Newsday
New York Post
NHK - Japan National TV
NY1-TV New York 1 News
Raleigh Observer
RAI - Italian National TV
Sacramento Bee
St. Louis Dispatch
St. Paul Pioneer Press
State News - Columbia SC WCBS-TV New York
WCVB-TV Boston
WHDH -TV Boston
WNBC-TV New York
WNYW-TV New York
Detroit Free Press

[Patsy Stone]

15. I wrote an e-mail to Steven Levy

at Newsweek. He had written an article about the voting machines, and I thanked him. A nice, positive, something could be wrong here, article.

Perhaps an e-mail to him requesting he take a peek-a-loo at the poll data, if possible?

[skids]

16. kick, for the researchers. n/t

[dargondogon]

17. Remind Mr. Mitofsky of pollsters' code

[email protected]
I urge you to abide by the polling standards promulgated by the
American Association for Public Opinion Research by publicly:

-- Releasing data to allow verification of your exit poll results by other researchers.
-- Discussing possible sources of error.
-- Describing any special scoring, editing, data adjustment or
indexing procedures (or calibration) used.
-- Describing your sample design.
-- Describing your sample selection procedure.
-- Delivering a full accounting of the final outcome of all "sample
cases."
-- Describing how non respondents differ from respondents.

The data no longer have proprietary value to your subscribers.
Election Night is long over, they hired you to help them promptly
report results. Now it's time to make a transition to public service,
Mr. Mitofsky.