Exit Poll Discrepancy

[jkd]

The NEP 2004 survey on question M broke down the response by religion to eight categories.

.ftp://ftp.icpsr.umich.edu/pub/FastTrack/General_Election_Exit_Polls2004/Datasets/Utah/Documentation/UT04G.pdf

The 20004 election results used the old model that was used prior to this year.

Protestant/Other Christian 82 18 79 1
Catholic 5 - - -
Jewish 1 - - -
Something else 4 - - -
None 9 64 33 -
http://www.washingtonpost.com/wp-srv/politics/elections/2004/graphics/exitPolls_utah.html

What is the relationship of the old model to the new data?

[minvis]

1. Added values for 2004

f. Changes from VNS 2000 variables

The following variables have been changed since the VNS 2000 exit polls:

Income – Two new codes have been added to the income question, making an 8-part response set used in every state. 6 and 7 were added and 9 was changed from 100,000 or more to 200,000 or more.

2003 total family income:
1 – Under $15,000
2 - $15,000 - $29,999
3 - $30,000 - $49,999
4 - $50,000 - $74,999
5 - $75,000 - $99,999
6 - $100,000 - $149,999
7 - $150,000 - $199,999
8 - $200,000 or more


Religion – Two new codes have been added to the religion question, making an 8-part response set used in every state. Codes 3 and 6 were added.

Are you: 1 – Protestant
2 – Catholic
3 – Mormon/LDS
4 – Other Christian
5 – Jewish
6 – Muslim
7 – Something else
8 – None


Time of decision – The response set for the question on time of decision on presidential vote is different than in 2000.


VNS 2000:
When did you finally decide who to vote for in the presidential election?
1 – In the last three days
2 – In the last week
3 – In the last month
4 – Before that

NEP 2004:
When did you finally decide for whom to vote in the presidential election?
1 – Just today
2 – In the last three days
3 – Sometime last week
4 – During the last month
5 – Before that

[jkd]

2. Thanks, I didn't notice the other changes.

The results and the survey question were both changed, expect for the religion question. Does anybody know why Mitofsky didn't change the religion results to conform to the new categories in the survey?

[minvis]

3. He did

I've looked at the raw data files from the ftp site (U. of Mich.). At the end of the files are several compilations of the data from some of these questions. For instance, there is a compilation of income data where it notes how many had income under $100,000 and how many over 100,000. They did the same thing with the religion data. They put the Mormon data into the Other Protestant category and Muslim in the Something Else category.

I hope this clears it up for you.

[jkd]

4. Let me restate the question? I'm not making it very clear.

Why would NEP change the categories on the survey, collect the data according to that change and then revert to the old model by putting Muslim in something else and Mormon into Other Christian? Why not keep the new categories in the results? In fact the new results combine Mormon, Protestant, and other Christian into one group.

[minvis]

5. It's what the WP is reporting

Actually, the hangup is what the Washington Post is reporting. The Washington Post is using the combined information, probably to make it like it similar to the 2000 information. The information with Mormon and Muslim divided out is there in the NEP data. The Washington Post just chose not to report those numbers.

[jkd]

6. Thanks for your patience.

I'm glad your looking at the data for Ohio and I hope you can identify the precincts that were polled.

I tallied the Utah vote from the raw data and found that the weightings actually reduced the Kerry vote by 3%. The weighted data came much closer to the actual vote, however.

[Time for change]

7. The weighted exit poll data also had Kerry winning by 3%

According to the Mitofski-Edison report