TIA and other statisticians - Need HELP!

[math_is_for_girls]

I am looking at precinct level data in my state, and the data looks odd - such as percentage vote for Bush compared to percentage vote for Kerry. WITHOUT exit poll data, how do you take into account known factors which can skew results?

For example, 1) in a Republican county you would expect the data to tend towards Bush, and in a Democratic county towards Kerry. 2) In an incumbent race, people cross parties far more often to vote for the incumbent. 3) In a previous race, a large percentage voted for Nader; who do these people vote for when Nader's not on the ballot?

How do you think about these things when looking at the data?
Thanks. :bounce:

[mgr]

1. A couple of questions first

Which state are you looking at? What data are you looking at, and from where did you get it.

First off you are looking at the 'real' data, e.g. the population that the exit poll is to estimate. I would recommend translating the percentages into real numbers(the actual number). Now, I am not sure what you are looking at, is it registration, or voting outcome? This would determine the next step to take.

I can help further, PM me with your e-mail address since you are a newcomer, and for some reason newcomers' PMs do not operate until a certain threshold of posting or activity is exceeded. I should be able to help directly.

Mike

[Bill Bored]

2. Welcome to DU. What precincts? OR votes by mail. Pls explain. nt

[math_is_for_girls]

3. Oregon still has precincts,

I suppose because lots of contests are specific to certain districts, like House of Repres, state races, etc. And of course, Oregon uses machines to count all those paper ballots. :)

[Bill Bored]

4. OK, so you're a truth seeker.

But keep in mind that your state is touted by the exit poll true believers as the BEST example of how paper ballot totals agreed with those polls, at least at the state level. Is this coincidence or do they actually agree at the "precincts" too?

Keep thinking and posting about this.

It's a big mistake in my view to restrict investigations to states where the exit polls didn't agree with vote totals, or were outside the MOE.

Central count Op Scans with ballots carried via US Mail is a far cry from a secure system in my book. If the exit polls didn't match in OR, you'd have heard plenty of screaming about that to be sure, but because they did match, it's assumed that everything was on the up and up. Good luck!

[TruthIsAll]

5. Hope this helps.

Step 1:

Get the data into spreadsheet as follows:
DemReg = Democratic registration
RepReg = Democratic registration
DemRep = DemReg/RepReg

KV = Kerry Vote
KP = Kerry Vote%
KVR= Kerry Vote/Kerry Reg
etc

Precinct DemReg RepReg DemRep KV KP BV BP KV KVR BVR
-
-
-

Total - - - -


1-See how the totals/ match up to the state/national exit poll numbers

2-Compare to 13047 NatExit Poll Party ID: 38 Dem /35 Rep /27 Independent

3- Compare the party vote/ registration ratios to prior elections, if you can get the data. Look for discrepancies in the ratios.

And go from there..

You:
In an incumbent race, people cross parties far more often to vote for the incumbent.
Me:
What is your evidence for that statement

You:
In a previous race, a large percentage voted for Nader; who do these people vote for when Nader's not on the ballot?

Me:
From the 13047 national exit poll, approximately 70% of Nader 2000 voters went to Kerry, 20% for Bush, 10% for other.