From Richard Charnin:
I have uploaded the Excel True Vote Model as a Google doc. It is now available to everyone on the Internet who wants see how the True Vote is calculated for all elections from 1972-2008.
Here is the direct link:
We distinguish between the True Vote (how people actually voted) and the official, recorded vote. It is an undeniable fact that in every election, the True Vote is never equal to the recorded vote. This is self-evident since the number of votes cast is never equal to the number recorded and therefore the True Vote shares cannot be equal to the recorded shares.
In the eleven elections since 1968, there have been approximately 80 million net uncounted votes. Net uncounted votes have declined from 10.6 million (10%) in 1988 to 5.4 million (5%) in 2000 to 3.4 million in 2004 (3%). Given that the vast majority of uncounted votes are Democratic, the Democratic recorded vote always understates the True Vote.
In the 1968-2008 elections, the average presidential recorded vote share was 49-45% in favor of the Republicans. We will show that the average True Vote share was 49-45% in favor of the Democrats.
Go here for more information:
Running the Model
To run the model, just two inputs are required:
1. Choose the election by entering the election code (1-10) in cell B6
2. Enter the calculation method (1-5) in cell B7
Election codes range from 1 for 1972 to 10 for 2008.
There are 5 calculation methods. Vote shares can be calculated for all elections since 1972 using the following methods:
1. Final National Exit Poll (forced to match the recorded vote)
2. Returning voters based on previous election recorded vote
3. Returning voters based on previous election votes cast
4. Returning voters based on previous unadjusted state exit poll aggregate
5. Returning voters based on previous election True Vote (the best measure)