The 1988-2008 State and National True Vote Model

17 Mar

Getting Started with the 1988-2008 State and National True Vote Model

Richard Charnin (TruthIsAll)

Oct. 9, 2011

The 1988-2008 State and National True Vote Model (TVM) calculates the True Vote for all state and national presidential elections since 1988. It contains a unique database of total state and national votes cast, unadjusted state exit polls and final national exit polls. The TVM has a built-in sensitivity analysis to calculate vote shares and margins over a range of scenarios. It automatically determines vote shares of new and returning voters required to force a match to the state’s recorded vote. These capabilities provide a forensic tool to determine the likelihood of vote miscounting in any election.

The True Vote

It is incontrovertible that the Final National Exit Poll (NEP) never reflects the True Vote. The Final is always forced to match the recorded vote even though millions of votes are uncounted in every election. It stands to reason that if the Final is forced to match an impossible recorded vote, then the Final must also be impossible – and therefore all the demographic cross-tabs reported in the media must be invalid.

Simple arithmetic shows that in 1972, 1988, 1992, 2004 and 2008 the number of returning Nixon and Bush voters were not only implausible, but also physically impossible. In 2004 preliminary state and national exit polls were downloaded from the Internet after midnight. They all showed Kerry leading by substantial margins. But final exit polls show that impossible adjustments were made to the number of returning voters (“the mix”) as well as to the corresponding vote shares. The reason is obvious, but goes unreported in the mainstream media: It is standard operating procedure to force all exit polls to match the recorded vote – regardless of whether or not the votes were miscounted.

Therefore, to obtain a good approximation of the True Vote, the Final NEP returning voter mix must be replaced by a feasible, plausible set based on voter mortality and the best estimated turnout percentage of living previous election voters.

A feasible mix produces a True Vote that is quite close to the unadjusted state and national exit polls. Yet the average Democratic True Vote share usually exceeds the unadjusted exit polls by 1-2%. Why the discrepancy? The exit poll sample is based on prior election recorded votes. But as we know, the recorded vote always differs from the True Vote.

In the eleven presidential elections since 1968, over 80 million votes were cast but never counted. The Republicans led the average recorded vote by 49-45%. The Democrats led the True Vote by exactly the reverse: 49-45%.

The 1988-2008 recorded vote, exit poll and True Vote trends are displayed in tables and a bar graph. Deviations between Democratic and Republican vote shares are indicated, along with the Democratic exit poll discrepancies from the recorded vote.

An additional feature is the reconciliation of true vote and exit poll discrepancies from the recorded vote. The Net Uncounted Vote (NUV) is the difference between uncounted and stuffed ballots, both of which are usually unknown. The user can input an estimated uncounted vote rate. Based on the rate and the NUV, the model will calculate the number of stuffed ballots and switched votes in order to reconcile the discrepancies. In Florida 2000, there were 185,000 spoiled, uncounted ballots. In Ohio 2004, RFK, Jr. reported approximately 350,000 uncounted votes.

Using the Model

Online users are assigned anonymous numeric codes for the online chat feature.

There are just three essential inputs (click the “Input” sheet tab).
1, Election year: 1988-2008
2. State abbreviation code (NA for national)
3. The calculation method determines the number of returning voters from the previous state or national election:

Methods 1-4:
1 – Official Recorded vote
2 – Votes Cast (recorded + net uncounted)
3 – Unadjusted state exit polls
4 – Recursive True Vote (calculated)

The final state or national exit poll is also displayed at the right and forced to match the recorded vote.

Optional Data Input

The model has built-in defaults for the turnout of previous election voters in the current election and allocation of uncounted votes in the previous election.The default assumptions may be overridden. It is often of interest to view the changes to the base case shares of new and returning voters that are required to match a given total share. The model calculates the required vote shares automatically.

Forcing a Match to a Recorded or Hypothetical Vote Share

For Democrat, Republican and third-party (Other):
Total vote share = ∑ return voter mix (i) * vote share (i), where I =1 for DNV (new) voters, 2 = returning Democrat, 3 = returning Republican, 4 = returning Other.

Methods 1-4 calculate a plausible return voter mix based on previous election vote shares adjusted for voter mortality and turnout. The corresponding vote shares are adjusted automatically. This is a forensic tool for exposing likely vote miscounts.

Enter code 1 in the indicated cell to determine the vote shares required to force a match to the unadjusted exit poll.

To force a match to an arbitrary state or national vote share, enter the share. Shares of returning and new vote shares will be automatically adjusted.

In calculating the National True Vote (methods 1-4), the model uses Final NEP vote shares. An exception is the 2004 election in which preliminary 12:22am vote shares are used. For each method, the adjusted return voter mix replaces the Final NEP (impossible and/or implausible) mix.

Since the unadjusted “How Voted” exit poll cross-tabs are unavailable for the states, the model automatically generates a return voter mix and vote shares that force a match to the recorded vote. The vote shares are derived from the NEP shares by applying a state to national vote share ratio.

Sensitivity Analysis

In 2004, Bush won the recorded vote by 50.7-48.3%. The model shows that Bush needed 109.7% of returning Bush 2000 voters to match his recorded vote.

According to the 12:22am National Exit Poll timeline(the base case), Kerry had 57% of new voters, 91.0% of returning Gore voters and 10% of returning Bush voters. In the Final NEP, he had 54%, 90.0% and 9%. The Final was forced to match the recorded vote.

The sensitivity analysis is built-in. The first set of tables display 25 scenarios of Kerry vote share, margins and win probabilities. The scenarios are in increments around his base case share: 57% of new (DNV) voters and 10% share of returning Bush voters. The second set of tables are in increments around Kerry’s 91% base case share of returning Gore voters and 10% share of new voters.

To execute each scenario separately would be prohibitively time-consuming.

Base case (most-likely) vote share calculation
Kerry share = 53.6% = .17*.57 + .414*.91 + .377*.10 + .031*.64
= ∑ return voter mix (i) * vote share (i), where I =1 for DNV (new) voters, 2 = returning Democrats, 3 = returning Republicans, 4 = returning third-party (Other).

Results for the first set of scenarios (three 5×5 tables):
Worst case scenario (Kerry 53% of DNV; 8% of returning Bush voters):
Kerry has 52.1% and a 7.0 million vote margin.
Popular vote win probability: 96.8%

Most likely (Kerry 57% of DNV; 10% of returning Bush voters):
Kerry has 53.6% and a 10.7 million vote margin.
Popular vote win probability: 99.8%

Best case (Kerry 61% of DNV; 12% of returning Bush voters):
Kerry has 55.1% and a 14.4 million vote margin.
Popular vote win probability: 100.0%

True Vote Model Sheets
Click the tab at the bottom of the screen to view the sheet.
– Input Input data sheet
– Main Output summary
– TrueVote Calculation Detail.
– StateMod National and state recorded vote, exit poll and True Vote. The last two columns show the discrepancies between the True Vote vs. the Recorded Vote and Exit Poll.
– HistWPD 1988-2008 State Exit Polls: Within Precinct Discrepancies
2004Exit state exit poll timeline: Unadjusted WPE, Best GEO , Composite and Final (recorded vote). The WPE measures were provided by pollsters Edison Mitofsky. Enter 1 for VNS, 2 for DSS, 3 for IMS (aggregate Kerry shares 51.84%, 52.17% and 51.97%, respectively).
– StateExitWPE 1988-2008 votes cast, recorded and exit polls. The adjusted vote share is based on the uncounted vote allocation. The adjusted state and aggregate national share are reasonable approximations to the True Vote calculated in the StateModel.
– Trend 1988-2008 Democratic Recorded Vote and Exit Poll shares
– Machine04 2004 voting machines.
– Census – 1968-2008 Reported Voting rates


Posted by on March 17, 2011 in True Vote Models


2 responses to “The 1988-2008 State and National True Vote Model

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: