April 27, 2009
Pre and
Post-election Model Assumptions
The 2008 Election Model (EM)
and the Election Calculator (EC) consist of three basic components: recorded
(“official”) data, assumptions and methodology. The recorded, official vote
data and calculation methods are easily verified; it’s the estimates for the
base case assumptions that are the subject of debate. This comprehensive
summary will show that the assumptions are based on the best available data.
The base case assumptions are best estimates derived from the following data
sources: 2004 and 2008 official recorded vote, pre-election state and national
polls, unadjusted state and national exit polls, voter mortality tables,
historical returning voter turnout, Census total votes cast. Because of the
margin of error in the polling data and other assumptions, a sensitivity
analysis (”stress-test”) was provided in both models to examine the effects of
changes in the assumptions and determine which of the assumptions are most
critical.
The preliminary May 2008
model projection indicated that Obama would win by 12 million votes: 71-59m
(54.1-44.7%). The number of returning 2004 voters and vote shares was
determined by adjusted the 2004 recorded vote for a 4.8% voter mortality rate
and 3.45 million uncounted votes. An estimated 95% of these voters turned out
in 2008. The projected Obama and McCain vote shares of returning and new voters
were close to the preliminary, unadjusted 2004 National Exit Poll estimates for
Kerry and Bush, respectively.
The Monte Carlo simulation
projected Obama would win 365.3 expected electoral votes and a 53.1-44.9%
share. He won by 365-173 EV with a 69.5-59.9m popular vote margin
(52.87-45.62%). The base case assumption was that he would win 60% of late
undecided voters. A 75% undecided voter allocation produced a 53.8% share and 379.5
expected EV. Pollsters typically allocate 70-90% of the late
undecided vote to the challenger depending on the incumbent’s performance.
McCain was the de facto incumbent; Bush approval was 22%.
The
Election Model projected 365.3 expected electoral votes - but not the True Vote
The model exactly matched
the electoral vote and was within 0.2% of the popular vote share. But a post- election sensitivity analysis
using Final 2008 NEP vote shares with a plausible returning voter mix indicated
that Obama’s True vote share was 2-4% higher than the recorded share.
A
Conversation about the 2008 election
A
one on one discussion highlights the
statistical anomalies. The Final 2008
National Exit Poll, like the 2004 NEP, provides the primary “tell”. It’s standard procedure for the Final NEP to “adjust” demographic
weights and vote shares and force a match to the recorded vote. But an
impossible mix of returning voters was necessary in order to force the match –
just as it was in 2004.
True
Vote Election Calculator (new)
The True Vote Election
Calculator (TVEC) was designed for a state-based analysis of the 2004 and 2008
elections. An important feature is the calculation of the true vote for each
state. The prior election returning voter mix is based on mortality and turnout
rates applied to Census total votes cast in each state. If votes cast exceed
the recorded vote, the difference
(uncounted ballots) is allocated and added to the recorded votes. If the
recorded vote exceeds votes cast, the difference (“stuffed” ballots) is
deducted. Shares of returning and new voters are calculated by adjusting
preliminary and Final National Exit Poll
(or user entered) vote shares. The ratio of the recorded state vote
share to the national share is a key factor in the vote share adjustments. The
TVEC automatically calculates a popular and electoral vote sensitivity analysis
over a range of assumptions.
For the 2008 election,
two scenarios were analyzed. In the first, Kerry was assumed the winner in 2004
by the aggregate unadjusted exit poll (52-47%). The model indicated that Obama
had a 57.5% true vote share and won by 21.8m (75.5-53.7m) with 431 electoral
votes. In the second scenario, Bush was assumed the winner by the recorded
50.7-48.3% vote share. Obama had 55.2% and won by 15.8m votes (72.5-56.7m) with
414 electoral votes.
Using the preliminary
12:22am 2004 NEP vote shares, Kerry had 53.3% and won by 67.0-58.0m with 360
electoral votes. Based on the Final NEP, he had a 52.2% share and won by
65.6.0-59.6m with 336 electoral votes.
Prior
Election Returning Voters
Proving that Obama and Kerry
both did much better than the official recorded vote is really a
straightforward exercise in logic. The key measure is turnout of prior election
voters in the current election. When reasonable, plausible estimates are made
for returning (and new) voters, the resulting Obama and Kerry vote shares are
4% higher than the recorded shares. A plausible voter mix is derived by adjusting
the 2000, 2004 and 2008 recorded votes for a 1.2% annual mortality rate and
assuming a 95% turnout of prior election voters still living.
Is the
Media cherry-picking LV polls and ignoring the RV polls?
From 1992-2004, the
Democrat won new voters by an average 14% margin. Registered voter (RV) polls
include new voters – mostly young Democrats. Likely voter (LV) polls do not
include new voters. The Real Clear Politics (RCP) site fails to list the final 4
registered voter (RV) polls. Instead, RCP averaged the 15 final likely-voter
(LV) polls: 52.1-44.5. Why were the RV polls not listed? They show an
average Obama lead of 52.75-39.75%, a 13% margin before allocation of undecided
voters. Obama won new voters by 71-27%.
RV polls are superior to LV polls because they include new voters, a
16 % share of the electorate. The 131.37m official vote consists of new and
returning 2004 voters.
A good approximation of
the number of returning voters can be calculated based on a 95% turnout of 2004
voters net of 4.8% mortality. There were approximately 110m returning voters,
therefore 21m (16% of 131m) were new.
The Final RV
and LV Polls Confirm an Obama 20m Landslide
This
analysis shows that the final pre-election likely voter (LV) and registered
voter (RV) polls were both correct even though there was a 4% difference
in their projected shares. The LV polls, with undecided voters allocated,
correctly predicted likely
(returning) voter shares. But since LV polls exclude new voters,
they understated Obama’s total projected vote share. In contrast, the RV
polls sampled both returning and newly registered voters. Obama won
71% of approximately 20m new voters. After allocating undecided voters, the
average RV poll projected a 57.25% Obama share and a 20 million margin. It
is interesting to note that the LV projection, when combined with the National
Exit Poll shares of new voters, produced a total vote share that closely
matched the RV projection. The RV projected share was also independently
confirmed by the Election Calculator model, which used National Exit Poll vote
shares and an adjusted, feasible returning voter mix.
The tremendous Obama GOTV and new voter registration effort did not
produce the expected numbers of votes. The nine million net increase in
the 2008 recorded vote was dwarfed by the 17 million in 2004. An analysis of uncounted votes and exit poll
discrepancies in the 1988-2004 presidential elections revealed some very
interesting information. In the 4 elections from 1988 to 2000, the average
unadjusted state exit poll aggregate was within 1% of the recorded vote - after
the uncounted votes were included. But 2004 was different in kind and scope:
including the uncounted votes did not make up the difference. HAVA (Help
America Vote Act) was passed during the first Bush term.
2008
National Exit Poll Does Not Compute
The National Exit Poll “Voted in 2004” returning voter
mix (46 Bush / 37 Kerry / 4% Other)
implies that Bush voters comprised 46% (60.3m) of the 131.37 recorded 2008
vote. If the election was fraud-free, Bush had 62.0 million recorded
votes. Approximately 3.0m died, so at most only 59.0m could have voted in 2008.
Assuming that 95% turned out, about 56m voted. Where did the 4.3m Bush voters
come from? If the 2004 exit poll reflected the True vote (Kerry by 52-47%),
then Bush only had 57m votes and about 51m returned to vote – a 9m discrepancy.
The NEP also implies that 5.2m (4%) were returning third-party voters, but there
were only 1.2m third-party votes in 2004.
Where did the extra 4m third-party voters come from? One can only conclude that since an
impossible Final 2008 NEP was forced to match the recorded vote, the recorded
vote was also impossible. It is interesting to note that the National Exit Poll
indicates that new voters comprised just 13% (17m) of the electorate and
returning third-party voters 4% (5.2m). But 4% is impossible because only 1.2m
third-party votes were recorded in 2004. If the third-party percentage is
changed to the true 1%, then the other 3% applied to new voters raises the
percentage to 16% - where it should be. It means that there were 21m new
voters, not 17m as the NEP indicates.
The
2008 Election Calculator model determines the True vote based on the number of
returning 2004 voters and Final 2008 National Exit Poll vote shares. It
indicates that Obama’s True Vote was more than double his recorded 9.5m vote
margin. Anomalies abound in the state
and national pre-election polls, exit polls and late vote shares. The number of
uncounted votes is still unknown.
Matching
to the 2008 Recorded Vote
The 2008
National Exit Poll Bush/Kerry (46/37) returning voter mix, like the 2004
Bush/Gore (43/37) mix, was mathematically impossible. In order to match the
recorded vote, both required more returning Bush voters than were still living.
The 2008 return voter mix anomaly is further confirmation that Kerry won. An
impossible high returning Bush voter turnout and low Kerry turnout were
required to match the 2008 recorded vote. Like Gore and Kerry, Obama’s True
vote far exceeded his recorded vote.
In order to force a match to the recorded
2008 vote, the 2008 NEP implied that Bush won in 2004 by 53-43%. He won the official vote by 50.7-48.3%
(62-59m). Because the NEP returning voter mix was mathematically impossible, the
2008 Election Calculator revised the mix assuming two scenarios: the recorded
vote and the unadjusted exit poll. It determined that the vote shares
required to match the 2008 recorded vote were implausible in both scenarios.
Furthermore, the returning 2004 voter mix required to match the vote was
impossible.
Using Excel
Solver to Calculate the Required Vote Mix
The National Exit Poll asked
respondents whom they voted for in 2004 as well as in 2008. Respondent vote shares, in combination with a
feasible returning voter mix, did not match the official count. But the
Final NEP must always match the official vote count- come hell or high water. How
do the pollsters adjust (force) the mix and/or the vote shares to achieve the
match? The Excel Solver program used a “goal-seeking” search algorithm to
automatically adjust the returning voter mix and vote shares to match the
official count. But the mix was impossible: it required more returning Bush
voters than actually were recorded for him.
Obama’s vote was a function
of his share of returning Kerry, Bush and new voters. The Final NEP vote shares
were 89%, 17% and 71%, respectively. Due to the margin of error, it is
instructive to view the effects of incremental changes in these shares on the
total vote. The Election Calculator sensitivity analysis showed that Obama did
much better than the recorded margin indicates and may have won by as many as
23m votes. In order to derive a plausible returning voter mix, two scenarios
were analyzed: a) the recorded 2004 vote (Bush 50.7-48.3%) and b) the
unadjusted 2004 exit poll (Kerry 52-47%). The mix was calculated for both
scenarios using identical voter mortality, uncounted votes and turnout rates.
Obama won the recorded vote (unlikely) scenario by 17.6 million: 75.4-57.8m
(55.7-42.7%). He won the exit poll (most-likely) scenario by 22.6m: 77.9-55.3m (57.5-40.8%).
Election
Calculator Sensitivity Analysis
All models consist of
recorded data, assumptions (parameters) and calculations. Given that the
Election Calculator (EC) model computations are correct, the only uncertainties
are the input base case assumptions. These are estimated from voter mortality
tables, historical returning voter turnout percentages, Census votes cast, and
2008 National Exit Poll vote shares. Due to the margin of error inherent in
these estimates, the model examines the effects of changes in the base case
assumptions via a comprehensive set of sensitivity analysis tables. Each table
consists of 25 combinations (scenarios) of two input variables. The purpose is
to gauge the impact of each combination on Obama’s vote share and margin. A
very close approximation to the True Vote is somewhere in the table; the most
likely, base case estimate is the central cell. The range of plausible vote
shares can be narrowed from 25 to 9 by focusing on the combinations that lie
within the margin of error.
The 10 million
Late Vote Anomaly
How does one explain the
discrepancy in the Democratic vote share between the initial votes recorded on
Election Day and the millions of votes recorded later? On Election Day, 121.21m
votes were recorded and Obama led by 63.4-56.1m (52.3-46.3%). As of Jan. 5,
131.37m votes have been recorded. Obama leads by 69.5-59.9m (52.87-45.62%). He
has a 59.2-37.5% share of the 10.16m votes recorded since Election Day, a 7%
increase in vote share and 15% increase in margin. Gore and Kerry also had
late vote margins that far exceeded their Election Day margins. In the last 3
elections, the average Democratic late vote share was 7% higher than the
initial share.
A
Trifecta of 2008 National Exit Poll Anomalies: New, former Bush and third-party
voters
Apparently,
adjusting the returning Bush/Kerry voter mix to an impossible 46/37% was
not enough to force the 2008 National Exit Poll to match the recorded vote. Exit pollsters also had to adjust
Obama's share of new voters (first-timers and others who did not vote in
2004) from 73% to 71%. Add this anomaly to the impossible mix of returning Bush
and third-party voters and we have hit the 2008 Election Fraud Trifecta. The
Obama True Vote Landslide was denied. It made sense to adjust the returning
voter mix AND the vote shares to get the match. Adjusting the mix or the shares
(but not both) would have been too obvious; that would force exit poll deniers
to jump through even smaller hoops to explain the numbers. On the other hand, a
combination of changes to the returning voter mix as well as the vote shares
minimizes the radical surgery required to force the National Exit Poll to
match to the recorded vote.
To Believe the
Official 2008 Recorded Vote, You Must Also Believe This
Who still believes that
elections are fraud-free? The Final National Exit Poll (NEP) is always forced
to match the recorded vote and therefore has assumed zero fraud in every
election since 2000. Since there was fraud, the Final NEP voter demographics
cannot be accurate. If you believe that 2008 was fraud-free and that Obama won
by the official recorded 9.5 million votes, then you must also believe the
following statements…
Projecting the
Electoral Vote and Win Probability
Most
election forecasters, media pundits and academics fail to apply basic
probability, statistics and simulation methodology in forecasting the electoral
vote and corresponding win probability. A
meta-analysis or simulation is not required to calculate the expected electoral
vote. Regardless of the particular method used to forecast the winner of
each state, the corresponding state win probabilities are necessary in order to
calculate the expected EV and win probability.
Election
Model: Monte Carlo Simulation
The expected state EV is the
win probability times the electoral vote. If the probability is 50% and the
state has 20 EV, then each candidate gets 10 electoral votes. The Total Expected EV is the sum of the
products for all the states. The math is exceedingly simple. The tough part is
accurately projecting state vote shares. The Election Model uses the latest
state poll average and allocates undecided voters to derive the projected
2-party vote. The state win probabilities are used in a 5000 election-trial
Monte Carlo simulation. The probability that a candidate will win the EV is the
number of winning election trials/5000.
Includes the following:
. Time-series regression models vs.
Monte Carlo polling simulation
. Final 2004 state and national
projections confirmed by the exit polls
. Analysis of 2004 registered voter
(RV) and likely voter (LV) polls
. Basic Polling Mathematics
. Overview of Monte Carlo Electoral
Vote Simulation
2008
Primaries: Footprints of Election Fraud
Confirmation
of a Kerry Landslide
2004-2006
Election Fraud Analytics