2004 Election Simulation
The version of the browser you are using is no longer supported. Please upgrade to a supported browser.Dismiss

ABCDEFGHIJKLMNOPQRST
1
2004 Monte Carlo Pre-election and Exit Poll Simulation Model
2
3
This is an introduction to election forecasting and exit poll simulation using the 2004 Election Simulation Model (ESM). It utilizes pre-election state and national polls and unadjusted (WPE), preliminary (GEO) and final ("Composite")state exit polls. The ESM provides strong circumstantial evidence that the election was stolen.
4
5
The model produces popular and electoral vote win probabilities based on 200 Monte Carlo election trials.
6
7
In the pre-election model, state and national polls are adjusted for the allocation of undecided voters. In the post-election model, the user can analyze unadjusted and adjusted state exit polls. Monte Carlo simulation calculates state and national vote shares, median and mean electoral votes. The popular and electoral win vote probabilities are a function of projected vote shares, user-entered undecided voter allocation and a "cluster effect" factor adjustment to the exit poll margin of error.
8
9
The National Election Pool is a consortium of six media giants which funds the exit polls.
10
11
In 2000, 110.8 million votes were cast but only 105.4 million were recorded. Gore won officially by 540,000 votes nationwide. But approximately 6 million uncounted ballots indicate that he won the True Vote by at least 3.0 million. In Florida 75,000 under-punched and 110,000 over-punched cards were uncounted. It is very likely that Gore won Florida by at least 50,000 votes since he was the clear choice on a solid majority of 110,000 punch cards which were double or triple-punched.
12
13
In every presidential election, millions of voters are disenfranchised and millions of votes are uncounted. Forecasting models should have the following disclaimer: The forecast will surely deviate from the official recorded vote. If they are nearly equal, then there must have been a) input data errors, b) incorrect assumptions, c) faulty model logic and/or methodology.
14
15
The True Vote always differs from the official recorded vote due to uncounted, switched and stuffed ballots.
16
Were the pollsters who forecast the recorded Bush win correct? Or were pollsters such as Zogby and Harris correct in projecting a Kerry the winner? Kerry won the preliminary state and national exit polls.
17
18
In 2004, Kerry had a slight lead in 18 pre-election national polls: 47.5-47.3% and a nearly identical lead in the state polls. After allocating undecided voters, he was projected to win by 51-48%. Kerry also won the unadjusted and preliminary state and national exit polls.
19
20
The Final National Exit Poll is always forced to match the recorded vote (Bush won by 50.7-48.3%). Either the pre-election and unadjusted exit polls were wrong or there was massive election fraud.
21
22
MODEL OVERVIEW
23
24
The workbook contains a full analysis of the 2004 election, based on four sets of polls:
25
26
(1) Pre-election State polls
27
(2) Pre-election National Polls (18)
28
(3) Post-election State exit polls
29
(4) National Exit poll
30
31
Click the tabs at the bottom of the screen to select:
32
INPUT: Data input and summary analysis.
33
SIMULATION MODEL: Monte Carlo Simulation of state pre-election and exit polls.
34
NATIONAL PRE-ELECTION POLLS: Projections and analysis of 18 national pre-election polls.
35
NATIONAL EXIT POLL: Timeline of Preliminary, Final and True Vote demographics
36
Summary graphs: Polling timeline, Election trial simulation, Kerry state vote shares
37
38
Calculation methods and assumptions are entered in the MAIN sheet:
39
1) Calculation code: 1 for pre-election polls; 2 for EXIT polls.
40
2) Undecided voter allocation (UVA): Kerry's share (default 75%).
41
3) Exit Poll Cluster Effect: increase in margin of error (default 30%).
42
43
4) State Exit Poll Calculation Method:
44
1= WPD: Within Precinct Discrepancy.
45
2= Best GEO: adjusted recorded vote geographic weightings.
46
3= Composite: further adjustment to weight pre-election polls.
47
48
49
Note: The Composite exit poll data set was posted at 12:40am on the CNN election site and downloaded by Jonathan Simon. The polls were weighted (oddly) to include pre-election polls. The next and final adjustment was to force them to match the recorded vote.
50
51
52
POLL SAMPLE-SIZE AND MARGIN OF ERROR
53
54
Approximately 600 were surveyed in each of the state pre-election polls (4% MoE). Of course, the national aggregate has a lower MoE since approximately 30,000 were polled nationally. In 18 pre-election national polls the sample-size ranged from 800 (3.5% MoE) to 3500 (1.7%).
55
56
In 2004, 76000 were sampled in the state exit polls. Respondents ranged from 600 to 2800 (1.8% MoE). Kerry won the state exit poll aggregate by 51.1-47.5%. He also won the National Exit Poll, a subset of 13,660 respondents) by 51.7- 47.0%. The National Exit Poll margin of error was 0.88% before adjusting for a polling "cluster effect". Assuming a 1.0% MoE, Kerry had a 98% probability of winning the popular vote.
57
58
The Law of Large Numbers is the basis for statistical sampling. All things being equal, polling accuracy is directly related to sample size - the larger the sample, the smaller the margin of error (MoE). In an unbiased random sample, there is a 95% probability that the vote will fall within the MoE of the sample mean.
59
60
The "cluster effect" is measured in the model as the percentage increase in the theoretical Exit Poll margin of error. As it is often not practical to carry out a pure random sample, a common shortcut is an area cluster sample in which primary Sampling Units (PSUs) are selected at random within a larger geographic area.
61
62
The polling Margin of Error (MoE) is equal to 1.96* Sqrt (P*(1-P)/n), where P and 1-P is the two-party vote split and n the sample size.
63
64
ELECTION FORECASTING METHODOLOGY
65
66
There are two basic methods used to forecast presidential elections:
67
1) Projections based on state and national polls
68
2) Time-series regression models
69
70
Academics and political scientists create multiple regression models to forecast election vote shares. And they run the models months in advance of the election. The models utilize time-series data: economic growth, inflation, job growth, interest rates, foreign policy, historical election results, incumbency, approval rating, etc. Regression modeling is an interesting theoretical exercise, but it does not account for daily events which affect voter psychology.
71
72
Polling and regression models are analogous to the current market value of a stock and its intrinsic (theoretical) value. The intrinsic value is based on forecast cash flows; it is rarely equal to current market value. The latest poll share is the equivalent of the current stock price; the regression model represents the theoretical intrinsic value which deviates from the current market value.
73
74
The historical evidence is clear: final state and national polls, adjusted for undecided voters and estimated turnout, are superior to time-series models executed months in advance.
75
76
Inherent problems exist in election models, the most important of which is never discussed: Election forecasters and media pundits never account for the probability of fraud. The implicit assumption is that the official recorded vote will accurately reflect the True Vote and that the election will be fraud-free.
77
78
MONTE CARLO SIMULATION
79
80
Monte Carlo is a random process of repeated experimental "trials" applied to a mathematical system model. The Election Simulation Model runs 200 trial "elections" to determine the expected electoral vote and win probability.
81
82
Statistical polling (state and national) ideally is an indicator of current voter preference. In the Election Model, after poll shares are adjusted for undecided voters, the resulting win probabilities are input to a Monte Carlo simulation. The probability of an electoral vote victory is simply the number of winning election trials divided by the total number of trials (200 in the ESM; 5000 in the Election Model).
83
84
The only forecast assumption is the allocation of undecided/other voters. Historically, 70-80% of undecided voters break for the challenger. If the race is tied at 45-45, a 60-40% split of undecided voters results in a 51-49% projected vote share.
85
86
ELECTORAL AND POPULAR VOTE WIN PROBABILITIES
87
88
In the simulation, 5000 election trials are executed to calculate the expected electoral vote and win probability. The win probability is calculated using the projected vote shares as input to the normal distribution function. The simulation generates an expected electoral vote that is unaffected by minor changes in the state polls.
89
90
The probability of winning the popular vote is a function of the projected 2-party vote share and polling margin of error. These are input to the Excel normal distribution function (NORMDIST).
91
92
Prob (win) = NORMDIST (P, 0.50, MoE/1.96, True)
93
94
The theoretical expected electoral vote for a candidate is a simple calculation. It is just the sum of the 51 products: state electoral vote times the win probability. In the simulation, the average (mean) value will converge to the theoretical value as the number of election trial increase.
95
96
A random number (RND) between 0 and 1 is generated and compared to the probability of winning the state. For example, say Kerry has a 90% probability of winning Oregon. If the RND is less than 0.90, Kerry wins the 7 electoral votes; if the RND is greater than 0.90, Bush wins. The procedure is repeated for all 50 states and DC. The election trial winner is the candidate who wins at least 270 EV.
97
98
In the Election Model, 5000 simulated election trials were executed to determine the electoral vote winner. The number of winning trials are calculated for each candidate. The probability of winning the electoral vote is just the number of winning trials divided by 5000.
99
100
The probability of winning the True Vote is directly correlated to the probability of winning the electoral vote. But popular and electoral vote win probabilities in models developed by most academics and bloggers are often at variance with projected vote shares and electoral vote.