|
Politech
mailing list archives
FC: Computer simulation predicts George W. Bush wins election
From: Declan McCullagh <declan () well com>
Date: Mon, 06 Nov 2000 09:32:50 -0500
http://www.cluebot.com/article.pl?sid=00/11/06/0520254&mode=nested
Computer Simulation Predicts Bush Wins Election
posted by cicero on Monday November 06, @07:15AM
from the computers-never-lie dept.
Congratulations, Dubya! If our electoral simulation is correct,
you're the new prez. This is based on what statisticians call
a Monte Carlo simulation, a standard technique in which virtual
elections are repeated over and over based on the latest poll
intelligence to get an average score. The result? According to our
in-house mathematician and polling analyst (read the writeup below):
"At the end of each list of trials the maximum, minimum, and average
over the 100 trials are reported for each candidate. Though 270
Electoral Votes are needed to claim the Presidency, in none of the 200
simulations here did Gore win over 257. The average result is Bush 340
and Gore 200." Alas, our author can't take the credit he deserves.
Because of unrelated, non-political consulting work he does for the
government, he wishes to remain anonymous.
Monte Carlo Simulations of the Electoral College Based on Rasmussen
Research's State Surveys.
In each state, independent random choices are made for each of the three
candidates from probability distributions with means chosen as indicated
below and with an appropriate standard deviation about half of the margin
of error. This probability distribution is the same one from which the
margin of error is computed. The standard deviation for Nader was
appropriately scaled to account for his smaller percentage. Finally, the
candidate's results are compared and that state's Electoral Votes awarded
to the one with larger percentage.
There are two sets of simulations below in which the means are chosen in
two different ways. In the first case the mean of the distribution for
each candidate and state is the latest survey result, as of Sunday, November 5.
In the second case, the mean is the survey result predicted for November 7
obtained by adding a correction to the last survey result. The correction
is the change in the candidate's national trend line between the date of
the last survey for that state and a (slightly) extrapolated November 7
value. The trend lines were constructed by eye, and arrive at election day
with Bush having 47.5% and Gore 42.5%. The effect is to increase the mean
of Bush's distribution relative to Gore's for states last surveyed before
the third debate.
The results of 100 simulations of each kind are presented here. Two totals
are presented for each candidate in each simulation. The first is the
national total, and the second is the total for the ten tossup states where
the difference between the candidates is currently 3 points or less.
At the end of each list of trials the maximum, minimum, and average over
the 100 trials are reported for each candidate. Though 270 Electoral Votes
are needed to claim the Presidency, in none of the 200 simulations here did
Gore win over 257. The average result is Bush 340 and Gore 200.
The largest variation is caused by the California result. Since California
has 54 electoral votes, and the rest total 94, if one candidate's tossup
total is greater than 94, that candidate has won California. This is true
for Bush five times in the first set of 100, and seven times in the second
set. Note that the maximum of the tossup totals for Gore is 132. This
means that Bush won some trials in which Gore got nearly all of the 148
tossup votes. Half of those times, Bush's total was 400 Electoral Votes or
more.
Tossup States
California 54
Delaware 3
Minnesota 10
Washington 11
Michigan 18
Tennessee 11
Arkansas 6
Florida 25
New Mexico 5
West Virginia 5
Simulations Based on Most Recent Survey
Trial# GoreEV BushEV GTossEV BTossEV
1 183 355 69 79
2 207 331 93 55
3 156 382 64 84
4 225 313 111 37
5 202 336 81 67
6 211 327 97 51
7 212 326 94 54
8 217 321 103 45
9 223 315 109 39
10 199 339 85 63
11 194 344 80 68
12 221 317 107 41
13 214 324 93 55
14 218 320 104 44
15 203 335 89 59
16 203 335 89 59
17 224 314 106 42
18 203 335 89 59
19 186 352 72 76
20 223 315 102 46
21 184 354 70 78
22 173 365 59 89
23 197 341 83 65
24 214 324 85 63
25 199 339 85 63
26 215 323 101 47
27 182 356 68 80
28 200 338 86 62
29 200 338 86 62
30 193 345 79 69
31 189 349 75 73
32 197 341 83 65
33 152 386 38 110
34 186 352 72 76
35 182 356 68 80
36 214 324 89 59
37 221 317 107 41
38 188 350 96 52
39 188 350 74 74
40 221 317 107 41
41 224 314 99 49
42 186 352 72 76
43 217 321 107 41
44 214 324 122 26
45 192 346 78 70
46 130 408 16 132
47 246 292 132 16
48 208 330 94 54
49 192 346 78 70
50 192 346 78 70
51 213 325 114 34
52 171 367 57 91
53 208 330 83 65
54 218 320 104 44
55 173 365 59 89
56 171 367 57 91
57 217 321 103 45
58 214 324 110 38
59 222 316 108 40
60 231 307 117 31
61 183 355 69 79
62 200 338 86 62
63 168 370 54 94
64 246 292 121 27
65 203 335 107 41
66 205 333 91 57
67 186 352 72 76
68 227 311 102 46
69 203 335 89 59
70 186 352 79 69
71 207 331 93 55
72 199 339 85 63
73 196 342 71 77
74 138 400 24 124
75 149 389 35 113
76 125 413 11 137
77 199 339 85 63
78 228 310 114 34
79 198 340 80 68
80 207 331 93 55
81 176 362 62 86
82 208 330 94 54
83 193 345 68 80
84 200 338 79 69
85 217 321 103 45
86 184 354 70 78
87 200 338 86 62
88 216 322 102 46
89 200 338 86 62
90 179 359 65 83
91 223 315 86 62
92 199 339 85 63
93 221 317 107 41
94 204 334 90 58
95 215 323 101 47
96 168 370 54 94
97 176 362 62 86
98 197 341 83 65
99 195 343 81 67
100 210 328 96 52
max 246 413 132 137
min 125 292 11 16
avg 198.92 339.08 84.57 63.43
Simulations Based On Date-Adjusted Data
Trial# GoreEV BushEV GTossEV BTossEV
1 200 338 86 62
2 207 331 93 55
3 221 317 107 41
4 171 367 79 69
5 216 322 98 50
6 211 327 97 51
7 211 327 97 51
8 223 315 102 46
9 179 359 65 83
10 203 335 85 63
11 197 341 83 65
12 246 292 132 16
13 199 339 85 63
14 203 335 89 59
15 230 308 116 32
16 235 303 121 27
17 196 342 78 70
18 147 391 29 119
19 192 346 78 70
20 218 320 104 44
21 153 385 39 109
22 203 335 89 59
23 209 329 91 57
24 197 341 83 65
25 221 317 107 41
26 189 349 75 73
27 207 331 93 55
28 188 350 74 74
29 192 346 78 70
30 179 359 65 83
31 190 348 65 83
32 211 327 97 51
33 158 380 44 104
34 200 338 86 62
35 208 330 94 54
36 197 341 83 65
37 224 314 110 38
38 212 326 83 65
39 198 340 73 75
40 216 322 91 57
41 192 346 78 70
42 195 343 81 67
43 202 336 88 60
44 132 406 14 134
45 218 320 104 44
46 177 361 63 85
47 247 291 129 19
48 204 334 90 58
49 246 292 132 16
50 195 343 81 67
51 257 281 132 16
52 178 360 64 84
53 243 295 129 19
54 150 388 29 119
55 223 315 109 39
56 205 333 91 57
57 127 411 13 135
58 201 337 78 70
59 239 299 107 41
60 186 352 72 76
61 246 292 132 16
62 217 321 103 45
63 214 324 96 52
64 210 328 96 52
65 200 338 86 62
66 211 327 97 51
67 138 400 24 124
68 200 338 86 62
69 244 294 107 41
70 182 356 68 80
71 225 313 111 37
72 203 335 89 59
73 178 360 64 84
74 212 326 91 57
75 190 348 67 81
76 218 320 104 44
77 189 349 75 73
78 193 345 79 69
79 203 335 89 59
80 218 320 93 55
81 221 317 96 52
82 205 333 91 57
83 192 346 78 70
84 218 320 104 44
85 229 309 115 33
86 225 313 111 37
87 211 327 97 51
88 213 325 88 60
89 210 328 96 52
90 217 321 92 56
91 234 304 109 39
92 209 329 95 53
93 228 310 114 34
94 215 323 101 47
95 189 349 75 73
96 223 315 109 39
97 237 301 123 25
98 235 303 121 27
99 186 352 72 76
100 197 341 83 65
max 257 411 132 135
min 127 281 13 16
avg 204.59 333.41 88.55 59.45
-------------------------------------------------------------------------
POLITECH -- the moderated mailing list of politics and technology
You may redistribute this message freely if it remains intact.
To subscribe, visit http://www.politechbot.com/info/subscribe.html
This message is archived at http://www.politechbot.com/
-------------------------------------------------------------------------
 By Date 
By Thread
Current thread:
- FC: Computer simulation predicts George W. Bush wins election Declan McCullagh (Nov 07)
|
Intro
