reply to post by Xcalibur254
I still think your logic is off. Sorry in advance for the long winded post.
Lets take your ice cream example (btw, as a chubby man this helps me). To anyone else reading this I apologize for the repititiveness, and for
sounding like fat man.
A niche flavor like Rocky road may have let say 1 percent of the total people like it most, whereas maybe vanilla is liked most by 50 percent. You
are right that in a small store or sample size it is more likely to get results that don't exactly match up with the overall data. But your
asserstion that they would more likely be rocky road supporters is false. It would be equally likely you would see no Rocky road supporters.
Regardless of the store your polling people at, the more people you poll, the closer you would get to the actual trend. So polling one Wal Mart with
50 people would be the same as polling 25 small stores with 2 people a piece. This is unless there is a differnce in the types of people that go to
these stores.
However, when we switch now to the election, there is no difference in the precincts other than some have a lot of people in them, and some have few.
I don't see why that changes how people would vote, which you seem to be asserting. The demographics seem to be the same no matter the size of the
precincts, in other words there are large rural precincts and small urban ones.
So instead of looking at wal mart vs speciallty stores, this example would be more apt. Some people were polled at wal marts that a lot of people
shopped at, some at wal marts few people shopped at, some at specialty stores with a lot of people, and some at specialty store with few people
shopping. In this example, wal marts are like urban precincts and the other stores are rural. Just like with the precints, despite rather its urban
or rural, both types have some stores with a lot of people, and some stores with few.
Whew, ok. Now what would've happened is that at all the stores with a few people, regardless of if they were a specialty store or wal mart,
miraculously had tons of rocky road supporters. If you were just looking at this data you would say "Ok, despite the different types of stores,
people seem to really like Rocky road". Then when you move on to the stores with a lot of people, all of the sudden almost no one likes rocky
road.
In fact, the number of people that liked it would have to be significantly lower than the overall percentage of the total people who like rocky road
in order to make up for the large amount the liked it previously. This makes no sense.
Now back to the elections. Regardless of political ideology, there should not be significant changes in a candidates votes purely because of precinct
size. If there were changes based on urban areas vs rural areas, taht would make sense, people in the country have different views than city people.
But this was not the case. The change was all based on size of precinct, not demographics.
Lets say a large precinct was large and had 500 people vote, and it was in an urban area. In the same urban are there also happen to be 10 smaller
precincts, which had 50 people a piece vote. There both in the same area, so in that vote, one would expect the results of the larger precinct to be
very similar to the surrounding precincts combined.
Both the one precinct and the ten precincts combined are from the same area demographically, and both have 500 people total, so the numbers for each
candidate should be around the same. Granted, an individual small precinct might stray from the curve, but as your sample size increases by adding
more small precincts, this should work itself out. In other words there is a good chance one small precinct would have an unusually large amount of
Paul supporters, but there would be an equal chance there would be one with a high amount of Romney supporters.
The only way your model works is if there is some reason small precincts and large precincts have inherently different types of voters, but seeing as
how they are from the same areas (again large urban precincts and small ones, same with rural), I don't see why this would be the case. As shown in
my previous post, without and explanation for this, the most likely explanation is vote tampering.
On a side note, I really appreciate the conversation and look forward to your reply.