First, I'd like to start out by saying that national polls are fundamentally flawed not only in the way the data is reported but also in collection and calculation of the data. Generally speaking, although the number may vary, about 600 people are polled for a national poll and about 500 for a statewide poll. Both generate a margin of error of about + or - 4%. Data from 600 people is collected and analyzed to interpret how the entire electorate will vote. Questions are asked about basic personal information. Things like state of residence, sex, race, education, income. In addition, surveys also collect data regarding your likelyhood of voting and parety affiliation. Are you a registered voter? Did you vote in the last presidential election? In 2000? Are you a registered Democrat/Republican? Do you see yourself as more liberal or conservative?
All of these questions help place the people surveyed in a category for which you (and several others) will represent. Census data is then used to extrapolate the answers from all respondents across all demographic groups and weight them to the population of each group. Put simply, in a 600 likely voter poll, if 300 vote for McCain and 300 vote for Obama, it does not (and likely is not) a tie. Depending on who was polled, and how each group of indiviuals is weighted, the actual number could sway wildly.
In addition, you have the problem that we don't actually elect a president nationally. Elections are run in all 50 states (and the District of Columbia). All states then assign electors based on the winner of that state. Each state (except Maine and Nebraska we'll talk about them in a minute) assign all of their electoral to the winner of the popular vote in their state. The number of electors is based on the total number of Senators and Representatives in each state (Washington D.C. does not have any Congressmen but are awarded 3 electoral votes). In Maine and Nebraska, the process is broken down even further. The winner of these states is awarded 2 electoral votes while the winner of each congressional district within each state hs a separate vote. It should be noted that both states have been using this method since 1968 and have never actually sent electors from both parties.
As such, only statewide polls are used in the calculation for both the electoral vote and the popular vote.
Polls Used
All polls are considered at one point or another in making a projection. A minimum of 3 polls are used in each state (where available). All likely voter polls are considered a poll, while other poll types (registered voters, adults, etc.) are considered 1/2 of one poll. The reason for this is that voter models are likely excluding many voters that will be voting for the first time, or the first time in a long time. There has been such an interestin this election that it is difficult to exclude anyone from polling data. Also, any polls that are sponsored by either of the parties is considered 1/2 of a poll. This is to limit the influence of parties in the projection process. Also, any poll that is less than 2 weeks old is considered in the projection even if this will result in more than 3 polls being included. Finally, only 1 poll (or 1/2 poll) will be considered from a single source. This is so that no individual pollster has more influence than another.
In the sample data above, the Civitas/TelOpinion dated 6/11-6/13 is used as 1/2 poll because it was registered voters (1/2 poll). Rasmussen dated 6/10 is counted as a full poll (1&1/2). PPP is counted as 1/2 a poll because it was sponsored by the Democrats (2 polls). Finally, the Survey USA poll dated 5/17 is used as a full poll (3 polls). All other polls are older than 2 weeks and the minimum requirement is filled so they are ignored.
Calculation for voting percentages
In order to make the math more simple each full poll is counted twice and each 1/2 poll is counted once. As such, the calculation begins below.
McCain
Civitas 45%
Rasmussen 45%
Rasmussen 45%
PPP 43%
Survey USA 51%
Survey USA 51%
Average 46.67%
Obama
Civitas 41%
Rasmussen 43%
Rasmussen 43%
PPP 40%
Survey USA 43%
Survey USA 43%
Average 42.17%
McCain leads Obama 46.67% to 42.17%, however, 11.16% are undecided.
First, as a general rule of thumb all third party candidates add up to 1% of the and are taken off. The remaining 10.16% undecided voters are apportioned based on the ratio of votes in the first calculation. As such, McCain receives 5.34% of the undecideds while Obama receives 4.83%. The end result is:
McCain 52.01%, Obama 46.99%, Others 1.00%
Calculation for Popular Vote and Voter Turnout
Using voting age population data from 2000 and 2004 for each state along with actual voter totals in each state from 2000 and 2004 I estimated the actual voter turnout rate in each state. I then used projected voting age population data in each state for 2008. I then estimated that in most states that are likely to be competitive, the would be an increase of 10% in the turnout from the 2004 election. That means that if the turnout was 60% in a given state and that state was expected to be competitive the projected turnout for 2008 was 66%. In non-competitive states I bumped the voter turnout to 5%. So a similr state would have a voter turnout of 63% vs. 66%. After that it's just adding everything up!
Safe, Lean, Barely & Toss-Up
State are considered safe if the total difference in voting preference is 10% or more.
Leaning is from 6-10%
Barely from 2-6%
Toss-ups are within 2%.
If anyone has any questions or suggestions, I'm open to anything. Thanks.
No comments:
Post a Comment