I.  Purpose
This formula's intent is to give a general sense of how the President would fare if the elections were held today.  As we move closer to election day, its usefulness as a prediction tool will increase.

II.  Objectivity
In developing the formula, I made every effort to be objective, rather than biased toward Bush.  Yes, I am a huge supporter of the President, but this formula does not, in any way, falsely inflate Bush's standing.

III.  Polls
I try to be thorough and objective in gathering polling data.  I use the Polling Report and RealClearPolitics for national polls and DC Political Report and RealClearPolitics for state polls.   However, there may be times when I overlook a poll.  If this happens, I urge my readers to alert me to such polls so that I can include them in the projection.  Though some have implied it, I do not pick and choose which polls will be used.  Even if a poll shows Bush's numbers to be a lot lower than others, I still use it.  There are types of polls I don't use, such as John Zogby's Presidential job performance polls.  He asks participants to rate the President's job performance as excellent, good, fair, or poor rather than asking if they approve or disapprove of the job he is doing.

The projections can be broken down into three distinct, independent factors.  I list them here and explain them in excruciating detail below.

• The Baseline Adjustment derived from the 2000 election result.
• The National Adjustment resulting from a compilation of various national polls.
• The State Adjustment calculated from the results of state polls.

The first task undertaken was to use the results of the elections of 2000 to arrive at a baseline adjustment which would reflect a zero positive or negative skew of the polling data.  In other words, what would be Bush's votes total if his job approval, head-to-head and right track numbers were all exactly neutral (that is, 50%, in most cases)?  In theory, he would get exactly the same votes as his challenger.  Before I go any further, I need to address the Nader candidacy of 2000.  Exit polls indicate 20% of Nader voters would have voted for Bush, 50% for Gore, and 30% would have stayed home.  So, Nader's run, as we all know, hurt Gore's chances.  Because I'm loathe to model this projection in a way that would favor Bush, I decided to allocate Nader votes to Bush and Kerry according to this exit polling data.  In so doing, I discount Nader's influence in this year's election while boosting Kerry's numbers just a little, assuming Nader does go the distance.  The result of all this number manipulation shows that, all polls being neutral, Bush and Kerry would both get 49.085%.  This is the starting point for adding in the current set of polls and reflects an adjustment of approximately 1.215% from Bush's total vote percentage in 2000.  Whew!  Fuzzy-brained, yet?  I know I am!

There are three categories of statistics that make up the national polling segment of the formula I use to project the outcome of the 2004 elections.  One requirement that any poll must meet before being included is that it must be timely.  Any poll older than a month is not used.

The big push from my readers to include state polls in my calculations is being rewarded.  I've added a weighted factor of state poll results to the projection formula.  Here are the rules:

• At most, three polls from a state are used.
• In the interest of currency, if there is a poll or polls that were conducted less than 14 days prior to the update, then only such polls will be used.  For example, if a poll was conducted in Iowa a week before the update, and another was conducted three weeks before, only the newest poll will be used.
• However, if the latest polls were conducted more than 14 days prior, they will be used, up to the maximum of three.
• In no case will a poll over 30 days old be used.
• State polling results will be weighted as follows:
    For polls less than 14 days.
      One poll used will be weighted half the national polling adjustment
      Two polls used will be equal to the national polling adjustment.
      Three polls used will weighted more than the national polling adjustment by one half.
    For polls more than 14 days.
      One poll used will be weighted one quarter of the national polling adjustment
      Two polls used will be half the national polling adjustment.
      Three polls used will be weighted 75% of the national polling adjustment.

• ((JobApproval * 7) + (HeadtoHead * 5) + (RightTrack * 2)) / 14

The result is used in conjunction with the state polling adjustment to determine each state's projected result.

Here's an example using Iowa: