Home     About Us     Contact Us
Our Purpose

The purpose of LatestPollResults.com (“LPR”) is to present an unbiased, statistically-based probability analysis of the Electoral College. Our projections are based exclusively on data from the latest poll within each of the 50 states, plus DC. Starting with the 2000 election, we've posted our daily projections on LatestPollResults.com, during the months leading up to Election Day.

Our Track Record

We correctly projected wins for: Bush in 2000, Bush in 2004, Obama in 2008 and Obama in 2012. In 2010, LPR extended its probability analysis to the Senate races and correctly projected that the Senate would remain controlled by the Democrats in 2010 elections and by the Democrats in 2012.

The Polls

All data is based on the latest poll results, as published by a variety of polling agencies, colleges and universities. All polls contain the following information:

  1. Polling agency and party affiliation, if any.
    LPR accepts the poll findings from every pollster, regardless of whether they may have a party affiliation.

  2. The date(s) the poll was conducted.
    If the poll was conducted on more than one day, LPR considers the midpoint of the dates as the “polling date.” LPR selects the poll with the most recent “polling date” within each state.

  3. The profile of the poll respondents.
    Pollsters generally select their findings to be representative of: “Adults,” “Registered Voters,” or“Likely Voters.” “Adults” include all voting age respondents, regardless of whether they are registered to vote. “Registered Voters,” include all registered voters, regardless of whether they are likely to vote. LPR only uses polls where results are presented based on the pollster's determination of those respondents who are “Likely Voters.”

  4. The sample size and the % vote reported for each candidate.

Calculating the probabilities for each candidate in a given poll

With its findings, every pollster reports the poll's margin of error, MOE*. For example, assume that a poll reports that Candidate A was polled at 46%, Candidate B at 38% and that the MOE for the poll is 4%. The pollster would conclude that the “true” % for Candidate A is between 50% and 42%, i.e. 46% plus or minus the MOE of 4%. Similarly, the “true” % for Candidate B is between 42% and 34%. Since Candidate B's % could NOT be higher than Candidate A's %, the conclusion is that the lead for Candidate A is “statistically significant.” The MOE is based on a 95% confidence level which means that there is a 95% probability that Candidate A's lead, for this poll, is between 0% and 16%. There is also a 2.5% probability that Candidate A's lead is greater than 16% and a 2.5% probability that Candidate B is actually ahead. Candidate A's lead being “statistically significant” at the 95% confidence level is equivalent to a 97.5% (or higher) probability that Candidate A is leading Candidate B, on the date(s) the poll is conducted.

In the above example, if Candidate B was polled at 41%, this means that the “true” % for Candidate B is between 45% and 37%. But since the “true” % for Candidate A is between 50% and 42%, Candidate A's lead over Candidate B, is NOT “statistically significant” at the 95% confidence level. Even so, there is a statistical probability that Candidate A is leading Candidate B. LPR calculates the statistical probability associated with any lead and the MOE for the poll. In this 46%-41% example with a 4% MOE, the probability that Candidate A is leading Candidate B is 89.4%.

*Note: The MOE reported for the poll is often rounded, inaccurate or imprecise. From the sample size and % from each candidate, LPR calculates the “exact” MOE and the probability of winning for each candidate in each state,

Calculating the probabilities for controlling the Senate

From each candidate's probability for winning in a given state, LPR determines the “safe” states, where one of the candidates has a “statistically significant” lead and thus the probability of winning for that candidate is at least 97.5%. The remaining states are the toss-up states.

Since each toss-up state will ultimately vote for either the Democrat or the Republican, there are two possible outcomes for each state. With 20 toss-up states, there are 2 x 2 x ... x 2 possible outcomes, or in mathematical terms, 2 raised to the 20th power, written as 2^20 = 1,048,576 possible outcomes. For each of these outcomes, LPR calculates the number of seats captured by the Democrats and by the Republicans and the probability associated with the outcome. The sum of the probabilities for all outcomes where the Democrats capture at least 50 votes is the overall probability of the Democrats controlling the Senate. Similarly, the sum of the probabilities for all outcomes where the Republicans capture more than 50 seats is the overall probability of the Republicans controlling the Senate.