Tons of media coverage is devoted each election cycle to predicting the outcomes of upcoming races. In the weeks before the recent vote, a swarm of polling firms scoured the Maine electorate hoping to divine voter intentions in the high-profile presidential and senate elections. Each arrived at a different assessment of the state of Maine's major races.

Any pollster will tell you that a poll is meant to provide a snapshot of the race during the time it was conducted and is not always the best predictive tool. The point of conducting a poll, however, is to give some insight into the actual results, so checking to see how close those snapshots came to the final verdict is a worthwhile exercise.

To see which poll came the closest, I collected results from each firm that conducted a survey during the last month before the vote. I also included the projections from Pollster.com, FiveThirtyEight.com and Real Clear Politics, three websites that weight and average poll results, and are designed to give a more accurate picture of the electorate than a single poll. Just for fun, I also included the results of the student mock election conducted by the Maine Secretary of State and the average of the predictions made in the MainePolitics.net betting pool, which was open to anyone who can click a mouse.

Most pollsters only released results for the presidential and US Senate races, so those are the figures I used for comparison. I chose to ignore any uncommitted or third-party voters reported and instead focused on the horse race, comparing the margins of victory each pollster projected against the actual margin on election day.

The results are very interesting.

Pan Atlantic SMS Group, a Maine research company, was the only pollster to report results with a margin of error that included the actual margin of victory on election day. They also came the closest in matching the actual result of the senate race, where Susan Collins won by a larger margin than almost anyone predicted. Market Decisions, another Maine firm, was the second most accurate pollster.

Next came the polling averages and the Research 2000 poll commissioned by Daily Kos, and then the results from YouGov, which does online surveys rather than telephone interviews, and Rasmussen and Survey USA, which both use robocalls. Critical Insights, which apparently does not weight their data by demographics, came in last place among the three Maine firms. The Democratic Senatorial Campaign Committee poll, which had a large sample size and pegged the presidential vote perfectly, was way off on the Collins/Allen race. This poll was released to give the Allen campaign a boost and its likely they knew it was an outlier at the time.

Neither the betting pool average or the mock election results were particularly predictive, but some of the individuals who made predictions in the pool came closer than any of the polls. I'll be announcing the full results of those predictions at MainePolitics.net over the next few days.

The final verdict is that Pan Atlantic SMS is the best Maine pollster of 2008, perhaps because of their familiarity with the state and their in-house professional interviewers. In addition to their accuracy in the presidential and senate races, Pan Atlantic was one of the few firms to poll the other congressional races and two of the statewide referendums, making relatively accurate predictions in those races as well.

The views expressed on this Web site are those of the authors alone and do not necessarily represent the views of Down East Enterprise or its employees.