More times than not polls conducted even by reputable survey research organizations have different results. Public opinion professionals have a stable of standard answers to account for discrepancies in poll numbers including methods of sample selection, question wording, question order, use of weighting, and the old standby, margin of error. The recently released polls on the New York Senate race from The New York Times /CBS News and the Marist College Poll need additional clarification.

On Thursday (9/21) New Yorkers were presented with poll results from two surveys on Clinton vs. Lazio that varied greatly. The New York Times/CBS News findings gave Clinton a 9% lead over Lazio among likely voters, fueled by a shrinking advantage in the suburbs surrounding New York City where, according to this poll, Lazio had the support of 48% to 43% for Clinton among these voters.

In contrast, the Marist Poll also released that morning showed Clinton and Lazio tied overall among likely voters with Lazio holding a 2:1 lead in the suburbs. What’s a poll watcher to think?

First, The New York Times/CBS News poll was conducted over a 6-day period, commencing on September 14^th, the night following the candidate debate in Buffalo. Their poll results captured the maximum shock effect from Lazio’s march across the stage. The Marist Poll began on September 19^th and was finished the following night, well after the debate dust had settled. Interestingly, the Hillary Clinton support in both polls was similar, the variation was in the Lazio number with The New York Times/CBS News numbers showing a lower Lazio figure and a higher undecided in the head-to-head matchup. The Marist Poll had a higher negative rating for Lazio with those voters having a negative impression of him growing from 32% to 38% (the Clinton backers now understood why they weren’t voting for Lazio), but the soft Lazio support had returned to him in the tossup question by the time Marist was in the field. Although released on the same day, the polls covered different days and the difference mattered.

Second, the two polls used voter screens of likely voters that produced very different samples. According to The New York Times/CBS News poll, 88% of the registered voters were categorized as likely. In the Marist Poll, 67% of registered voters are expected to vote. In 1996, the previous presidential election, turnout was 63% of registered voters. The importance of this calculation is that Lazio fares better the lower the turnout. The Marist Poll may view the race as closer in part because of the share of likely voters in the poll.