One thing I like is that the newspaper linked to a detailed summary of the questions, methodology and results. This is no doubt because they conducted the poll. They also included a detailed summary of their methods. To me, linking to primary sources is a easy way to take advantage of the web and give a more comprehensive picture.
But polls are not always accurate. To evaluate the Times and CBS's poll, I put it to a list of 20 questions (minus the irrelevant ones) from the National Council on Public Polls. While they aren't famous names, the article was written by the cofounders of the NBC/Associated Press Poll, Sheldon Gawiser and G. Evans Witt.
Who did the poll and who paid for it? The New York Times and CBS News, both respected organizations with a history of doing polls.
How many people were interviewed? 1,301 were reached and 1,170 were registered to vote — and thus able to respond. Polling that many results in a respectable margin of error of plus or minus three percent.
How were people chosen? Numbers from exchanges that served residential neighborhoods were dialed at random. Pollsters also called cell phones in order to ensure the poll was representative of all Americans.
Are the results based on the answers of all the people interviewed? Yes. All questions were directed at every respondent.
Who should have been interviewed and was not? Pollsters used multiple attempts to try and reach likely voters. They called back at different times and different days to maximize the odds of reaching someone.
When was the poll done? The attack on the American embassy happened in the middle of the polling, on Sept. 11, though most Americans wouldn't have heard about the news or adjusted their opinion on the candidates yet. The events may have had some effect on the results, but probably not a significant one, considering the 3 percent margin of error. Nate Silver explores this issue in more detail, including the advantages Obama has.
How were the interviews conducted? The survey was done by telephone — generally considered a reliable method — and was not a robo poll.
What other kinds of factors can skew poll results? Weighing was done to ensure the people asked were representative. According to the detailed summary, many questions are the same as those done in the past, suggesting they have a good track record.
In what order were the questions asked? Questions about the candidates were separate from those about specific issues, avoiding the problem of one question skewing responses to another.
What other polls have been done on this topic? A Gallup poll from Tuesday shows President Barack Obama leading Mitt Romney, 50-44, compared to the Times' 49-46. Gallup's tracking poll, which runs through Saturday, showed 48-45. So the overall results are consistent with what you'd expect.
Unsurprisingly, the poll does well on every question. The pollsters behind it have plenty of experience conducting polls, and it shows. I'm also happy that the reporting focused less on the horse race numbers — who's leading whom — and more on other findings, like that Obama's doing better on the economy.