Imagine three polls. All of them covering Monday, Tuesday, and Wednesday, which is to say the Republican convention before Sarah Palin's speech. One, the Gallup poll, tracks a sample of 2,771 registered voters, and has a margin of error of 2 percent. Another, the Rasmussen poll, tracks 3,000 voters and has a similar margin of error. A third, the CBS poll, sampled fewer than 800 voters, and has a margin of error of four percent.You're a political reporter. You work for a news organization. The Gallup poll shows a seven percent Obama lead, the Rasmussen poll shows a five percent Obama lead, and the CBS poll shows a tie. What's your headline?Well, if you're David Paul Kuhn, and you work for The Politico, your headline is, "Poll: Democratic bounce gone, race tied." But wait! You're not a liar. Your third paragraph tells readers they've been misled. "Other polls have failed to show the same tightening of the race found by CBS. Neither the Gallup or Rasmussen daily tracking polls have registered a significant drop in Obama’s support from his post-convention bounce numbers. The Gallup tracking poll, for example, still has Obama ahead of McCain 49 percent to 42 percent." You're just a truth shaper. A context manipulator. And even though that headline might have been a direct contradiction of the available evidence, it was for a good cause:
As the wise men say
, asking someone to tell you the truth, and asking them to tell you as much truth as possible while still make the maximum in profit, are not the same things. Kuhn, of course, is safe: It's a good bet that Obama's lead will melt away this week, and McCain may even pull slightly ahead. Convention bounces are broadly predictable things. But this is the danger of a political environment with too many polls mixed with a media environment hungry for drama: The most prized commodity become outlier polls, because they make for the better story. Rather than making it easier for the media to give an accurate snapshot of the race, they make it easier for the media to give an inaccurate, but more exciting, snapshot of the race. And so too with this article. Given the universe of data available last night, the only conclusion was that the CBS poll was showing an odd result, and given its smaller sample size and wider margin of error, should be treated skeptically. Instead, it was the headline of the piece. More traffic that way.Incidentally: This seems like as good a time as any to say that you should be ignoring polls for the next week, just as you should have been ignoring polls for the last week. To get an idea of where the race sits, you need to wait till both conventions are incorporated into the polls, and then for the cycle of polls after that to see how voters feel when the coverage has settled down.
As the wise men say, asking someone to tell you the truth, and asking them to tell you as much truth as possible while still make the maximum in profit, are not the same things. Kuhn, of course, is safe: It's a good bet that Obama's lead will melt away this week, and McCain may even pull slightly ahead. Convention bounces are broadly predictable things. But this is the danger of a political environment with too many polls mixed with a media environment hungry for drama: The most prized commodity become outlier polls, because they make for the better story. Rather than making it easier for the media to give an accurate snapshot of the race, they make it easier for the media to give an inaccurate, but more exciting, snapshot of the race. And so too with this article. Given the universe of data available last night, the only conclusion was that the CBS poll was showing an odd result, and given its smaller sample size and wider margin of error, should be treated skeptically. Instead, it was the headline of the piece. More traffic that way.Incidentally: This seems like as good a time as any to say that you should be ignoring polls for the next week, just as you should have been ignoring polls for the last week. To get an idea of where the race sits, you need to wait till both conventions are incorporated into the polls, and then for the cycle of polls after that to see how voters feel when the coverage has settled down. 
