Pew Poll: Debunking Obama's 15 Point Lead

Today, Pew Research released a poll of the Presidential race with pretty shocking results. The poll said that Obama leads registered voters by a staggering sixteen percentage points (52%-36%) and likely voters by fifteen points (53%-38%).

These results are staggering and difficult to believe, especially when most other national surveys put Obama's lead at a much smaller number. One reason is probably Pew's party ID that is heavily skewed toward Democrats. Pew's tables are a bit confusing, so it's difficult to figure out their exact Party ID figures for their sample of likely voters.

If I am reading the Pew results correctly, when voters were asked to self-identify with a party, 24% said Republican, 39% said Democrat, and 32 said Independent. That is a staggering 15% difference which would be a shocking and unprecedented difference. In 2004, party ID was even in the electorate according to exit polls - 37% Republican, 37% Democrat and 26% Independent. Most state exit polls in 2006 showed either little change or a slight shift toward Democrats. I expect party ID to move in the Democratic direction given the very tarnished GOP brand, President Bush's dismal approval ratings and the financial crisis, but not so drastically.

Oddly, Pew later lists the number of voters in their sample: 361 registered Republicans, 490 registered Democrats and 411 Independent voters. The problem? When you add these figures up, you get 1262 voters, which doesn't match the number of voters in either the registered voter or likely voter pool. Among those 1262 voters, Republicans are 28.61%, Democrats are 38.83% and Independents are 32.57% (the likely voter pool only had 1198 people, so these figures clearly are not accurate for the likely voter results). A ten point gap still seems overly wide and could account for Pew's very strange results. If anyone can shed some light on Pew's methodology, feel free to comment below.

If there is a conflict between self-identification and actual voter registration, I would tend to rely on self-identification for the purposes of analyzing a poll anyway. Voter registration is a ministerial act that can sometimes be the result of historical artifact. For example, people often register as independents and later come to identify with one of the parties, whereas others change their views (and party) over many years. Self-identification is also useful for poll comparisons, as exit poll data uses that metric rather than actual voter registration.

The bottom line: I highly doubt Obama is ahead by 15 points, just as I doubt he is only ahead by 2% (according to Gallup's traditional likely voter model). To put it in perspective, Real Clear Politics' average of New Jersey polls gives Obama an 18 point edge. Are we to believe that Obama's lead nationally will be almost identical to his winning margin in New Jersey? Highly doubtful.

UPDATE: Welcome to Campaign Spot readers. You may also enjoy reading our State of the Race column, which handicaps the Presidential race, along with all Governor, Senate, and House races. We also run daily State of the Day columns with in-depth profiles of political races each state. We'll get started on handicapping the 2009 & 2010 contests shortly after Election Day.