The Campaign Spot

Why Are So Many Pollsters Oversampling Democrats?

My regular correspondent Number-Cruncher checks in, groaning about the latest Pew poll and that organization’s strange habit of including an unrealistic percentage of Democrats in their sample.

The latest one from Pew poll is a shining example of why our side gets so frustrated with polls. Every time a Pew poll comes out, the numbers appear out of whack. Of course if you are a number-cruncher and look to the cross-tabs, the results are clearly flawed. Pew, to its credit, tells us its history since 1988. Basically in 1988 they did a good job, calling the race almost perfectly, possibly even overestimating Bush support by 0.4% (keep in mind they round so 50-42 could be 7.6%). But since then, their results have been downhill.

Starting in 1992, EVERY Pew poll appears to lean to one directionalways towards the Democrat, and by an average of more than 5 percentage points. Worse this is a reflection of the “final” poll which even the Democratic firm, Public Policy Polling, usually gets right.

October 1988 — Bush 50 Dukakis 42; Actual Result Bush +7.6 (Call this one spot on.)

Late October 1992 — Clinton 44 Bush 34; Actual Result :Clinton +5.5 (Skew against Republican candidate +5.5)

November 1996 — Clinton +51 Dole 32; Actual Result Clinton +8.5 (Skew against Republican candidate +10.5)

November 2000 — Gore 45; Bush 41 (Skew against Republican Candidate +3.5)

November 2004 — Kerry 46; Bush 45 (Skew against Republican Candidate +3.4)

November 2008 — Obama 50 McCain 39 (Skew against Republican + 3.8)

After being wrong in the same direction so consistently, wouldn’t you think that Pew might attempt to adjust their sampling techniques to adjust their techniques to avoid under-sampling Republican voters?  Keep in mind the polls I have highlighted are the last polls in the race.  I find it interesting that not one of their poll statisticians came out and said, ‘Boss, these results look whacked out because the electorate is going to be more than 24 percent Republican, and self-identified Democrats aren’t going to outpace Republicans by 9 percentage points.’  The Democrats couldn’t even reach that margin in 2008 . . . and you wonder why so many people think Obama is going to win.  Didn’t Einstein once say the definition of insanity was “doing the same thing over and over again and expecting different results”.  So I ask are the people at Pew insane or just biased?


On another polling annoyance, I turn to your favorite punching bag PPP-D.  Here is the trend with PPP-D.  They do a terrific job polling our primaries (this is not sarcasm).  Their closing polls are usually okay.  But their tracks, should be involving a Republican versus Democrat race be viewed with healthy skepticism.  The latest poll release in NC showing Obama +1 indicated a Democrat advantage of +12.  I don’t know what they are smoking over there but I want some (just kidding) . . .

As I tracking every public poll, I have found it amazing on how many pollsters are oversampling Democrats.  On average, every poll is indicating a partisan ID similar to 2008.  Based on my analysis, the average gap is +6.4% Democrat which compares to the +7% which was in 2008.  Alas, for the RCP poll average to be correct you have to assume the self-identified party preference turnout will be similar to 2008, if the turnout is similar to 2004 or 2000 or 2010, Obama’s polling leads may as well be part of his “story telling.”  For  the pollsters that look at preference, Rasmussen has indicated that the self-identified party ID (for Adults – not likely voters) is about +1.4% Republican.

Based on my track if the election were held today Romney wins by 5 percentage points.

I’ll make a quite modest defense of PPP on North Carolina, in noting that there are a lot of conservatives down there who are still registered Democrats. The margin in the 2008 exit polls was 42 percent Democrat, 31 percent Republican. According to the latest voter-registration statistics, the state has 2,729,427 registered Democrats (43.4 percent) 1,972,428 registered Republicans (31.3 percent) and 6,284,428 total registered voters. Of course, not all registered voters actually cast ballots each year, and quite a few of those registered Democrats are pretty conservative. So yes, D+12 is probably too disproportionate, but the parties will probably not be evenly split on Election Day, even if Romney wins handily.


The Latest