WaPo/ABC poll: Check the sample
posted at 9:45 am on September 24, 2008 by Ed Morrissey
Two questions arise from the Washington Post/ABC poll. Since they find an unusually high gap — an outlier from all other polling — what kind of sample produced it? And does this really show a shift in likely voters to Barack Obama, when state-by-state polling shows the race tightening in the other direction? The answer is that a poor sample still could show some movement:
More voters trust Obama to deal with the economy, and he currently has a big edge as the candidate who is more in tune with the economic problems Americans now face. He also has a double-digit advantage on handling the current problems on Wall Street, and as a result, there has been a rise in his overall support. The poll found that, among likely voters, Obama now leads McCain by 52 percent to 43 percent. Two weeks ago, in the days immediately following the Republican National Convention, the race was essentially even, with McCain at 49 percent and Obama at 47 percent.
Last week’s near-meltdown in the financial markets and the subsequent debate in Washington over a proposed government bailout of troubled financial institutions have made the economy even more important in the minds of voters. Fully 50 percent called the economy and jobs the single most important issue that will determine their vote, up from 37 percent two weeks ago. In contrast, just 9 percent cited the Iraq war as their most important issue, its lowest of the campaign.
As some have already noted, the sampling in this poll could explain why. In the raw numbers, the poll sample consists of 38% Democrats and 28% Republicans. This doubles the actual spread on party affiliation, last surveyed by the more-reliable Rasmussen at 38.7%-33.6% D-to-R. When the WaPo/ABC poll takes leaners into account, the spread gets even more pronounced: 54-38. That suggests a rather strong bias towards Obama, and an almost insurmountable hurdle for McCain.
In contrast, the spread was smaller on 9/7, when McCain had a two-point edge. The base number gave an eight-point advantage to the Dems, but only a nine-point advantage with leaners. That’s a seven-point shift in two weeks within the sample, which would certainly account for a large shift towards Obama.
But does it account for all of the shift? That’s a tougher question. The seven-point shift in leaners should only result in a seven-point shift, maximum, in the end result — but McCain lost eleven points. The difference could be attributable to the margins of error that are built into the polling, or it could just be that Obama did really gain a few points, although not eleven as the WaPo/ABC poll suggests.
We’ll have to watch other polling (and their samples) to determine whether this reflects reality or a pollster bias. Rasmussen’s daily tracking poll has Obama up two points for the first time in a while, so there may be movement — but it still looks like anyone’s race.