WaPo/ABC ends sample transparency in national polling

The Washington Post and ABC News has a new national poll out today.  It purports to show that Barack Obama has a 50% approval rating and that he would beat Mitt Romney in a head-to-head matchup.  And heck, that might even be true, except for a couple of problems.  First, this is a poll of general population adults rather than registered or likely voters, so it’s not even a proper polling type for the predictive outcome they claim.

More importantly, though, the poll series has dropped its reporting of partisan identification within their samples.  It’s the second time that the poll has not included the D/R/I split in its sample report, and now it looks as though this will be policy from this point forward.  Since this is a poll series that has handed double-digit partisan advantages to Democrats in the past (for instance, this poll from April 2011 where the sample only had 22% Republicans), it’s not enough to just hear “trust us” on sample integrity from the Washington Post or ABC.

One cannot determine whether Obama’s improvement in this series is a result of the State of the Union speech, as Dan Balz and Jon Cohen suggest, or whether it’s due to shifting the sample to favor Democrats more so than in previous samples.  The same is true for the Post’s report that Obama “for the first time has a clear edge” over Romney head-to-head.  One would need a poll of registered or likely voters to actually make that claim (one has to register to cast a vote, after all), and one would need to see the difference in partisan splits between this and other surveys in the series to determine whether the movement actually exists or got manufactured by the pollster.

Essentially, the overall poll is worthless, and given the track record of this poll series, it’s easy to assume that the reason that the Post has ended its sample transparency is because they have something to hide.

Among the Republicans and independents the Post did survey, Romney has a big lead over the rest of the field, 39/23 over Gingrich, with Santorum coming in at 16% and Paul at 15%.  Since we don’t know the mix between Republicans and independents here, either, and we’re still dealing with adults in the middle of the primaries when the likely voter model should be used, this is also a worthless result.

Update: The description in their news reports show 879 registered voters as a subsample.  However, their sample report only mentions the 1,000 adults of the overall sample, from which Obama’s 50% approval rating is taken, as well as their approval rating for the GOP rhetoric, which is prominently mentioned in the news report.  They do, however, report on both general population and registered voters for the Obama-Romney head-to-head — but again without any indication of the sample composition at all.

And once again, why anyone is polling adults in the middle of an ongoing primary is a complete mystery.  That should be likely voters.