Adjustment of Washington Post/ABC Poll for Party Affiliation

The Washington Post/ABC News presidential poll published on 1 October 2012 says that Obama leads Romney among likely voters by 49% to 47% and among registered voters by 49% to 44%.  The poll was based on 1002 respondents between 26 and 29 September.  But the respondents broke down as follows:

Democrats 33%
Republicans 28%
Independents 34%
Other 4%
Don't Know 1%

Because only about 9% of the people asked to respond to a poll will do so and because there were 58 questions in this poll, it is very likely that those willing to respond are not necessarily a good representative sample of those who will vote.  For instance, what fraction of small business owners, struggling very hard these last four years to keep their businesses going in the Obama economy, will take the time to answer these questions?  Yet, this group of people is heavily Republican.  In general, one also can imagine that the sample of people willing to take this time on the phone may be more heavily weighted toward the unemployed than the employed.

So, we need to find a source of how many people are Democrats, Republicans, Independents, and Others.  Rasmussen Reports has such data here.  They do not have their September numbers up yet, so I will use the August numbers as I did in my prior adjustment of Washington Post/ABC News presidential poll results from 11 September.  The Rasmussen party affiliation numbers are:

Democrats 33.3%
Republicans 37.6%
Other 29.2%, of whom I expect 25.2% are Independents and 4% are Libertarians, Greens, Constitutional Party, Social Workers Party and other micro-parties

We can use the registered voter results of the WaPo/ABC poll to see what percentage of Democrats, Republicans, and Independents will vote for either Obama or Romney and apply the Rasmussen party affiliation weightings to adjust for the proper sampling of voters by party.  According to this poll,

Democrats: 89% for Obama, 7% for Romney
Republicans: 7% for Obama, 87% for Romney
Independents: 47% for Obama, 47% for Romney

With the Rasmussen party weightings among registered voter respondents:

Votes for Obama = (0.89)(33.3%) + (0.07)(37.6%) + (0.47)(25.2%) = 44.1%

Votes for Romney = (0.07)(33.3%) + (0.87)(37.6%) + (0.47)(25.2%) = 46.9%

The Washington Post 49% to 44% advantage to Obama becomes a 47% to 44% advantage for Romney among registered voters.  If we use the WaPo/ABC poll adjustments for likely voters we get:

Votes for Obama = (0.91)(33.3%) + (0.06)(37.6%) + (0.45)(25.2%) = 43.9%

Votes for Romney = (0.07)(33.3%) + (0.90)(37.6%) + (0.49)(25.2%) = 48.5%

The recalculated likely voter advantage for Romney is 4.6% compared to his recalculated registered voter advantage of 2.8%.  The WaPo/ABC advantage of likely voters was given as a 2% Obama advantage by them.

Of course this change in the advantage hinges upon the Republicans having a registration advantage over Democrats. There have been a few people who have checked on this in some of the states that will report how many people are registered as Democrats, Republicans, and Other or Independents.  The general story seems to be that Democrat registrations are way down and registrations as Independents are way up compared to 2008.  Republican registrations are more constant, but generally up a bit.  Here is one such report by John Hinderaker at Powerline on registrations in Ohio.

We should also be aware that there are many more Republican governors, Senators, Representatives, and state legislators now than there were in 2008, so it is very probable that there has been a trend away from Democrat to Republican party affiliation since then.  Even in this election, Real Clear Politics has the Republicans in the House with 226 seats leaning to them, the Democrats with 183 seats leaning to them, and only 26 toss-up seats of which more than half are now Republican.  If the Republicans win in half of the toss-up seats, they will still have a substantial House majority.  Such a majority would be unlikely if the Democrats had the 5% registered voter edge implied by this WaPo/ABC poll with their sample of registered voters.  That cannot be the case.

It should also be noted that this is an unusual poll result in that it implies that among Independent registered voters the split for Obama/Romney is 50-50, or more precisely 47-47.  Almost every other poll has shown Romney to be winning among Independents.  It will be interesting to see if future WaPo/ABC polls continue to show a trend toward equality of Obama and Romney among Independents.  Personally, given the continued dire state of the Obama-mismanaged and abused economy and Obama's failure to provide adequate security in our embassies and consulates as the Mideast sinks into chaos, I expect Romney to improve his advantage among Independents.

The Democrats are claiming that Romney was greatly hurt by his fundraiser comments that he could not expect many of the votes of the 47% not paying income tax and of those highly dependent upon government welfare programs.  No doubt he was hurt, but one of the questions in this WaPo/ABC survey was whether people thought too many people were dependent upon the government and they answered heavily in favor that yes, too many were dependent.  This poll also noted that more people believed the economy would recover with Romney as President, despite the skewed sample.  That has to be a most important Romney advantage.