Explaining the "dip" in Obama approval rating: They polled more Republicans!

If you watch any of the morning news shows on TV, or read the news first thing with your morning coffee, you know today's poll narrative in the media: In the midst of rising gas prices, Obama's approval ratings dip. That is the headline from the ABC News/Washington Post poll released today, showing President Obama's approval rating at 46% and disapproval at 50%, a reversal of fortunes from early February. Hell, it even shows Romney beating Obama by a point. The explanation offered by pundits and headlines? Gas prices, which 66% of Americans are very concerned about, the poll finds.

But what's behind the headlines? What's beyond the pundits and the pontifications? Why did President Obama's approval numbers really go down? Or did it even actually go down? What really accounts for the reversal from last month? The real answer: they polled more Republicans this time than last. Ta-da! I decided to compare the party ID numbers attached to this poll vs. the ones attached to the last poll. Look what I found:


From the last poll to this one, there is a net 7 point gain for GOP identified voters as opposed to Democratic ones, and there is a net 8 point loss in the President's approval rating. Hmm, looks like an awfully close correlation to me. If we assume that independents lean roughly the same way as the party ID numbers (really, very few voters are truly independent), GOP and GOP lean voters get a representation bump another 3 percentage points net, moving the GOP party ID vs. Democrats to a net +10 points as compared to the last poll. Given that about 80% of GOP and GOP-leaning voters oppose President Obama, the entire 8 point swing in the poll can be accounted for by the additional representation of Republican and Republican leaning voters.

But what are the real numbers on voter registration in this country? From the most current data from states that allow registration by party, Democrats outnumber Republicans by a 12 point margin, 43% to 31%, with independents coming in at 24%. Granted, only 29 states and DC allow registration by party, so take that data with that caveat, but I will note that the February poll had a much closer party ID difference (D +11) to the known actual national data (D +12). In this poll, that has dwindled down to a D +4, which is obviously a significant over-representation Republican and GOP-leaning voters, and an equally significant under-representation of Democratic voter registration advantage.

It is important to note here that the analysis above is not meant to show that ABC News and Washington Post pollsters somehow "cooked" the numbers. These variations - 3 points here and 4 points there, especially given the poll's margin of error of +/- 4 percentage points - are normal statistical anomalies. But it is intellectually dishonest to point to this poll and find a gas price related "dip" for the president's approval without looking at the very obvious factor of over-representation of GOP voters in the poll.

That is even more true when the analysts have no basis for comparison. When the last poll showed the president's approval rating at 50%, questions about gasoline were not involved, and thus there is simply no way of knowing, based on the numbers, whether gasoline prices actually are affecting the president's approval rating, if they are affected at all. This month, the poll shows 66% of Americans "very" concerned about gas prices, but we do not know if this number was higher, lower, or the same last month. The analysts are basically comparing approval numbers based on another number (concern about gas prices) that wasn't even present in the last poll. This is how logical fallacies are made.

Here is another piece of information from the poll the media narrative drivers don't want you to pay attention to: by a 54-40 margin, even this GOP-skewed group expect President Obama to win re-election. That is a net 17-point gain in Obama's favor since January. This means that a large portion of even Republican and GOP-leaning voters believe that none of their candidates can win in November. I doubt this would be the case if any factor other than the over-representation of GOP voters were moving the approval and head-to-head matchup numbers in this poll

As a final thought, the media buzz seems to be more a proof of the proverb that there are lies, damned lies and then there are statistics than it is about an actual dip in the president's numbers. Statistical analysis is invaluable when it is done properly, analyzing and comparing all the factors. But when the same "numbers" are used to push a political talking point instead, a wealth of information buried in a survey gets ignored in favor of pushing a given narrative. That is the dangerous part. We need our media talking heads to leave behind narratives and look at real numbers and present real analysis.


Note: You may notice that the February numbers are 2/12 and not 2/4, which is when the last poll was released. But given that these were the exact numbers attached when the 2/4 polls came out (and with a 2/12 label even then), I'm thinking that's a typo from the pollsters.