Free Markets, Free People
A Primer on polling
As I’ve mentioned any number of times, you have to be careful about what polls you consider as worthy of believing and what polls are likely not particularly accurate. Jay Cost has a great article about that with the added point that the media doesn’t understand what he tells you and so doesn’t understand the races in the various states.
He points to this from the Hill as an example:
President Obama is retaining his commanding lead over Mitt Romney in Pennsylvania, topping the Republican presidential nominee by 12 points in a poll released Wednesday by Franklin & Marshall College. Obama would win the favor of 48 percent of Keystone State voters, versus just 36 percent for Romney, according to the poll.
1. The president is under 50 percent in most swing state polling averages. It’s not an ironclad rule that Obama cannot rise in the polls, but common sense suggests that it will be tough. He’s been the president for three years – if you’re not inclined to vote for him now, what will five months of a campaign do?
That’s an important point – if you’re the incumbent and polling under 50%, you’re in trouble regardless of the type voter the poll uses. Also note that the Romney candidacy isn’t even official yet. He would likely see a rise in preference once he is officially the GOP candidate.
Another point I’ll expand on later – favorability. Cost says this:
It’s worth noting as well that most of these polls show the president getting roughly his job approval, which is all we should expect him to receive in the general election (maybe a little less). And his job approval rating has consistently been under 50 percent for two-and-a-half years.
Not good. Not insurmountable, but certainly not an indicator of a strong candidate. More on favorability later.
2. Most polls are of registered voters. This matters because the actual electorate will only be a subset of registered voters, and will probably be more inclined to vote for the GOP. So, these polls probably overstate Obama’s “lead,” such as it is.
With a state like Pennsylvania, using registered voters, you most likely get an oversampling of Democrats. Which side is most likely to be motivated this time? The GOP. So the number quoted in the Hill story is probably considerably lower than claimed (remember Wisconsin? A state carried by 14 points in 2008 is now showing 7 points).
3. There is no “blue wall.” This is a common point pundits will make – the list of states that have not voted Republican since 1988 amounts to a “blue wall” for the president. Nonsense. It’s better to say that these states have Democratic tilts, some of them pretty minimal.
We’ve seen evidence of that in landslide elections. There’s no “red wall” either. It’s all about tilts. Some states tilt more than others but all states, at some point, are in play. Think preference cascade.
The states with a Republican tilt of at least 1 point total up to 253 electoral votes, based on the 2008 results. The states with a Democratic tilt of at least 1 percent total up to 257 electoral votes.
In other words, it’s a wash.
And this is key:
4. The “horse race” metaphor has its limits. Take this from the guy who used to write the Horse Race Blog: The concept of a horse race does not capture the idea of voter psychology very well at this point. Roughly 85 percent or so of the electorate is locked in – though they may not be admitting it to pollsters – while the final 15 percent has barely started the decision-making process. So, the idea that Obama has a “lead” in the polls is really a non sequitur. The gettable voters are not yet engaged, so there really is no race going on at the moment
The fight is for the 15% and they’re not even really paying attention yet. My guess is the 15% probably have a preference, but can be swayed. But for the most part, the majority of the electorate is already engaged (and again, this is one reason WI has national implications).
Finally, favorability. Obama supporters like to point to his favorability rating vs. Romney. That’s pretty much useless as Morris Fiorna explains:
Over all, in the 13 elections between 1952 and 2000, Republican candidates won four of the six in which they had higher personal ratings than the Democrats, while Democratic candidates lost four of the seven elections in which they had higher ratings than the Republicans. Not much evidence of a big likability effect here. In most elections, however, the electorate did not give a large personal edge to either candidate. In four elections they did.
So it is a very mixed bag concerning favorability or likeability. The most pertinent recent example:
Jimmy Carter’s 1980 job approval was flirting with lows established by Harry S. Truman, Nixon and later, George W. Bush, but the electorate rated Carter’s personal qualities as the highest of the Democratic candidates between 1952 and 2000. The same electorate rated Ronald Reagan as the lowest of the Republican candidates. The Ronald Reagan of October 1980 was not the Reagan of “morning again in America” in 1984, let alone the beloved focus of national mourning in 2004. Many Americans saw the 1980 Reagan as uninformed, reckless, and given to gaffes and wild claims. But despite their misgivings about Reagan, and their view that Carter was a peach of a guy personally, voters opted against four more years of Carter.
Fiorna sums it up this way:
“Voters didn’t like my personality” is a loser’s excuse.
As the campaigns progress, we’re likely to hear how Obama’s favorability rating is higher than Romney’s and that such a rating is “significant”. Don’t buy into that. It is likely not that significant at all.
In summary, if the candidate is under 50% in a state in which registered voters are polled, he’s not as strong (or weak) as the polling might indicate. If the poll is of registered voters, take it with a grain of salt. All states are in play and the fight is for the uncommitted 15%.
Favorability? Disregard. It’s about job performance. (That said, here’s POLITICO trying to make something of Obama’s favorability rating).
Hopefully this will help you navigate the worth of the umpteen polls you’ll have thrown your way in the next few months. You should be able to quickly get their measure and then just as quickly figure out if the media has any idea of what it is talking about.
Most likely you’ll find they don’t. But then, that shouldn’t particularly surprise you, should it?