Most Primary Polls Inflating Clinton's Advantage
by Chris Bowers, Thu Apr 12, 2007 at 07:58:33 AM EDT
Looking through the internals of the new Los Angeles Times / Bloomberg poll, I was struck by this little nugget (emphasis mine):
Absurdly high projections of Democratic primary turnout can be found in other recent national polls as well. Gallup's latest poll sampled 1,008 adults nationwide, including 491 people in their Democratic trial heat. That projects a Democratic primary turnout model of over 50% of registered voters. Time's latest poll sampled 1,264 adults nationwide, and included 511 people in their Democratic preference trial heat, for a primary turnout model of over 40% of registered voters. Pew sampled 1,503 adults in their latest survey, and included 614 people in their Democratic preference sample, thus projecting a national primary turnout model of over 40%. CNN sampled 1,027 adults nationwide, and included 447 Democrats in their presidential preference trial heat, thus projecting a national primary turnout of over 45%. None of this makes any sense. None of these news outlets and polling firms are making any attempt to actually try and poll anything resembling an accurate makeup of the Democratic primary and caucus electorate nationwide. Not only are their turnout models way, way too high (they would even be too high for general elections in presidential years), but I bet not a single one of these polls makes any attempts to adjust for the primary / caucus rules specific to the state where people are being polled. For the most part, these national polls simply cannot be considered serious attempts to determine the current state of opinion among the Democratic primary electorate.
The reason this matters is because there are important, significant differences between public opinion among people who simply self-identify as Democrats, and people who will actually participate in Democratic primaries and caucuses. Consider the following in Iowa:
The fact is that most self-identified Democrats, or even most registered Democrats, will not participate in the 2008 presidential primaries. Since there is a noticeable drop within Clinton's performance among those Democrats who are most likely to participate in either primaries or caucuses, it seems that the three-way is even more competitive than we had previously surmised. Polls that project Democratic primary turnout in the 40-50% range among all registered voters, which right now is what most national primary polls are doing, inflate Clinton's perceived national advantage. These polls are conducted because they are cheaper than polls which include a decent sample of likely Democratic caucus goers, so basically what we are seeing here is a case of garbage in, garbage out. You get what you pay for in polling, and if you are not willing to pony up for a more expensive poll that is able to accurately sample the Democratic primary electorate, don't expect that your poll will end up being an accurate reflection of the Democratic primary electorate.
It is also worth pointing out that Giuliani performs far worse, and Romney performs far better, in polls that sample likely Republican primary goers, rather than all self-identified and / or registered Republicans.
The Los Angeles Times/Bloomberg poll contacted 1,373 adults nationwide by telephone April 5- 9. Included are 1,246 registered voters, among them 557 Democratic primary voters and 437 Republican voters.Ummm... their Democratic primary voter pool includes 45% of registered voters nationwide (557 of 1,246)? Pardon my French, but that is just fucking absurdly high. The Democratic primary electorate will not even come close to making up 45% of registered voters in any state, much less nationwide. In New Hampshire in 2004, the state which boasts the highest presidential primary turnout of any state in the nation, where members of any party can choose to vote in any primary they like, and in a year where there was no seriously contested Republican primary, the Democratic primary electorate represented only about 25-26% of the registered voting population (220,000 of 853,000). In 2008, many states holding primaries and caucuses on February 5th or earlier will have closed primaries open only to registered partisans, every state will feature a seriously contested Republican primary, and every state will have a lower voter turnout than New Hampshire. We are talking about, at best, 15-20% turnout of the registered voting population in most states for Democratic primaries and caucuses, no where near 45%.
Absurdly high projections of Democratic primary turnout can be found in other recent national polls as well. Gallup's latest poll sampled 1,008 adults nationwide, including 491 people in their Democratic trial heat. That projects a Democratic primary turnout model of over 50% of registered voters. Time's latest poll sampled 1,264 adults nationwide, and included 511 people in their Democratic preference trial heat, for a primary turnout model of over 40% of registered voters. Pew sampled 1,503 adults in their latest survey, and included 614 people in their Democratic preference sample, thus projecting a national primary turnout model of over 40%. CNN sampled 1,027 adults nationwide, and included 447 Democrats in their presidential preference trial heat, thus projecting a national primary turnout of over 45%. None of this makes any sense. None of these news outlets and polling firms are making any attempt to actually try and poll anything resembling an accurate makeup of the Democratic primary and caucus electorate nationwide. Not only are their turnout models way, way too high (they would even be too high for general elections in presidential years), but I bet not a single one of these polls makes any attempts to adjust for the primary / caucus rules specific to the state where people are being polled. For the most part, these national polls simply cannot be considered serious attempts to determine the current state of opinion among the Democratic primary electorate.
The reason this matters is because there are important, significant differences between public opinion among people who simply self-identify as Democrats, and people who will actually participate in Democratic primaries and caucuses. Consider the following in Iowa:
Among a subsample of likely Democratic caucus-goers polled before the Edwards announcement, Edwards led with 30.2 percent, followed by Clinton at 24.4 percent and U.S. Sen. Barack Obama (D-Ill.) at 22.1 percent. After the announcement, both Edwards and Clinton moved up, while Obama dropped. Among likely caucus-goers after the Edwards announcement, the results were: Edwards 36.4 percent, Clinton 33.9 percent and Obama 14.4 percent.Edwards does better among likely caucus-goers in Iowa, while Clinton does better among the wider net of "registered Democrats." Further, in February, Pew found striking differences in potential support between Clinton and Obama among people who were paying close attention to the campaign, and people who were not. Specifically, Clinton led Obama 56-48 among potential supporters who were following the campaign closely, but held a much larger, 42-23 lead, among people who were not paying close attention. In other words, Obama's potential support increased by more than 100% among people following the campaign, while Clinton's increased by only 33%. And then there is this little nugget from today's LA Times / Bloomberg poll:
Among all registered Democrats (not just likely caucus-goers) asked who they would support for president in 2008, before the Edwards announcement the front-runners were: Clinton - 21.1 percent; Edwards - 19.7 percent; Obama - 18.8 percent; don't know - 32.9 percent. After the announcement, some change was seen for both Edwards and Clinton, with a substantial move away from "don't know" and Obama: Clinton - 29.5 percent; Edwards 23.2 percent; Obama - 11.4 percent; don't know - 28.7 percent.
Though Obama ran even with Clinton among college graduates, Clinton did better among less-educated voters.Again, Clinton is weaker among high-information Democrats who are more likely to vote in Democratic primaries than she is among all self-identified Democrats. This goes a long way toward explaining the huge difference found in some primary polls, such as Zogby and Rasmussen, which poll "likely" Democratic primary voters, and all other polls which are throwing an absurdly wide net for the 2008 Democratic primaries. According to Zogby and Rasmussen, Clinton's lead is typically around 8-10% nationwide (5-12% is the wider range), whereas her lead in most other polls approaches twice that amount, 15-20% (10-24% is the wider range). Considering the differences in the types of samples these polling firms are using, I have no choice but to conclude that the Zogby and Rasmussen polls are by far better indicators of the current state of opinion among the Democratic primary electorate. Clinton's lead is almost certainly about half the size that most polls project it to be.
The fact is that most self-identified Democrats, or even most registered Democrats, will not participate in the 2008 presidential primaries. Since there is a noticeable drop within Clinton's performance among those Democrats who are most likely to participate in either primaries or caucuses, it seems that the three-way is even more competitive than we had previously surmised. Polls that project Democratic primary turnout in the 40-50% range among all registered voters, which right now is what most national primary polls are doing, inflate Clinton's perceived national advantage. These polls are conducted because they are cheaper than polls which include a decent sample of likely Democratic caucus goers, so basically what we are seeing here is a case of garbage in, garbage out. You get what you pay for in polling, and if you are not willing to pony up for a more expensive poll that is able to accurately sample the Democratic primary electorate, don't expect that your poll will end up being an accurate reflection of the Democratic primary electorate.
It is also worth pointing out that Giuliani performs far worse, and Romney performs far better, in polls that sample likely Republican primary goers, rather than all self-identified and / or registered Republicans.
Tags: Democrats, inflated Clinton poll theory, polls, President 2008, Primary Elections (all tags)
-----------
myDD - skin
-----------
23 Comments