Bookmark and ShareShare
Wednesday, April 18, 2012

Gallup’s National Election Tracking

Gallup Daily tracking of the general presidential election began April 11, and our first five-day rolling average result, reported Monday, showed Romney leading by two points, 47% to 45%, among more than 2,200 registered voters. However, the trend has been moving in Romney’s direction.

Within the first five nights of tracking, from April 11-15, Obama led by a small margin on the first two nights. Romney did better over the weekend, ultimately shifting the five-day average around to where he was slightly ahead. Romney’s advantage expanded further on Monday night, resulting in a 48% vs. 43% advantage for him in Gallup Daily tracking from April 12-16. We will see this morning whether last night’s interviewing shows Romney’s momentum continuing to build, leveling off, or reversing. Any of these is possible given the fast moving nature of today’s political environment. Regardless, the daily trends within our reported five-day averages show a definite move toward Romney since the weekend. The interesting point is why that occurred and what it indicates, if anything, for the campaign ahead. 

Gallup invests in state-of-the-art methodological techniques in its sampling and interviewing. One decision we made this year was to lengthen our reporting period to five-day rolling averages. Each day we drop off the oldest interviewing and add in the newest day of interviewing. Plus, we ask the registered voter question of a full random half sample a night, N=500, yielding about 440 registered voters each night, for a combined random sample of around 2,200 each reporting period. This gives us a very large sample size and stable estimates. The tracking is thus a little less responsive to short-term changes in the political environment, but makes the estimates more consistent. Our intention is to focus on broader changes in voter sentiments.

As you would expect from Gallup, we are continually refining our sampling and interviewing process to stay at the cutting edge of survey methodology. We, for example, were the first major poll in the U.S. to institute the inclusion of cell phones in January 2008, and now at least 40% of each of our samples are conducted on cell phones. We interview in Spanish language to a subset of respondents who want to be interviewed in that language. We make multiple calls on numbers in the sample. We now have our presidential ballot question at the front of the survey, in order to replicate the experience voters have in the voting booth, and in order to avoid any context effects. Other polls still ask the presidential ballot after other sets of questions, which may influence or bias it. All of the Gallup tracking interviews are conducted by Gallup’s own trained interviewers in our own interviewing facilities around the country.

These techniques have fared well in the past. In 2008, Gallup’s final estimates were within two percentage points of the actual vote share obtained by Barack Obama and John McCain. Similarly, in 2004, our final estimate was within one to two points of the final vote share for John Kerry and George W. Bush. In both instances, our final allocated estimates were slightly higher for the Democratic candidates involved than for the Republican candidates.

We are also as transparent about what we do -- under the assumption that we all benefit when interested parties are able to examine and analyze our data in detail and in different ways. We are happy to share the internals of our data with interested journalists, analysts, and others who want to find out more about what the data are showing. Since our first tracking report on Monday, we have sent internals and tabs to a number of journalists -- and also to the Obama re-election campaign’s pollster at his request (more on this below). Starting in a couple of weeks, we will be posting on our website each week detailed breakdowns of the vote within a wide range of subgroups, based on rolling averages of combined samples of over 6,000 registered voters.

This is my sixth presidential election cycle here at Gallup, and if there is one thing I have learned over the years, it’s that politics has evolved into a world in which the attempt to control the message is everything -- at all costs.

Now that we have begun tracking the presidential race on a nightly basis, the political reactions come out -- as I would expect based on history. In the current situation, we see Tuesday’s “tweet” by Obama campaign strategist David Axelrod to the effect that Gallup’s methodology is somehow flawed because it has shown a close race with Romney slightly ahead.

Criticizing a poll's “methodology” is a useful technique for a political campaign because it diverts attention from the message. Our Obama job approval in the first five nights (April 11-15) of tracking interviewing was 47%, consistent with other national surveys around the same time. If there were methodological problems with Gallup’s polling, they would logically be reflected in presidential job approval as well as the ballot.

We did not see tweets about Gallup when The New York Times' poll in early March showed Obama job approval dropping suddenly to 41%, while our tracking survey (the same one we are using now) showed Obama’s approval staying high and constant in the upper 40% range during the same time period. Nor did we see a tweet when we reported that the majority of Americans support the so-called Buffett Rule, which Obama has heavily pushed for.  On the other hand, we presumably should expect to see a tweet this morning criticizing the new New York Times/CBS News poll showing Obama and Romney tied at 46% each, essentially what we reported on Monday. And so on. The tweets come out when the results of a poll are counter to what is in the best interest of a campaign narrative. That’s a natural outgrowth of the intense political environment in which we operate today.

We are great believers in “collective wisdom.” That’s that we do. So if we can learn from others’ input, we welcome it. We thus welcome any campaign’s pollsters to come over to Gallup and to be briefed on our Gallup methods and to ask any questions they might have. I would hope that all pollsters, including those who work for campaigns, end up seeing that our focus is on providing the highest quality, top-drawer methodological polling. We are always happy to explain our methods to anyone.

As noted, a number of people have requested more information about our poll, and we supply that to everyone who requests it. It’s certainly possible that, with many minds pouring over our data, even our own analysts may learn something.

Ronald Brownstein of the National Journal and Mark Blumenthal of Huffington Post, for example, both wrote insightful pieces looking at the Gallup data and other polls that have come out recently.
Some discussion centers on the proportion of certain demographic groups in the sample -- which are always going to vary at least marginally across polls. Mark Blumenthal points out that all of the recent polls show the same high percentage of the vote among those who identify themselves as Republicans and Democrats going for their candidate -- basically about 90% on both sides. (This is a key because it shows that despite the rigors of the Republican nomination campaign, it appears that Republicans are having no problem indicating that they will end up voting for Romney -- and that most Democrats will have no problem indicating that they will end up voting for Obama). Mark ends by pointing out "While tracking polls will continue to produce ups and downs, the underlying structure of the race appears to be in place. The Obama-Romney contest is close and likely to remain that way for the foreseeable future."

Ron Brownstein looked at the data in a different way, focusing on race, gender, and class lines. One of his focus points is the percentage of the sample that is “non-white.” That’s significant, as Ron notes, given that in today’s political environment, race is a huge cleaving variable. The significant majority of whites vote for the Republican candidate and the significant majority of non-whites vote for the Democratic candidate. Ron has some speculation about what the final voting percentage of non-whites might be this November. But, of course, no one knows that at this point. We certainly don’t weight our sample to projected turnout proportions, in the sense of making assumptions about what the electorate will look like on Nov. 6. What the electorate will look like, of course, depends on what transpires between now and then. And what it looks like even then is dependent on what another poll -- the exit poll -- estimates. There are no official election records of the race, age, gender, etc., of voters as they vote.

The 2008 election may well have been a high-water mark for the percentage of non-white voters in the electorate, given the presence on the ballot of the first major party African-American candidate in history. It’s not clear that the percentage of non-whites this Fall will grow larger than 2008. That’s still an unknown.

Gallup has refined its race/ethnicity question in great detail in recent years, and we basically follow the U.S. Census Bureau procedures. Additionally, and this is a key, we weight each of our samples to the census parameters on race. That, in turn, means that the registered voters we use are the result of an overall weighted national adult sample. The race composition in the national adult pool reflects the base census figures.

I have no doubt that there will be other issues that arise as the election season plows ahead. Gallup’s purpose in investing in tracking and making it widely available on Gallup.com is to provide interested parties a solid, scientific basis for understanding the dynamics of this important election. In general, we welcome any and all input and -- as I’ve noted -- believe that we all benefit from a shared focus on our data.

0 comments:

Post a Comment

Comments are moderated by Gallup and may not appear on this blog until they have been reviewed and deemed appropriate for posting.

Copyright © 2010 Gallup, Inc. All rights reserved. | Terms of Use | Privacy Statement