Bookmark and ShareShare
Tuesday, December 13, 2011

Tracking the GOP Race

People have questions about our Gallup tracking of the vote preferences of Republicans. One question I often hear: Since the GOP nomination will be one on a state by state basis through primaries and caucuses, why pay attention to the national numbers?

Here are some answers to that:

1. The only state numbers that mean much now are Iowa and New Hampshire, and the voters' views in those two states are changing and possibly will change more between now and their caucus and primary dates.

The positioning of the candidates in the rest of the states will change in cascading effect after actual votes begin coming in from these two early states. The national voting intentions of Republicans will also change in cascading effect once the state results begin to come in.

Monitoring this cascading change using the national numbers provides us a larger sample base and a broader audience than keeping up with each state individually. It’s a more efficient mechanism for monitoring the impact of campaign events and election results.

2. The election process has been more nationalized than usual this year, primarily because there have been so many more nationally-televised debates. Local, retail politicking in the individual states has become less valuable or necessary, at least up to this point.

It therefore makes sense to measure Republican voters' sentiments on the broad, national stage. We are measuring the candidates, as it were, in front of a big, national test audience -- since they, in essence, are playing to the big, national audience. The views of the national audience of Republicans provides an excellent indicator of exactly how the campaigns are playing out.

3. So far, the shifts we have seen in the national numbers are being duplicated at the individual state level. Most recently, as we have seen the surge in movement for Newt Gingrich nationally, we have also seen it in Iowa, South Carolina, and Florida.  So the national tracking essentially provides a measure of the same trends seen in the individual states.

4. It’s easier and much more cost efficient to conduct daily tracking on the national level than it is to provide daily tracking at individual state levels.

We also receive questions about the methodology we use for our tracking.

At this point, our national tracking of Republicans’ preferences for their party’s nominee is based on five-day rolling averages of over 1,000 interviews each day. Every day we drop off the oldest day’s interviews and add on the most recent day’s interviews. That means from one day to the next, 80% of the interviews being reported are the same. For example, Monday’s results were based on the combined interviewing for Wednesday, Thursday, Friday, Saturday, and Sunday among Republicans nationwide. On Tuesday, we reported the combined interviewing for Thursday, Friday, Saturday, Sunday, and Monday. So the interviewing conducted on Thursday, Friday, Saturday, and Sunday remains in both samples. The Wednesday interviewing was replaced by the Monday interviewing on Tuesday.

Our decision-making on this rolling average is based on the need to balance interest in keeping the data relatively stable, while at the same time allowing it to pick up actual changes in the minds of voters.

All of our experience, going back to presidential election tracking in 1992, shows that surveys of voters will be volatile at the day-to-day level. Some of this is sampling error. That means that even if the population of Republicans nationwide hasn’t changed their views at all, the voter preferences measured on Monday will be at least somewhat different than the voter preferences measured on Tuesday due to the fact that any sample from a population will not perfectly represent that population. So, if we simply reported the results of each night’s interviewing, even with a larger sample size, we would get fluctuations that would not, in reality, be indicative of underlying change.

At the same time, there are real, day-to-day changes in voters’ preferences for candidates. Some observers have found this hard to comprehend, in my experience, but it’s true. Voter preferences are not totally stable on a day-to-day basis. There are enough voters who are on the fence about their preferred candidate that even if we were able to interview everyone in the population daily, we would find change. That’s particularly true when we are interviewing people in one party only (about their primary nominee choice) as is the case now, since these people are not anchored by party identification and are, in many ways, freer to jump around. We have certainly seen that this year so far.

So, if we “go long” and combine two weeks' worth of interviewing and each night dropped off the oldest night and added in the most recent night, we would have a highly stable estimate each night, but one that would be very slow to reflect real world changes in the population. For example, we anticipate that the preferences of Republicans nationwide for their nominee will change after the results of the Iowa caucuses come in on Jan. 3. If the time frame for our tracking is too wide, it would take a long time for those changes to begin to appear in our daily reports.

If we “go short” and combine one to three days of interviewing, we would have a highly sensitive estimate each night, but one that would show too much volatility and would not be an optimal guide to what is actually changing.

We hit on the five-day rolling average as the best compromise and midpoint between these two extremes. Even with that, we are seeing some variation in the daily five-day number we report as the voter preferences for the GOP candidates. This reflects both sampling error and some actual real world volatility in Republicans’ views on who they support.

We will monitor this as we go. Right now, the basic structure of the race has remained roughly the same since we initiated tracking on Dec. 1. Gingrich is receiving about 33% support from Republican registered voters nationally, with fluctuation around that point each day, while Romney’s support has been remarkably stable around 23%. The other candidates are all fluctuating slightly down in the single digits.


Anonymous said...
December 16, 2011 at 5:10 PM  

Why no poll results so far today, Friday, 12/16?

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