Poll Position
By Shawn Healy
If you are among the sliver of America already tuned into the 2008 election, you are without doubt deluded on a daily basis with public opinion polls, both of the national and state variety. It's sometimes difficult to get a handle on what all of this information means, and a number of caveats abound as we interpret their findings. What follows is a few tips from a self-confessed political junkie and numbers geek.
Polls are a mere snapshot in time. They take a picture of public opinion over a few days, but fail to account for the fact that public sentiments are continuously shifting, at least as they relate to candidates for public office (it is remarkably stable relative to issue positions). Rather than placing too much weight in a given poll, look for trends. Who's moving up, and who's falling through the floor. For instance, Mike Huckabee is gaining at the right time with the Iowa caucuses a mere four weeks away, while Fred Thompson's decline may represent an early death knell.
For this reason, pay attention to longitudinal data to circumvent the cross-sectional nature of polls. I visit pollster.com on a regular basis because they show candidate poll performance across time, and also aggregate a series of polls at once. They use data points from each to arrive at the average position for each candidate.
This brings me to my next point: Don't place all of your eggs in one basket. A single poll is not necessarily an accurate depiction of the race. All polls account for errors, as they survey only a small sample of the electorate and make projections off of this data. In general, larger samples, so long as they are randomly constructed, equal lower margins of error.
Margins of error are statistical terms to reflect the possibility that the numbers reported in the sample vary from the general population. When a survey suggests a statistical dead heat, this typically means that the candidates are within the margin of error of one another. For instance, Barack Obama leads in Iowa according to the most recent Des Moines Register poll, but Hillary Clinton may be ahead of him when accounting for margin of error, and third place candidate, John Edwards, may even be ahead of Hillary. The projections offered in the poll, however, are the best-guess estimates of the current state of the race.
Early opinion polls, especially of a national nature, are indicative of name recognition and little more. This is why Hillary and Rudy Giuliani led the pack early on and maintain national leads. In the early caucus and primary states of Iowa, New Hampshire, and South Carolina, where voters see the candidates up close and personal on a daily basis, the numbers vary significantly. True, Hillary leads in New Hampshire and South Carolina, but Barack is closing in, and the Illinois Senator leads in Iowa (at least according to some polls). On the Republican side, Huckabee leads in Iowa (suggested by three recent polls), and Romney in New Hampshire and South Carolina.
The momentum gleaned from these early contests will likely translate into success elsewhere as a result of free media, momentum, and money flowing into campaign coffers, so for better or worse, discount national polls for their early state counterparts.
A crucial variable is who is being surveyed. John Edwards, for instance, finished second in the 2004 Iowa Caucuses and has a legion of supporters who have already showed up on caucus night for him, while Barack Obama is tremendously popular among young people in the state, many of whom tend not to vote when push comes to shove (colleges will also be on winter break on January 3rd). All the poll administrators can do is ask voters if they intend to caucus. Delivering them to the polls is the responsibility of campaigns. Turnout will decide who wins on election night, not intentions, meaning Edwards' support may indeed be more realistic.
On the other hand, Obama's support among young people may be underestimated in polls given their tendency to shun land lines for cellular phones. I for one have not had a landline for a decade, and 16% of households share this habit. Polling administrators tend to focus on home phones, and to date, I have received only one phone call conducting a poll for a local aldermanic race. In short, my opinions are typically not accounted for in opinion polls, and along with my counterparts, this may represent bias and add to the errors inherent in opinion polls.
What then, should we make of this barrage of data? Think of polls as a public thermometer taking the temperature of voters at a moment in time. As a native of Wisconsin, I was taught that if I didn't like the weather, I should wait five minutes for it to change. Although public sentiments aren't as fickle as a Midwestern spring, they are fluid enough for us to caution calling an election before it has even begun.
I would be remiss if I failed to mention our current special exhibit at the Freedom Museum, Vote4Me!: Inside a Presidential Election. One portion of the exhibit addresses the issue of polling data and presents many of the caveats I echoed above. It asks the visitor to play the role of presidential candidate and craft issue positions relative to three hot button issues: Iraq, taxes, and gay marriage. Then, these positions are placed alongside those of our previous visitors and the candidate is asked to emphasize those issues that resonate with the public, and downplay those without widespread backing. I encourage you to come and take your own temperature!
Polls are a mere snapshot in time. They take a picture of public opinion over a few days, but fail to account for the fact that public sentiments are continuously shifting, at least as they relate to candidates for public office (it is remarkably stable relative to issue positions). Rather than placing too much weight in a given poll, look for trends. Who's moving up, and who's falling through the floor. For instance, Mike Huckabee is gaining at the right time with the Iowa caucuses a mere four weeks away, while Fred Thompson's decline may represent an early death knell.
For this reason, pay attention to longitudinal data to circumvent the cross-sectional nature of polls. I visit pollster.com on a regular basis because they show candidate poll performance across time, and also aggregate a series of polls at once. They use data points from each to arrive at the average position for each candidate.
This brings me to my next point: Don't place all of your eggs in one basket. A single poll is not necessarily an accurate depiction of the race. All polls account for errors, as they survey only a small sample of the electorate and make projections off of this data. In general, larger samples, so long as they are randomly constructed, equal lower margins of error.
Margins of error are statistical terms to reflect the possibility that the numbers reported in the sample vary from the general population. When a survey suggests a statistical dead heat, this typically means that the candidates are within the margin of error of one another. For instance, Barack Obama leads in Iowa according to the most recent Des Moines Register poll, but Hillary Clinton may be ahead of him when accounting for margin of error, and third place candidate, John Edwards, may even be ahead of Hillary. The projections offered in the poll, however, are the best-guess estimates of the current state of the race.
Early opinion polls, especially of a national nature, are indicative of name recognition and little more. This is why Hillary and Rudy Giuliani led the pack early on and maintain national leads. In the early caucus and primary states of Iowa, New Hampshire, and South Carolina, where voters see the candidates up close and personal on a daily basis, the numbers vary significantly. True, Hillary leads in New Hampshire and South Carolina, but Barack is closing in, and the Illinois Senator leads in Iowa (at least according to some polls). On the Republican side, Huckabee leads in Iowa (suggested by three recent polls), and Romney in New Hampshire and South Carolina.
The momentum gleaned from these early contests will likely translate into success elsewhere as a result of free media, momentum, and money flowing into campaign coffers, so for better or worse, discount national polls for their early state counterparts.
A crucial variable is who is being surveyed. John Edwards, for instance, finished second in the 2004 Iowa Caucuses and has a legion of supporters who have already showed up on caucus night for him, while Barack Obama is tremendously popular among young people in the state, many of whom tend not to vote when push comes to shove (colleges will also be on winter break on January 3rd). All the poll administrators can do is ask voters if they intend to caucus. Delivering them to the polls is the responsibility of campaigns. Turnout will decide who wins on election night, not intentions, meaning Edwards' support may indeed be more realistic.
On the other hand, Obama's support among young people may be underestimated in polls given their tendency to shun land lines for cellular phones. I for one have not had a landline for a decade, and 16% of households share this habit. Polling administrators tend to focus on home phones, and to date, I have received only one phone call conducting a poll for a local aldermanic race. In short, my opinions are typically not accounted for in opinion polls, and along with my counterparts, this may represent bias and add to the errors inherent in opinion polls.
What then, should we make of this barrage of data? Think of polls as a public thermometer taking the temperature of voters at a moment in time. As a native of Wisconsin, I was taught that if I didn't like the weather, I should wait five minutes for it to change. Although public sentiments aren't as fickle as a Midwestern spring, they are fluid enough for us to caution calling an election before it has even begun.
I would be remiss if I failed to mention our current special exhibit at the Freedom Museum, Vote4Me!: Inside a Presidential Election. One portion of the exhibit addresses the issue of polling data and presents many of the caveats I echoed above. It asks the visitor to play the role of presidential candidate and craft issue positions relative to three hot button issues: Iraq, taxes, and gay marriage. Then, these positions are placed alongside those of our previous visitors and the candidate is asked to emphasize those issues that resonate with the public, and downplay those without widespread backing. I encourage you to come and take your own temperature!
0 Comments:
Post a Comment
<< Home