Poll Position, Level One
By Shawn Healy
In our coverage of the 2008 presidential election, polling data has entered our discussion an untold number of times. As consumers of daily news, we are literally bombarded with data on the so-called horse race as media outlets trip over one another to release their latest temperature of the electorate. These numbers are significant, but must be interpreted with several caveats in mind. What follows is Part One in a series of simplistic breakdowns of public opinion polls in the context of the 2008 presidential election.
As polling data is collected and analyzed, keep in mind that public opinion is never stagnant, but instead in continuous flux. Polls are snapshots of how the electorate feels at a specific moment in time. In order to make meaning of this deluge of information, we are better served by longitudinal looks at public opinion where trends become evident and day-to-day fluctuations balance out. Gallup provides a great service through their daily tracking poll, allowing us to view the contest along a longer trajectory. Also, notice that their daily polling figures are a compilation of three days of data, providing a longer time horizon.
Any responsible polling organization releases its margin of error when presenting survey data. This figure is a product of sample size, and accounts for the fact that samples are but a small fraction of the general population. If samples are representative of the population as a whole, or of likely voters, the margin of error allows for wiggle room in either direction. The aforementioned Gallup poll, for example, finds Senator John McCain leading Senator Barack Obama by a mere 2 points, 47%-45%. With a margin of error of plus or minus 2%, this means that Obama could actually be leading the race, 47%-45%, or that McCain is up by a larger margin, 49-43%. The race, according to Gallup, is a statistical tie. However, the 2 point spread represents their most accurate projection of the national electorate over the three-day period studied.
As mentioned earlier, as consumers of public opinion polls, it is best to be wary of any single poll as a predictor of pending election results. In a sense, we try to avoid placing all of our eggs in one basket. Pollster.com and RealClearPolitics.com both do a great job of aggregating these polls. Take Pollster's compilation of national polls in the presidential race, for instance. Each of the dots on the chart represent data points from individual polls, the red ones representing McCain, the blue ones Obama. The red and blue lines denote the mean of these individual data points. Listed below are the series of individual polls, along with dates compiled, incorporated into this longitudinal look at the 2008 contest.
As the election nears, it is also important to be wary of last-minute polls, for their may push undecided voters into a decision they are ill-prepared to make. Moreover, undecided voters at a late hour are much less likely to actually vote on Election Day than those already committed to a candidate. Also, undecided voters may break decisively to one candidate or the other, throwing off any number of previous polls.
This election cycle actually has a small, but still significant number of undecided voters who will without doubt determine the winner. I attribute this to the fact that the race is 20 months old and has been the top news story since the Spring of 2007. Many of us are very familiar (if not tired of) the two major party candidates and the central narratives in this close contest.
An unfortunate variable that must be considered in this year's clash is the so-called Bradley Effect, named after former Los Angeles Mayor Tom Bradley, who ran and lost a race for Governor of California even though late polls showed him with a healthy lead. The hypothesis states that low income white voters (<$50,000 in annual income) tend to over report their proclivities to support African-American and other minority candidates to poll administrators, later voting for the white opponent upon entering the voting booth. Given that Obama is the first African-American candidate for president, the Bradley Effect must remain on our radars. True, Obama did exceedingly well in a number of predominantly white states, specifically Iowa and Wisconsin. However, New Hampshire is similarly homogeneously white, and polls there predicted an Obama victory in the primary over Clinton only to see her squeak out a 2 point win.
A final word about polling methods is also warranted. I have long been concerned about pollster's almost exclusive reliance upon land lines to conduct surveys, partially because I have been almost exclusively a cell phone user myself for a good part of the last decade. The typical defense was that cell phone only voters were decidedly young, and more fickle in their turnout on Election Day. More recently, in 2004 we witnessed an uptick in voting amongst the 18-29 demographic, and early indications in this election cycle predict a record turnout among young people, a seismic shift that indeed could decide the election. Some polling organizations have incorporated cell phone reliant users into their samples, negating my concerns, but I caution you to be vary of those who have not.
This initial overview was just that, and I invite you to forward poll-related questions my way that I promise to address in future posts. Leave a comment below, or send me an email at shealy@freedommuseum.us. Wednesday's edition will move from national to state polls, taking into account the fact that this election will be decided by a handful of swing states. Until then, pardon the pun, but I'm polling away.
As polling data is collected and analyzed, keep in mind that public opinion is never stagnant, but instead in continuous flux. Polls are snapshots of how the electorate feels at a specific moment in time. In order to make meaning of this deluge of information, we are better served by longitudinal looks at public opinion where trends become evident and day-to-day fluctuations balance out. Gallup provides a great service through their daily tracking poll, allowing us to view the contest along a longer trajectory. Also, notice that their daily polling figures are a compilation of three days of data, providing a longer time horizon.
Any responsible polling organization releases its margin of error when presenting survey data. This figure is a product of sample size, and accounts for the fact that samples are but a small fraction of the general population. If samples are representative of the population as a whole, or of likely voters, the margin of error allows for wiggle room in either direction. The aforementioned Gallup poll, for example, finds Senator John McCain leading Senator Barack Obama by a mere 2 points, 47%-45%. With a margin of error of plus or minus 2%, this means that Obama could actually be leading the race, 47%-45%, or that McCain is up by a larger margin, 49-43%. The race, according to Gallup, is a statistical tie. However, the 2 point spread represents their most accurate projection of the national electorate over the three-day period studied.
As mentioned earlier, as consumers of public opinion polls, it is best to be wary of any single poll as a predictor of pending election results. In a sense, we try to avoid placing all of our eggs in one basket. Pollster.com and RealClearPolitics.com both do a great job of aggregating these polls. Take Pollster's compilation of national polls in the presidential race, for instance. Each of the dots on the chart represent data points from individual polls, the red ones representing McCain, the blue ones Obama. The red and blue lines denote the mean of these individual data points. Listed below are the series of individual polls, along with dates compiled, incorporated into this longitudinal look at the 2008 contest.
As the election nears, it is also important to be wary of last-minute polls, for their may push undecided voters into a decision they are ill-prepared to make. Moreover, undecided voters at a late hour are much less likely to actually vote on Election Day than those already committed to a candidate. Also, undecided voters may break decisively to one candidate or the other, throwing off any number of previous polls.
This election cycle actually has a small, but still significant number of undecided voters who will without doubt determine the winner. I attribute this to the fact that the race is 20 months old and has been the top news story since the Spring of 2007. Many of us are very familiar (if not tired of) the two major party candidates and the central narratives in this close contest.
An unfortunate variable that must be considered in this year's clash is the so-called Bradley Effect, named after former Los Angeles Mayor Tom Bradley, who ran and lost a race for Governor of California even though late polls showed him with a healthy lead. The hypothesis states that low income white voters (<$50,000 in annual income) tend to over report their proclivities to support African-American and other minority candidates to poll administrators, later voting for the white opponent upon entering the voting booth. Given that Obama is the first African-American candidate for president, the Bradley Effect must remain on our radars. True, Obama did exceedingly well in a number of predominantly white states, specifically Iowa and Wisconsin. However, New Hampshire is similarly homogeneously white, and polls there predicted an Obama victory in the primary over Clinton only to see her squeak out a 2 point win.
A final word about polling methods is also warranted. I have long been concerned about pollster's almost exclusive reliance upon land lines to conduct surveys, partially because I have been almost exclusively a cell phone user myself for a good part of the last decade. The typical defense was that cell phone only voters were decidedly young, and more fickle in their turnout on Election Day. More recently, in 2004 we witnessed an uptick in voting amongst the 18-29 demographic, and early indications in this election cycle predict a record turnout among young people, a seismic shift that indeed could decide the election. Some polling organizations have incorporated cell phone reliant users into their samples, negating my concerns, but I caution you to be vary of those who have not.
This initial overview was just that, and I invite you to forward poll-related questions my way that I promise to address in future posts. Leave a comment below, or send me an email at shealy@freedommuseum.us. Wednesday's edition will move from national to state polls, taking into account the fact that this election will be decided by a handful of swing states. Until then, pardon the pun, but I'm polling away.
0 Comments:
Post a Comment
<< Home