Monday, September 8, 2008

Are Poor Age Samplings Skewing the Polls?

Now, obviously, things aren't looking as good for Barack Obama as they were a month or so ago. The United States is, in fact, seriously at risk of four more years of decline in economic and political strength, unconstitutional power grabs, and environmental degradation under a McCain/Palin presidency. The polls range from a tie between McCain and Obama to a poll (probably an outlier, but still worrisome) showing McCain ten points out.

However, in looking at the sample used by The Diageo/Hotline poll (the only recent poll to have released data from its sample), I can't help feeling hopeful that the sample is skewed to make McCain look more favorable.

The reason I think this is that their polling sample skews far older than the likely electorate. The U.S. Census Bureau reported the following breakdown of those voting in 2004 by age (See Table 1):

18-29: 16.00%
30-44: 27.33%
45-54: 21.32%
55-64: 16.32%
65+: 19.03%

For the Diageo/Hotline poll, the samples are as follows:

18-29: 6%
30-44: 20%
45-54: 20%
55-64: 22%
65+: 28%
Unknown/Refused: 4%

That's a pretty significant skewing. Their sample has 50% of registered voters (like some other firms, they haven't bothered trying to construct a likely voter sample; USA Today/Gallup did, and they had a 6 point gap between their results registered voters (McCain +4) and likely voters (McCain +10)); I have not been able to locate crosstabs) over the age of 55, though this age group made up only 35% of those voting in 2004, and has voters under 45 making up only 26% of voters, though they made up 43% of those voting in 2004.

While the electorate has continued to age, given the massive registration drives by the Obama campaign and the significant enthusiasm among younger voters this election, it's highly, highly unlikely that those over 55 will increase their share of the electorate by even more than 5 points. More than 10 points is pretty much out of the question, much less 15. It's far likelier that their share will remain constant or even stagnate than increase by more than 5 points.

For instance, between the 2000 and 2004 elections:

The percentage of citizens over 18 who were over 55 grew from 29.27% to 31.40% due nearly entirely to the oldest baby boomers (defined here as those born between 1946 and 1964) starting to reach old agehood during that time period; those between 55 and 64 went from 12.2% to 14.3% of those over 18, while citizens over 65 just barely increased from 17.07% to 17.1% (well within the margin of error, though a small increase is likely).

Meanwhile, the percentage of citizens over 18 who were under the age of 45 shrunk from 51.83% to 48.80%, largely due to all but the very youngest boomers having left the 30-44 age group (down from 30.72% to 27.94%); the share of 18-29 year olds remained fairly constant to within the MOE, 21.1%-20.86% as more Generation Yers reached voting age;

Given that even more baby boomers will have moved into this category by November 2008 (including my parents), I'd expect another spike, placing voting-age citizens over 55 at between 32% and 35% of the population, and a decrease in voting-age citizens under 45, placing those citizens at between 45 and 48% of the population.

However, that will likely be lessened by voting trends. For instance, the percentage of those casting ballots who were under 45 decreased only a little, from 44.55% to 43.33%, due to a major increase in registration (and turnout) among those under 30; from 54.87% of citiens in 2000 to 60.03% of citizens in 2004, and a smaller increase (but still larger than for those 45 and over) among those between 30-44.

Similarly, those voting increased from only 55.45% to 56.67%.

Given registration trends in those few states with numbers by age and enthusiasm factors, I expect the percentages of those voting in these age groups to change by less than 1%.

I'm not sure whether polling firms had similarly poor sampling in 2004. However, if it had shown up, it wouldn't have made that much of a difference. Looking at CNN's 2004 Exit Poll (which, while not without its own problems, is far, far more likely to have its margin at within < 5 points of the actual numbers than a telephone opinion poll)

While Kerry did do significantly (15 points) better among those under 30 (won 54-45) than he did among those 30 and up (lost 47-53), among those over 30, there was very little if any differences based on age, where every group listed (30-44, 45-59, 60+, and 65+) was between 51-48 and 54-46, a 5 point margin difference.

This is not going to be the case. Barack Obama has excited the youth of this country, who are more progressive (in particular on social and environmental issues [which we have to live with longer] as well as foreign policy) than older Americans; Obama is expected to do particularly well among those under 30, who came of political age during the era of disgustingly hyper-partisanship in either the good years of Bill Clinton or the awful, awful years of George W. Bush, tired of the way things are in DC but wanting to change them.

On the flip side, several factors are combining to hurt Senator Obama among those over 65. First of all, his being African-American. Those whites who were 65 or older formed their political views either during the civil rights era [those who will be 65 by this November were old enough to vote for Lyndon Johnson in 1964 unless they were born a day or so too late in November] or before, may well have participated in white flight, and for various reasons are more likely to have negative view of African-Americans.

Second of all, older people tend to want "experience," at least aside from the ones who remember and resented the same charge being thrown at JFK (though he and Nixon actually both had 14 years in federal office [though Nixon moved up to the Senate 2 years earlier and then spent 8 years as VP]); at any rate, those generally considered to be our 3 greatest presidents didn't have all that much government experience.

  • George Washington had been appointed a surveyor at one point; other than that, he'd just been a soldier (albeit reaching the office of General).

  • Abraham Lincoln had a grand total of 2 years in office, in the House of Representatives over 10 years before running.

  • FDR had 2 years as Governor of New York before beginning his run for the presidency (he'd also been Assistant Secretary of the Navy for 8 years and a state Senator for 2).

I guess in these scenarios, here's who each person would be.

McCain = Stephen Douglas ("little", long-time Senator, faux maverick who caused somewhat of a breakaway [with Ron Paul/Bob Barr = combination of Breckinridge and Bell; though with less support], died soon after the 1860-election)


George Bush and John McCain combine as Herbert Hoover.

Back to older people and Senator Obama. The third problem is that they're less used to having to be skeptical of the media, because for most of their life, it functioned very well and only recently descended into its current form. Moreover, they are more likely to believe (and be scared by) the "He's an anti-patriotic Muslim" e-mail due to this trust in media and general less understanding of technology.

The point of this long piece being that poor age sampling may be overestimating John McCain's percentage of the vote.

No comments: