Pollster.com

Articles and Analysis

 

Is Youth Being Served?

Topics: Age Weighting , Cell Phones , Disclosure

My NationalJournal.com column, on whether pollsters should be weighting by age, and if so, to what value, is now online. It follows on last night's post on changing weighting of the GWU Battleground survey.

Two pieces of added context. First, yesterday I emailed a handful of national pollsters to ask about the age compositions of their samples, and I got quick responses from those listed below. The table includes the estimates of 2004 turnout from both exit polls and the U.S. Census Current Population Survey (CPS). See my column for more discussion of the 2004 estimates. Note that most pollsters report on the percentage of 18-29-year-olds, which is included below:

2008-10-01 18-29-year-olds.png

Second, I want to re-post the very helpful graphic produced by the Pew Research Center last year that shows how how the rise of cell-phone only households has dramatically affected the ability of pollsters to reach younger respondents when relying on random samples of land-line telephones. The percentage of 18-34 year olds in Pew's un-weighted national samples declined by roughly ten percentage points between 2002 and 2006 -- the same time period in which the percentage of adults living in wireless only households has grown from 3% to 12% (I originally posted this chart as part of an entry last year on cell phones and their impact on political surveys).

Keeter%20graph.png

 

Comments
bythesea:

This is interesting information as it would seem to suggest that Obama may have more support than is showing in the polls. Daily Kos which has the highest % of young folks (18% vs the 2004 performance of 17.4%) has the largest margin for Obama at 9/30 (51% to 41%). Dem. Corp and Pew which seem to underweight young people (14%,15% vs. 17.4%) shows a narrower Obama edge (4% or 6%). IF youth actually turns out then they will have quite an impact.

____________________

UofI32:

I think this is a great point to make that very few polls are able to contact the hundreds of thousands of college aged students that are going to vote come Nov. 4th. I run a voter registration drive for college aged students and i can tell you with great confidence that less than 1% of the more than 5,000 students weve registered have been polled. While most elections this age group has shown up in almost pathetic numbers, this generation is more than happy to vote for their future and understand what this election means.

____________________

EarBucket:

The map on your front page should read 250 EVs for Obama, not 260. Your addition's wrong.

____________________

Ryan in MO:

it's not wrong as soon as you put MN blue like everyone else has it

____________________

FlyOnTneWall:

Mark:

Thanks for your column, and your post. They combine to answer, rather nicely, a question I posed last week - and I appreciate it.

I did want to offer a reaction, though. Your column focuses on the importance of weighting by age in this cellular era. And its points are well taken. But I was particularly delighted by the data you present in this ancillary post, because for ordinary folks like me, it's all but impossible to obtain. Remarkably few pollsters release cross-tabs on age, and so it's rarely possible for even well-informed commentators to evaluate their conclusions. Just tonight, I noticed that the NYT/CBS poll doesn't bother to break down its sample by age brackets.

In any year, that would be insane. This year, it borders on malpractice. Age has been one of the strongest predictors of voter preferences both in the primaries and the general election. We're witnessing, in many ways, a generational divide. The numbers above are a revelation - most pollsters are undersampling young voters even relative to the 2004 turnout and registration, when the best data we have suggests that registration has soared since then. But add in the Pew findings - suggesting that those without cellphones tilt even more strongly Democratic than their age-cohort - and you have the makings of a polling fiasco.

The Battleground survey had been badly out of line with the other national tracking polls - simply tweaking its (partisan) model pulled it four points closer to them, mostly because its age sample came closer to reality. I found the strength of that effect rather stunning.

Wouldn't it be great if pollsters released their sampling crosstabs, including age, so we could all adjust their results to match our own sense of the composition of the electorate?

____________________

CTPonix4BHObama:

Who would've thought Kos would be the only one to really get it right.

____________________

nelson:

Is there any evidence on whether young people with land lines and those with only cell phones have similar political views? If not, simply weighting the land line responses won't compensate for not reaching the cell-phone only people.

____________________

msippi:

They should be looking at 1992 and 2004 to estimate youth participation. I saw one exit poll estimate from '92 that put this segment at just over 20% of voters. Like '92 there is a candidate with strong appeal to young voters. Kerry was not so appealing as it was more anti-war and anti-Bush and a good ground game. Now we have those plus a more inspiring candidate and even better ground game. In 2004 Pew, found that the youth %'s were even higher in battleground states, where the GOTV was strongest.

I understand why pollsters are conservative with these numbers. It's hard to go out on a limb when there's not a lot of history to back it up, but I think there going to be off on Nov 4.

____________________

Jon Coit:

Not a statistician--but doesn't the decline in the number of adults age 18-34 raise the possibility of sample size error (i.e. that random fluctuations in responses could have a disproportionate effect in the survey)?

It's also interesting that they've allowed the weighted sample to decline as well. Thanks for this-

____________________



Post a comment




Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.

MAP - US, AL, AK, AZ, AR, CA, CO, CT, DE, FL, GA, HI, ID, IL, IN, IA, KS, KY, LA, ME, MD, MA, MI, MN, MS, MO, MT, NE, NV, NH, NJ, NM, NY, NC, ND, OH, OK, OR, PA, RI, SC, SD, TN, TX, UT, VT, VA, WA, WV, WI, WY, PR