Friday, October 1, 2010

Why Texas Polls Showing Rick Perry Leads Bill White Might Be Wrong

Earlier this week a Belo poll suggested Republican Gov. Rick Perry leads Democratic challenger Bill White by a 50 percent to 36 percent margin. John Reynolds reports in a Quorum Report article (subscribers only) that Lone Star Project's Matt Angle believes a Belo poll screening question disproportionately disqualified likely Bill White voters and thus discredits the survey. Specifically, Angle said that the survey screened respondents by asking if they voted in most or all school, local and primary elections. Angle said that is a very different screening question from the more common "Are you likely to vote in the next General Election?"

Angle said the Belo screening question eliminated voters who might not vote in every election, but would come out for a higher profile gubernatorial election. In addition, he said the screen question concentrates Republican respondents who tend to be the higher income homeowners who regularly turn out for local school elections. He said that the poll fell outside the range seen in other polls because it is unintentionally skewed to favor Perry.

Angle is correct in his assessment that Belo's screening question heavily skews the poll's coverage to favor Perry, but that is not the whole story. A second factor further skews toward the Republican side of the question in this and most other national and Texas polls.

Most pollsters, from national pollster Rasmussen to small Texas pollsters, call only people with published number landline telephones. Political polls that do not include respondents who subscribe exclusively to wireless cell phone service produce results that skew six or more points toward the conservative or Republican side of the question. The adoption of wireless-only phone service by major segments of the U.S. population has occurred so rapidly that political experts and pollsters are scrambling to adjust to the ramifications of this telecommunications earthquake.

Over the last 10 years pollsters have increasing relied on statistical weighting of landline telephone samples to correct for the omission of wireless-only households that new research reveals have a decidedly progressive tilt. According to Nate Silver of the FiveThirtyEight blog statistical weighting no longer sufficiently compensates for the omission of the 30% of U.S. households that have dropped traditional landline telephone service. (It is estimated that 30% of all U.S. households have become wireless-only during the last half of 2010.)

The rapid increase of wireless-only Americans now has a quantifiable impact on political polling results. Specifically, excluding wireless-only adults from political surveys has a statistically significant, negative impact on Democratic performance in political polling. This was confirmed in a recent study by Pew, which compared the national generic ballot preference of a landline-only sample of 4,683 registered voters with a combined landline and cell-phone sample of 7,055 registered voters:
In the landline sample, Republican candidates have a 47%-to-41% margin over Democratic candidates on the 2010 generic horse race, but in the combined sample voters are evenly divided in their candidate preferences for this November (44% for each party). A majority of cell-only voters (52%) say they will support the Democratic candidate in their district.
There is still a margin of error in a poll with such a large sample size, but it is just barely over plus or minus 1%. As such, with an overall six-point gap, the survey shows a statistically significant difference between polls that include cell-phone only adults and polls that do not.

Telephone surveys since the 1960's, when they first grew to prominence, have traditionally relied on samples from published landline telephone numbers. The explosion of unpublished number wireless-only phone service over the last 10 years, and more recently Internet-based VOIP phone service, places a rapidly growing number of "unpublished phone number" Americans out of reach of those surveys. Nationally, more than one-in-four U.S. households now have no "published number" landline telephone, considerably more than in the early 1960s when telephone surveys were considered unreliable because so many households were unreachable by telephone.

While some pollsters are starting to include wireless-only respondents in their surveys most do not because of the higher cost to reach those wireless respondents. Many pollsters, particularly smaller local pollsters don't include wireless-only respondents because the cost of calling wireless numbers is more than double the cost of calling published number landline phone respondents.

It is difficult and more expensive for pollsters to interview wireless-only Americans because of the provision in the 1995 Telemarketing and Consumer Fraud and Abuse Prevention Act (TCPA) that places restrictions on unsolicited calls to mobile phones. The TCPA forbids calling a cell phone using any automated telephone dialing system (autodialer) without prior express consent. This rule applies to all uses of autodialers and predictive dialers, including for survey and opinion research. So, when live-interviewer pollsters want to interview respondents on their cell phone, they must first surmount the problem of compiling a list of unpublished cell phone numbers. Then the live-interviewers must place the calls by manually dialing each number.

The most recent survey from the CDC covering the period of July-December 2009 shows that 24.5% of US households, and 22.9% of US adults, were wireless-only by the second half of 2009. This is a sharp increase over the past two years.

Adults living in U.S. households with wireless-only phone service.
Source CDC/NCHA, National Health Interview Survey.

In the second half of 2009 nearly half of adults aged 25–29 years (48.6%) lived in households with only wireless telephones. More than one-third of adults aged 18–24 or 30–34 (37.8% and 37.2%, respectively) lived in households with only wireless telephones. As age increased from 35 years, the percentage of adults living in households with only wireless telephones decreased: 23.9% for adults aged 35–44; 14.9% for adults aged 45–64; and 5.2% for adults aged 65 and over. Adults of all ages living in or near the poverty level, Hispanic adults and non-Hispanic black adults are also more likely to live in households with wireless-only phone service. The most progressive segments of the population are rapidly going wireless only.

When the CDC releases its second half of 2010 wireless-only report it is expected that more than 30% of all U.S. households will have unpublished number wireless-only phone service. Last May Nate Silver of FiveThirtyEight, wrote an interesting piece examining the effect wireless-only households now have on political polls. Silver predicts that if the current adoption trends hold, the percentage of wireless-only households could be in the mid to high 30s by election day November 2012. Furthermore, the CDC wireless-only figure does not fully reflect so-called "cellphone-mostly" households. Cellphone-mostly households are households that do have a landline, but that line is used for FAX or home security systems and it rarely or never used to receive incoming calls; another 15% of the population falls into this category.

The recent explosion of unpublished number Internet-base VOIP landline service subscriptions further increases the percentage of unpublished phone number households in certain metropolitan areas with high broadband Internet coverage. Collin Co. has very high residential broadband Internet coverage.

Previous research has shown that Rasmussen's use of a likely voter screen is not the reason why their polls now differ strongly from the trendline of all other polls. Since Rasmussen excludes wireless-only adults from their surveys (possibly due to restrictions on automated phone calls to cell phones), it is likely that the wireless-only effect is one of the main reasons that Rasmussen's likely voter polls are about six points more favorable to Republicans than other likely voter polls. Also, Rasmussen polls of all adults are six points more favorable to Republicans than other polls of all adults. This six-point pro-Republican tilt in national polling results is exactly the gap found by Pew in their landline-only sample.

State-level estimates for 2007 show that wireless-only adults are particularly prevalent in Arkansas, Kentucky and Texas. While 13.6% of the nation as a whole was wireless-only in 2007, Arkansas was already 21.2% wireless, Kentucky was 21.6% wireless and Texas was 20.9% wireless (PDF, page 5). Wireless-only households in these states likely continue above the national trend line, which means the percentage of wireless-only households in these states may already be in the low to mid 30's. The lack of wireless-only adult survey coverage by Rasmussen may explain why Rasmussen polls in Kentucky and Arkansas have skewed toward the most conservative candidates in primary and and general election surveys.

While Texas is one of the states leading in wireless-only adoption the metro areas surrounding Dallas and Austin lead most other Texas counties in unpublished number wireless-only phone coverage. Given the trends nationwide, it is likely that roughly one in three of all adults in Arkansas, Kentucky and Texas are now wireless-only. This would make for an even more pronounced localized landline only coverage effect gap than the national discrepancy found in Pew's 2010 study.

So, Belo's poll showing Republican Gov. Rick Perry leading Democratic challenger Bill White by a 50% to 36% margin is very likely skewed more that six points toward the Republican side of the question by the landline only coverage effect found in Pew's study, in addition to the right skewing nature of the Belo's "likely voter" screening question.

A just released Texas Lyceum Poll shows Perry leading White 48 percent to 43 percent, a margin of only five points. Adjusting the Texas Lyceum poll results for Pew's six point landline only coverage effect puts White dead even with Perry. If the landline only coverage effect is indeed greater than six points in Texas because Texas residents have a significantly higher wireless-only adoption rate, then White could even lead Perry in a dual-frame survey including both landline and wireless-only respondents. (also see Texas Lyceum Summary of Trial Ballots)

The rapid rise in wireless-only adults, along with the confirmation that those adults have a decidedly progressive tilt, helps explain some, and possibly all, of the recent right skewing of polls from Rasmussen and other polling firms. Americans are dumping landlines at a rapid rate, and those Americans do skew heavily toward progressive political viewpoints. Pew's "landline only poll coverage effect" finding puts a lie to the national media's mime that the nation's political mood has shifted so far right that the Tea Party movement now represents mainstream America. The truth is that Tea Party supporters still rely on old "copper wire" landline phones and they are the people being polled while everyone else who has gone totally wireless are not being polled.

No comments:

Post a Comment