Polling Propaganda

While “straw” polling had been used in presidential campaigns since early in the Republic, its use to study political opinion became commonplace in the 20^th century and its scientific underpinnings were much improved by George Gallup, a pioneer in the field.

The 1948 presidential election pitted incumbent Democrat Harry Truman against Republican challenger Thomas Dewey. For that election, Gallup decided, for the first time, to use telephone polling instead of more expensive door to door pollsters historically used to conduct interviews. Additionally, Gallup ceased polling for the Truman-Dewey election in October.

Based on his polling data, Gallup predicted that Dewey would win the election with 50% of the vote and Truman would lose with 44% of the vote. The remaining 6% would be divided among third party candidates. So convinced of the inevitable winner, the Chicago Tribune went to press early with its November 4, 1948 edition and the headline "Dewey Defeats Truman" later made famous by Truman holding the paper for a photo-op.

In fact, the actual voting results were almost the opposite of Gallup’s prediction. Truman got 50% of the votes, Dewey got 45%, and the third party candidates got 5%. Gallup unknowingly made two polling blunders that have been memorialized in textbooks on sampling errors. First, the decision to poll by telephone, a device which in 1948 tended to be in homes of well-off families, often Republican in their politics, introduced a pro-Dewey bias into the polling. Second, the decision to stop polling several weeks short of the election failed to spot last minute vote changes and decisions that favored Truman. Post-election research revealed that 14% of Truman’s voters didn’t decide for him until two weeks before the election and 3% didn’t decide for him until Election Day.

Polling methodologies and technology have improved since Gallup began to develop his craft, but it remains an imprecise mixture of art and science. For a number of reasons, even the best efforts of poll designers can’t eliminate all sources of error that contaminate the reliability of their findings. The expense of polling – recently around $50 to $100 per person interviewed – forces pollsters to work with the smallest samples that predict outcomes with tolerable margins of error, usually about +/- 3% in samples of around 1,000 people. Polls must be paid for or conducted by the organization that wants the results, which can introduce bias into sample selection. And because the target of the polling – voters in the case of elections – can change their mind or delay deciding until the last moment, polling has to be repeated often. This makes national election polling especially expensive.

Because people are jealous of their time and privacy, telephones are about the only practical way to conduct a poll today. Yet, about a third of all homes no longer have land line phones, relying exclusively on cell phones whose numbers aren’t published and can’t be robo-dialed under federal law. Cell phone-only households tend to be younger people with Democrat leanings, so the difficulty of getting input from a third of the households introduces a bias. Moreover, land line telephone polling oversamples the elderly, who are at home more often than the working age population. Poll designers in recent years are finding that 38% of their telephone calls aren’t answered and 53% of the calls that are answered won’t participate, leaving 9% of the attempted calls willing to answer questions. That means only one in ten of a sample can be reached.

A truly random sample – especially with a sample size around 1,000 – may not reflect the makeup of the expected voter turnout. For example, a registered voter is not necessarily a likely voter. Some who are registered to vote people will sit out elections. Women represent 51% of the population, the racial makeup of the population is 67% white, 13% black, and 15% Hispanic, and the age makeup is 13% over 65 and 24% under 18. Suppose a truly random sample of 1,000 people was created but it consisted of 60% females, 75% whites, and 30% over 65. Its randomness doesn’t reflect the population and could produce misleading voter preferences. If adjustments are made to reflect the population, what gives? Randomness.

One of the most difficult things to simulate in polling samples is the party affiliation turnout. It’s generally believed that long-term political identification averages out to about a third each for Democrats, Republicans, and Independents. But this varies from election to election. Polling of 15,000 adults by Scott Rasmussen last month showed self-identified Republicans at 37%, Democrats at 34%, and Independents at 29%. The 2012 voter turnout, however, could be quite different.

Based on exit polls, the 2008 turnout was 39% Democrat, 32% Republican, and 29% Independent (which is referred to as D+7) whereas the midterm 2010 turnout was 35%/35%/30% – i.e. a D+0 turnout in which Republicans gained seats in the Senate, won the majority in the House, and gained majorities in state legislatures.

The 2008 election landslide for Obama was extraordinary for a number of reasons, not the least of which was the black and Hispanic voter turnout. Blacks, who historically cast 11% of the presidential vote, cast 14% in 2008, and Hispanics, who historically cast about 7% of the vote, cast 9% in 2008. The turnout among college age voters was double the historic norm.

When we look at the 2000 turnout, which was 39%/35%/27% (D+4), and the 2004 turnout, which was 37%/37%/26% (D+0), we see how unusual the 2008 turnout was with its D+7 weighting. Then, after the Obama administration in office for two years, the 2010 midterm had a D+0 turnout. If we want to know which candidate is ahead at the moment, we must guess at what the turnout will be. But the 2008 D+7 tilt most certainly will not be repeated in 2012 – which brings me to the point I want to make in this blog.

When you look at the major polls, the edge one candidate has over the other varies considerably from poll to poll. How can this be? There are several reasons. One is the difference in polling methods – there are tracking polls, such as Rasmussen, Gallup, and Investor’s Business Daily (IBD) and there are interval polls. Tracking polls are conducted daily and presented as a moving average – often over a moving three-day period. Interval polls are conducted over, say, a week or two week period. A tracking poll is continuously updated; an interval poll isn’t. It’s important to pay attention to the date of interval poll because something substantive may have happened after it was conducted – a debate performance, a political story develops legs, an economic report is released – making the poll worthless since it hadn’t factored in new developments.

Another reason poll results aren’t comparable is who they poll. Some polls survey adults – those 18 years old and older – some survey registered voters, and some survey likely voters. Over the last 30 years a little over 50% of the adults and about 70% of the registered voters have voted in presidential elections. Only likely voter polling is reliable in comparing the relative positions of candidates. Polling among the other two groups is a popularity contest.

Yet another reason for the variance among polls is the turnout model used. As shown above, the pollster can oversample race, gender, and age, but even more important is oversampling political affiliation. Some polls don’t give the internal details of the survey like sample size, margin of error, and the turnout model used. You can guess that the margin of error of most polls is +/- 3 or 4. A poll less accurate than that is no better than throwing darts. But if the sample is less than about a thousand or there is no information on the turnout assumptions, ignore its estimate of who is leading whom. A poll is worthless if it won’t reveal the turnout assumptions on which its sample composition is based.

In the six presidential elections that preceded the Obama election, the turnout average was less than D+3. Yet last week the Washington Post – which is tied with the New York Times to win the Obama Outstanding Media Cheerleader Award – came out with a poll using a D+9 turnout model that gave Obama a 3-point lead over Romney. Keeping in mind that the 2008 Obama Coronation was a rare D+7, the Washington Post poll at D+9 is laughable, and even with its unprecedented oversampling of Democrats, it couldn’t get Obama’s lead outside of the margin for error.

Each percentage point which oversamples Democrats takes a point from Romney, so if the Washington Post poll is adjusted to a more reasonable D+3, Obama would go from a 3-point lead to a 3-point deficit, and a D+3 or less is probably the true state of the popular vote race nationally right now. Why? Because of at least three reasons. First, the second debate was essentially a tie, so Romney lost no ground and Obama gained none. If anything, Candy Crowley’s boorish correction of Romney’s Benghazi assertion hurt Obama. Second, Romney has the “enthusiasm advantage” among his supporters versus Obama’s (over 60% support Romney enthusiastically – double the number McCain had in 2008.) And third, Romney has a 10-point lead among Independents, and Independents decide election outcomes. All things considered, we could see a turnout of D+0 which would assure Romney a large victory.

All of this is known to the Obama camp, including their sycophants in the media and pollsters. So I believe there is another reason we are seeing and will continue to see (until the election) more polls showing Obama ahead of Romney or polls with a Romney lead stuck in the margin of error – i.e. essentially tied. That reason is to demoralize Romney supporters. More than in any past election, polls in the 2012 election are being used by the media to impact the election rather than to report its status.

That’s not paranoia. One important voter group is still in play. The undecided voter. Depending on which poll you look at, the percentage of likely voters who are still undecided is between 3% and 5%. Let’s split the difference and call it 4%. In the last several presidential elections, undecided voters have broken for the challenger. Obama-biased polling provides fodder to the mainstream media to trumpet that Obama will be reelected … four more years … so the undecided voter needs to go with the winner and get on the Obama bandwagon. So far this tactic is succeeding. Even though the electorate is about equally split between the two candidates, about two-thirds of them believe Obama will be reelected. It’s a mind game. It’s polling propaganda.

The D+7 political party mix of 2008 will not reoccur in 2012 – for several reasons.

Despite last week’s phony jobs report, which proved a fluke when this week’s numbers were released, half of black teenagers can’t find jobs and the unemployment rate among minorities is 14% – almost twice the national average. Obama will not get a repeat of the black, Hispanic, and Independent vote he got in 2008. Obama is currently polling 85% among blacks, down 10 points from the 95% he got in 2008. College graduates can’t find jobs. They aren’t likely to vote for Obama in the historic proportion they did in 2008.

Romney is now polling above the historic norm for Republicans among Hispanics and Jewish voters. In the gender demographic, Obama led McCain 56% to 43% among women in the 2008 election; Romney is tied with him.

Gasoline was under $2 a gallon when Obama took office. It’s almost twice that expensive now. Obama has declared war on fossil fuels, especially coal. The United Coal Miners Union refused to go to the Democrat Convention. Are they and the others laid off by Obama’s energy policies going to vote for him (again) in two weeks? I don’t think so.

Seniors are the most reliable voting bloc in the electorate. They are having problems finding doctors who will take Medicare patients. With all they’ve heard about ObamaCare and the theft of over $700 billion to help pay for it, do you think they will vote to keep him in office? I don’t. Romney has a 5-point lead among over-65 voters.

How many people do you know who didn’t vote for Obama in 2008 but will vote for him in 2012? How many do you know who will admit to voting for him in 2008 but have said they won’t vote for him again? Obama is set to lose votes he had in 2008.

My advice is to ignore the polls in the last two weeks of this campaign. They’ve been weaponized. But if you’re hooked on polls, then pay attention only to tracking polls like Rasmussen and Gallup and to UnskewedPolls.com.

But setting polls aside, ask yourself instead why aren’t you seeing 2012 Obama bumper stickers on cars? Where are the Obama signs along roads and in yards? Why are Obama’s crowds getting smaller and why is his intake of cash declining? If Obama is cresting a wave, why can’t he get his job approval ratings above 50% and why has his job approval fallen 20 points since he took office?

Assume that Obama’s desire for a second term imposed some modicum of restraint on his first term agenda. There will be no restraint on his second term agenda because he can’t run again. Maybe that’s why he won’t talk about where he wants to take the country if given four more years. So ask yourself this. Do you believe that a guy whose ideology created an economic disaster at home and a foreign policy disaster abroad can persuade voters to give him the blank check that a no-restraint second term represents? At this point in time – late October – Obama had a 6-point lead in 2008. Today Gallup gives Romney a 6-point lead, Rasmussen a 2-point lead, and UnskewedPolls.com gives Romney a 5-point lead.

Obama is headed for defeat on November 6. If the candidates are tied today and all or most of the undecided voters break for Romney, Romney wins a popular vote spread of three or four points. If Romney is ahead today, as I believe he is, the spread will be greater. But Romney has to win at least 270 electoral votes. Among the 11 battleground states, three – VA, NC, and FL are no longer in play. Romney is leading there. I will concede CO, OH, PA, and WI to Obama – a total of 56 electoral votes – even though I believe Romney can win WI and CO. If he doesn’t, Romney wins with 276 electoral votes. If he wins WI and CO, Romney wins with 295 votes.

So far, Obama hasn’t spent money or campaigned in PA. If he does either, he’s in trouble. PA has 20 votes.

November 6 is your chance to rise up. Make your voice heard. Vote. Change the future. Send this blog to your friends. And if you know of someone who plans to vote for Obama, ask this question, “What about the last four years have you liked so much that you want four more years of it?”