PPIC’s New Poll – Plus a Primer on ‘Oversampling’
Democrat Hillary Clinton now leads Republican Donald Trump by a staggering 54-28% among likely voters in the Golden State, according to the latest survey from the Public Policy Institute of California, and Kamala Harris leads sister Democrat Loretta Sanchez by nearly as much – 42-20%.
No surprises there, except for the whopping margins and the fact that Clinton is not only winning women 59-25% but also beating Trump among men, 48-32%. Oh, and that somehow Mr. Border Wall is winning 12% of Latinos (compared to 71% for Clinton) – voters who apparently didn’t get the memo.
The survey also reveals how important education is in the Clinton-Trump contest. She leads 55-26% among voters with only a high school education but 62-21% among college graduates. As a favor to the Abacus Department at Calbuzz, PPIC ran a special crosstab that shows that among men, Clinton leads by just 43-38% among non-college grads but 55-23% among college grads; among women, she leads by a hefty 52-28% among non-college grads but a stunning 70-20% among college grads.
If you want to see more results from the foreseeable PPIC survey, you can find it here.
The Trump-Putin theory of polling. The Calbuzz Green Eyeshade and World Peace Division has done excessive eye-rolling over Trump’s recent whining about national polls that show him losing, which is as wrong-headed as everything else he’s had to say this election season.
Among his “evidence” that the 2016 presidential election is “rigged” is Trump’s argument that by oversampling Democrats in their surveys of voters, pollsters are creating a false narrative that he is losing to Clinton. On this, he’s either a liar or he’s ignorant. Most likely both.
The Real Clear Politics average of polls – which even includes losers like the LA Times/USC and Gravis Marketing surveys – shows Clinton with about a 6 percentage point lead, while analysts like Nate Silver give Clinton an 84% chance of winning with an Electoral College landslide in the range of 330-200.
But for Trump, this is all part of a conspiracy to fix the election.
“When the polls are even, when they leave them alone and do them properly, I’m leading,” he said this week at a rally in Florida. “But you see these polls where they’re polling Democrats. How’s Trump doing? Oh, he’s down. They’re polling Democrats. The system is corrupt and it’s rigged and it’s broken.”
The “smoking gun” Trump cites is a 2008 email from Clinton campaign chairman John Podesta’s account, hacked by Trump’s pal Vladimir Putin and posted by WikiLeaks, in which a prominent Democratic strategy firm recommended “oversampling” certain voters in their polls, including blacks, Hispanics and millennials.
Oversampling 101. Since polling is a mysterious process, even for honest, intelligent voters, it’s worth taking time to understand what oversampling is all about.
It’s not about just adding certain groups to a survey in order to skew the results in favor of one candidate or another. It’s a respected, legitimate and widely-used technique to ensure that the opinions of certain groups who would normally be a small proportion of a random sample are queried in adequate numbers so that their views can be reliably understood.
Sampling works like a blood test: you don’t have to look at all the blood in a person’s body in order to analyze what’s in the blood. Instead you take a small sample that tells you what’s in the bloodstream.
In a true random sample of 1,000 voters, each with an equal chance of being chosen, probability theory tells us that the results of a survey will have a margin of statistical error, plus or minus about 3%. This gives pollsters a high degree of confidence that what they find accurately reflects the views of a much larger population – say, all the likely voters in the country.
Smaller samples, by and large, have bigger margins of error. That means less confidence.
If 12% of the likely voters (and we’re making up a percentage here) are expected to be Latinos, then in a true random sample of 1,000 likely voters only about 120 of them will be Latinos. The margin of error, however, for that 120 respondents is more like +/- 9% — not a very high degree of confidence.
But if a campaign really wants to know what Latinos think about something, one way to know is to sample more Latinos while doing the survey. They might, for example, make sure they interview 500 Latinos (chosen at random) which would create a margin of error of about 5% for them – a much more reliable degree of confidence.
Then, when reporting the results of the entire survey, those 500 Latinos are weighted so that each respondent counts just 24% so that there are only 120 Latinos in the total sample of 1,000 (500 x .24 = 120). Voila!
And now this special message. Here’s how Pew Research explains it:
For some surveys, it is important to ensure that there are enough members of a certain subgroup in the population so that more reliable estimates can be reported for that group. To do this, we oversample members of the subgroup by selecting more people from this group than would typically be done if everyone in the sample had an equal chance of being selected. Because the margin of sampling error is related to the size of the sample, increasing the sample size for a particular subgroup through the use of oversampling allows for estimates to be made with a smaller margin of error. A survey that includes an oversample weights the results so that members in the oversampled group are weighted to their actual proportion in the population; this allows for the overall survey results to represent both the national population and the oversampled subgroup.
For example, African Americans make up 13.6% of the total U.S. population, according to the U.S. Census. A survey with a sample size of 1,000 would only include approximately 136 African Americans. The margin of sampling error for African Americans then would be around 10.5 percentage points, resulting in estimates that could fall within a 21-point range, which is often too imprecise for many detailed analyses surveyors want to perform. In contrast, oversampling African Americans so that there are roughly 500 interviews completed with people in this group reduces the margin of sampling error to about 5.5 percentage points and improves the reliability of estimates that can be made. Unless a listed sample is available or people can be selected from prior surveys, oversampling a particular group usually involves incurring the additional costs associated with screening for eligible respondents.
We return you now to our regular programming.
The PPIC Statewide Survey was conducted by telephone among 1,704 California adult residents—half (852) interviewed on landline telephones and half (852) on cell phones from October 14–23, 2016. Interviews were conducted in English or Spanish, according to respondents’ preferences. The margin of error is ±4.3 percent for the 1,024 likely voters.
subscribe to comments RSS
There are no comments for this post