Tag: statistics
2016 Could Mark Telephone Poll’s Last Stand

2016 Could Mark Telephone Poll’s Last Stand

By Shawn Zeller, CQ-Roll Call (TNS)

WASHINGTON — Ask any pollster how confident they are that voters will really choose Donald Trump, the demagogic businessman, as the Republican presidential nominee, and they’ll say, “Not very.”

Then again, Trump is so far ahead, how could the polls be wrong? The lack of confidence is striking, but it’s because the traditional telephone poll has become a creaky, expensive and outdated system for gauging Americans’ preferences.

With cheaper, more flexible Internet polling gaining precision, this could be the phone poll’s last stand — especially if the pollsters mess it up. And there’s a good chance they will. The problem, simply put, is that people just don’t answer their phones for pollsters anymore. The decline of landlines, prohibitions on autodialing cellphones and the proliferation of caller identification systems all contribute to the weaker response, as does a general suspicion of solicitations from strangers.

Like a broad swath of industries before it, polling is on the verge of succumbing to the Internet’s transformative force. “I think the Internet poll is where we are going,” says Cliff Zukin, a professor of public policy and political science at the Eagleton Institute of Politics at Rutgers University. “Four years from now, maybe we won’t have telephone polls.”

That’s a bold statement coming from a past president of the American Association for Public Opinion Research, the polling industry’s professional society and upholder of its standards. When The New York Times decided to work with the Internet polling firm YouGov during the 2014 campaign, the association issued a scathing critique of its abandonment of scientific rigor.

But efforts to forestall the inevitable haven’t helped the music, newspaper or book retailing industries. Each has had to adapt, and the polling industry will too.

The conventional wisdom has it that the phone poll’s gravest problem is that people no longer pick up for callers they don’t know. And that is a big problem. When pollsters call nowadays, less than 10 percent of their calls reach someone willing to answer their questions. As recently as 20 years ago, they were successful a third of the time.

But the more serious threat is actually the Internet’s power as a communications tool, specifically its ability to reach many more people than a telephone at a fraction of the cost.

“The restrictions are getting more and more severe” for phone polls, says Douglas Rivers, the chief scientist for YouGov, which has also worked with CBS News and CQ Roll Call’s parent company, The Economist Group. “The phone poll is disappearing. How long it will take isn’t clear.”

Polling has always been one part information gathering and one part analysis of that information. When people answered their phones, polling was more science (information gathering) than art (analysis). Now it’s the other way around.

Rivers’ title aside, Internet polling lacks the scientific rigor of the traditional phone poll. The best phone polls take a random sample of all phone numbers and call them. The typical Internet poll lures respondents with Web ads. The results are not a random sample.

But that distinction, once considered a critical differentiator between a good poll and a bad one, is fuzzier now that so few people answer their phones.

“If you have a random sample and combine that with a 90 percent nonresponse rate, you now have a nonrandom sample,” says Andrew Gelman, director of the Applied Statistics Center at Columbia University and an advocate of Internet polling.

Adds Rivers: “The claim that it’s random is ludicrous.”

Campaigns are already experimenting with Internet polls, if not to gauge their standing in the race, then to test advertising messages. And the political parties are looking at them as well: “I don’t care where we get our data as long as it’s the most accurate data collection possible,” says Daniel Huey, a senior adviser at the National Republican Senatorial Committee.

Of course, Internet pollsters must analyze their data and make projections about voter turnout, just like phone pollsters. But they have a big advantage: They can collect vastly more data and do it cheaply.

With all the problems of phone polling, and the cost incentives that favor the Web, it will be tough for candidates and parties to resist the urge to switch, especially when the results are equivalent or better for the Web pollsters. And they are getting there: When polling analyst Nate Silver issued his report card for pollsters following the 2012 campaign, Internet polls held their own, filling four of the top seven spots in the rankings. Gallup, long the gold standard in phone polling, finished last, having predicted a narrow victory for Republican candidate Mitt Romney.

©2016 CQ-Roll Call, Inc., All Rights Reserved. Distributed by Tribune Content Agency, LLC.

Photo: Who wants to answer a pollster’s calls? Maria Teresa Ambrosi via Flickr 

 

Census Bureau Wants To Stop Collecting Marriage, Divorce Data

Census Bureau Wants To Stop Collecting Marriage, Divorce Data

By Adam Belz, Star Tribune (Minneapolis) (TNS)

The U.S. Census Bureau is proposing to eliminate a series of questions about marital history from its ongoing American Community Survey, now the only reliable source of information on marriage and divorce rates in the United States.

Demographers and sociologists are asking the Census Bureau to keep the questions, pointing out that if they are dropped, the United States will become the only country in the developed world that does not generate annual age-specific rates of marriage and divorce, and would lose its only reliable measure of divorce rates.

“It’s an example of the federal statistical system breaking down,” said Steve Ruggles, director of the Minnesota Population Center.

One of the questions up for elimination — “In the past 12 months did this person get married?” — was first asked in 1850, said Ruggles, a historian at the University of Minnesota who studies divorce and marriage. He says the questions and what they tell us about society are irreplaceable, and now more necessary than ever.

“The drop in marriage among young people is just extraordinary. I’m projecting that about a third of the people who are currently 20 to 24 years old are never going to get married, and that’s completely unprecedented in American history,” Ruggles said. “So this is a bad time to stop collecting any data on it. It’s an amazing transformation.”

The survey, sent annually to about 3 million U.S. households, asks whether in the past 12 months a person has been married, widowed, or divorced. It also asks how many times a person has been married and what year they were last married, which gets at how long American marriages last.

Those five questions about marital history are among seven on the chopping block as the Census Bureau tries to cut costs.

Since the answers to the marital history questions come with information on the age, race, education, and income level of the people divorcing or getting married, they allow demographers to analyze divorce and marriage rates in connection with a series of economic and cultural factors.

Researchers are able to figure out divorce rates by ethnic group, or by education. Sociologists use the data to investigate divorce rates and marriage duration during recessions, or in states with different levels of religious conservatism. They can come up with millennial divorce rates in the top 25 cities.

When people call Susan Brower, Minnesota’s state demographer, with questions about marriage, they often want to know what older adults’ households look like, what their economic situation might be and whether they’re on their first, second or third marriage.

“All of these things have real consequences when you’re thinking about how people set up their households, how they share income, how they care for each other,” Brower said. “These are going to be really important questions in coming years as we move into the aging of the population.”

The Census Bureau wants to drop the questions on marital history because they are not required by law and are not used for county-level, tract-level, or metropolitan-area analysis. As Ruggles points out, they were never meant for local analysis because the sample is too small. “These questions aren’t designed to do that,” he said.

Two other questions are proposed for elimination. One asks whether a business or medical office is located at the property where the respondent lives, and one asks what bachelor’s degree the respondent has earned. Demographers argue the bachelor’s degree question is important, too, since it is used to study the characteristics of the workforce.

The comment period for the proposed changes to the American Community Survey ends on Dec. 30.

“The Census Bureau is in a tough spot,” Brower said. “But overall my sentiment is that every single question on this survey is very important.”

Photo via WikiCommons