Why an online poll? And why SurveyMonkey?

Nov. 7, 2015 3:51 PM PT

Why an online poll?

Dating back decades, we’ve done phone polling, in which surveyors randomly dial phone numbers and pose a series of questions to people who answer and are willing to participate. (See our September poll of the presidential race as an example.) A successful interview can last 15 to 30 minutes, sometimes more.

These days, fewer people have landlines and many who do won't answer them or participate in a telephone survey. Reliable telephone surveys call cellphones in addition to landlines, but many cellphone users won't answer a call from a number they don't recognize. In many cases, fewer than 10% of people contacted by phone now complete a poll interview. That response rate used to routinely be above one-third.

As the response rate has dropped, more and more experts on polling have grown concerned that the accuracy of telephone polls is in jeopardy. Indeed, traditional polling has generated some significant errors in recent elections, some favoring Democrats, other Republicans.

Moreover, as response rates have dropped, getting an accurate sample requires making more calls, which increases costs.

The Internet provides an alternative way to reach people and survey their opinions. But it requires a different technique, just as the telephone required a different approach than the door-to-door surveys that pollsters started using generations ago.

Online, there’s no way to randomly reach people. Instead, what polling firms try to do is contact a large cross-section of the population, then winnow the sample to match known demographic measures, such as age, race, gender and education.

Debating the methods

There has been a lot of research about online surveys, and more political and media organizations are using them to take a snapshot of sentiment.

Online polling improves a survey's ability to reach a sufficient number of younger people, often underrepresented in phone polls. It also gets around a known problem of telephone surveys -- people may give false answers to avoid disclosing a view that they fear an interviewer might disapprove of.

SIGN UP for our free Essential Politics newsletter >>

Web-based surveys also can be more dynamic — respondents can be shown a political ad, for example, and be asked to respond to it. Images, videos and other graphics aren’t possible in telephone polling.

But online polling also can undersample groups of people who have less access to the Internet or who simply do not want to take surveys online. About 10% to 15% of the population either can't or won't take a survey online. They tend to be older, poorer, less educated and more likely to be members of a racial minority than people who do take an online survey. Online polls need to adjust for that.

Pew Research tackles that subject here.

Why partner with SurveyMonkey?

Many people know SurveyMonkey as a company that allows groups to conduct "do it yourself" surveys of their members. But in recent years, as the company's customer base has grown, SurveyMonkey has begun to do research projects on topics such as elections.

It has compiled an impressive track record.

Earlier this year, for example, SurveyMonkey's polls of the British election were the only ones to accurately forecast the result. SurveyMoney also correctly forecast the Republican victory in the 2014 congressional elections, which many traditional telephone surveys got wrong.

Because SurveyMonkey's customer base is so large, it has overcome one of the biggest hurdles to a successful online poll -- being able to reach a big enough universe of respondents to get a representative sample. About 3 million people take a SurveyMonkey poll each day, giving the company a huge cross-section of the population to start with.

How does it work?

When a person goes to the SurveyMonkey site to take a poll for whatever group he or she belongs to -- a church, a bowling league, a neighborhood association -- he or she may get a message about participating in a separate survey on the election. A computer program chooses which people to invite as the first step toward getting a representative sample.

In addition to questions about the election, the survey also asks basic demographic facts, including the respondent's age, race, gender and education level. When the results come in, they are weighted to make sure the sample matches what we know from the Census about those demographics. Similar weighting takes place with telephone surveys.

In this survey, respondents were recruited in English and Spanish and could participate in either language.

Don't online polls have a bad reputation?

A lot of organizations conduct surveys online which they call polls, but which make no effort to provide a representative sample of the population. Those surveys just reflect whoever decides to answer the questions and are not reliable gauges of public opinion.

This poll, by contrast, was carefully developed to provide a sample that reflects the voting public as a whole.

California is a particularly good state in which to conduct online polls because the share of the population which is online is among the highest in the country.

Going forward

This is the second online survey that the Los Angeles Times has published, so it's still a bit of an experiment for us. We'll continue to try different approaches to make sure we can give you the most accurate information about how Americans feel about issues of public policy and about candidates for elected office.