Joey: You don’t know what these numbers told you. I’m an expert, and I don’t know what these numbers just told you.

Josh: Numbers don’t lie.

Joey: They lie all the time.

Numbers lie all the time. Yet the business of numbers is big business. Big data helps companies make better decisions by extracting key insights from piles of information. Polling does the same for politics.

And like big data for business, though results can be ambiguous and notoriously hard to interpret, polls are essential tools for political campaigns. When pundits and politicos talk about “the polls,” they’re referring to a bevy of companies and universities that perform specialized research.

Modern political polls are generally conducted through telephone surveys that target population samples based on demographic and psychographic criteria. Poll recipients are run through a barrage of carefully worded questions about personalities, messaging, and policy. Political data site RealClearPolitics publishes a question example.

READ: The case for building a collaborative organization (Tech Pro Research story)

Quality polls use random sampling to determine who is called. Random digit dialing, a widely used tactic, is based on manual selection of the area code and phone number prefix, then picks the last four digits of a phone number at random. This tactics is great at getting usable local responses to questions, but bad because it doesn’t exclude business, Skype, and other non-human numbers.

Registration-based sampling, another common polling method, is based on available data, usually voter registration lists and other data provided by the party. These polls can be less expensive, because voter data is already captured.

Tracking, policy, benchmark, and opinion polls are often commissioned by campaigns digging for information about voter segments. The final product, presented to the campaign or client, is typically a document that breaks down recipient responses by age, gender, location, income, and political association.

Big data played a major role in informing the Ted Cruz campaign who to microtarget and poll during the primaries. Computers are prohibited from calling cell phones, but humans can manually dial numbers from spreadsheets generated by computers.

Are polls accurate? Poll results can seem inconsistent because the business of polling is diverse. Gallop, a respected data firm that provides deep analytics to large organizations, routinely publishes a survey on its election polling accuracy.

Data site FiveThirtyEight tracks the accuracy of dozens of polling agencies, ranging from local regional specialists to national firms and large research universities. While accuracy depends on the type of poll and the agency conducting poll research, political polls have been fairly accurate.

A Bloomberg survey seems to confirm that, during 2016 political cycle campaign, polls have been reliable. Bloomberg used RealClearPolitics aggregate polling data in a month-long survey during the primaries, which showed that 86 percent of polls accurately forecast election winners.

READ: 2016 by the numbers: Should Donald Trump be winning? (CBS News)

Though polls have been accurate this cycle, the industry appears to be in a period of transition. Mobile devices have had a significant impact on poll response rates. Because many polls are conducted over the phone, consumers are more able to block or dismiss unknown calls. A Vanderbilt survey found that response rates to robocalls are abysmal.

Big data companies are hungry for policial dollars and attempting to augment or supplant polling data with web and mobile user data. During the recent Republican National Convention, several polling experts indicated that the polling and big data industries could someday merge. A number of private data companies–Cambridge Analytica, Wilson Perkins Allen Research, L2 Political–already provide campaigns with mission critical data. Several political data companies see politics as a stepping-stone to other industries. After the general election, Cambridge Analytica intends to apply its personality modeling data research to the financial services, advertising, and healthcare industries. “We chose [politics] because those candidates were the best way to enter the market,” said CEO Alexander Nix.

Because of the broad range of polling agencies, using polls to track the general election can be challenging. According to the above sources–Bloomberg, RealClearPolitics, FiveThirtyEight, and Pew–these are five of the most accurate political polling agencies to follow this year.


Type of organization: Private firm

Core competency: Metrics-based enterprise performance optimization.


Type of organization: Nonpartisan research institution

Core competency: Public policy-focused analytics

Quinnipiac University

Type of organization: University

Core competency: Research

Monmouth University

Type of organization: University

Core competency: Research

Reuters Ipsos

Type of organization: Media

Core competency: Syndicated news, research, and data organization

ABC News/Washington Post

Type of organization: Media

Core competency: Broadcast news

CBS News

Type of organization: Media

Core competency: Broadcast news

Note: CBS is TechRepublic’s parent company.

SEE: Seven ways to build brand awareness into your digital strategy (Tech Pro Research report)

Like big data, polls are useful, but are not Delphic orbs. “Nothing can predict the future, and polls won’t tell you what’s going to happen,” said political scientist and TechRepublic research partner William Stodden. “Polls only provide a snapshot of the current date and climate.”

Read more