Polling Data: A few cautions

Polls are intelligent but inexact predictions of voters’ attitudes and behaviors based on generally accurate information reported by a mostly representative sample of said voters.

And that’s polls at their best. As you can see, there’s substantial room for error.

What can polls tells us?

If you work with low-income people and/or people whose home language is not English, the attitudes of the people you know best are almost certainly not going to show up in the polls. It’s notoriously diffcult to remove language, socioeconomic and racial bias from the polling questionnaire. It takes a highly sophisticated polling scheme and truly representative sample to discover the attitudes of such people; in an election, the costs are usually too great. And because marginalized populations don’t routinely vote, polling your base will be a low priority that may be slighted even by your allies.

Furthermore, results of early polls of low-income people and non-native English speakers are frequently wrong. In 1994, polls in California showed most Latinos favored the anti-immigrant Prop. 187 as late as the September before the November vote. In fact, over three quarters of voting Latinos voted “no” on Prop. 187 – and outrage over the initiative led to massive Latino registration and voting in 1996. All that the early polls had found was that Latinos hadn’t started to think about the election yet. This is a normal result in groups that are not heavily engaged with elections. You are likely to have a better idea of how the people you work with are going to vote and why than remote pollsters.

Where do polls come from?

When confronted with a poll that seems to require a particular electoral strategy, pay close attention to who commissioned and designed it. Why? Because:

• Polling uses sophisticated statistical models to process numerical data, but design and interpretation of polls is not a pure science.
• To greater and lesser degrees, we all think inside our boxes; pollsters and people who pay for polls are no different.
• The findings and conclusions of polls are often influenced by the preconceptions of their designers. Commissioners of polls may choose pollsters whose methods and findings reflect their shared worldview.
• And finally, both pollsters and their backers may be honest in their methods and intention but fail to frame what they find in terms that community groups can agree with or use.

In targeting, you want as much information as you can get; sophisticated demographic data brings geographic information to life. Community Voices Heard, Henry Serrano says, supplements geographic and basic demographic information with findings published by various institutions, but they maintain a healthy skepticism of the source as well as the data itself.

So the bad news is that polling data is far from perfect and you probably won’t be able to pay for your own poll, but the good news is that you already have available to you massive data about electoral behavior – and it’s free. Previous elections are larger samples than any poll. If you can identify previous campaigns that involved similar public attitudes, you can learn a tremendous amount about what works, with what groups and where.

Click to hear from an organizer

Henry Serrano: sophisticated demographic data can be biased