An FAQ on KQED's New Poll on Democratic Candidates

KQED’s Political Breakdown team is partnering with Change Research to gauge Californians’ support for Democratic presidential candidates ahead of the 2020 primary. Change Research will conduct polls monthly through the March primary, and KQED will be the first media outlet to publish the results.

KQED may also separately commission additional polling questions on state election issues.

Why is KQED commissioning polls?

With so much interest in the 2020 election, we see an opportunity to provide our audience with insights into the race, as well as other key issues coming before the electorate in California. We are also excited to have the opportunity to track changes in public opinion over time, as the campaigns progress and we develop a better idea of what else will be on the 2020 ballot.

How is this poll different from other polls?

Traditionally, polls have been conducted by phone — which is typically a resource-intensive, expensive and slow process. Change Research polling, however, is conducted exclusively online so it can be completed within a few days and released quickly — hopefully making the snapshot in time as accurate as possible.

FiveThirtyEight, the well-respected news site that relies on statistical analysis, gives Change Research a C+ rating based on seven of its publicly released polls, finding that it accurately predicted election results 57% of the time. But the polling firm has conducted over 800 polls and argues they have a stronger record of accuracy.

Who else does Change Research work with?

Media partners in other states include The Nevada Independent and The Post and Courier in South Carolina. The San Francisco Bay Area-based firm also has paid clients, including nonprofits, foundations, advocacy organizations, Democratic candidates and Crooked Media, which produces “Pod Save America” and other programs.

How does the polling work?

Change Research pays to put advertisements online — on both websites and social media platforms — asking people in certain geographic areas to respond to polling questions. Respondents are asked information such as their age, where they live and who they voted for in the last primary, in addition to specific questions about candidate preference. The firm uses algorithms developed by studying past voter trends to ensure it gets a large sample that reflects the demographics of a likely electorate — in this case, California voters.

After the responses are collected, Change Research again uses models based on tens of thousands of historical data points to weight (or adjust) the responses, so the poll more accurately reflects the likely electorate in California, or even a specific congressional district or city if we decide to poll in smaller contests.

Why is KQED partnering with a group that works primarily with Democrats?

While Change Research’s paid candidate clients are mostly Democrats, its polling is conducted in a nonpartisan fashion. KQED, which is not paying for the poll, is reviewing questions before surveys are conducted to make sure they live up to our nonpartisan, ethical standards. FiveThirtyEight actually rates Change Research’s statistical bias as slightly leaning Republican (meaning they have historically overrated the performance of GOP candidates); Change Research says that in 2018, it had a slight Democratic bias.

Why should anyone pay attention to polling after how off pollsters were in the 2016 election?

Many polls in 2016 were actually pretty accurate in predicting the popular vote — it’s just that presidential elections are based on the Electoral College. We hope that by focusing the poll on California’s electorate, we will get an accurate snapshot of how voters here feel — and that we can track changes in those opinions over time by conducting these surveys monthly. We are also excited about the unique opportunity to target smaller districts and gauge public opinion in races that might not traditionally be polled.

How seriously should I take any poll?

Honestly, with a grain of salt. Experts say it’s best to look at an average of many polls; and keep in mind, presidential races move quickly. A poll, again, is just a snapshot in time, and given the nature of this race, results are sure to change month-to-month.