Observational Data

Elections data: We rely on elections data that come from city and county elections offices. Elections offices collect information about voters, such as whether or not they cast a ballot in a given election, as well as their age, party affiliation and address. These data are observational, not experimental.

Analyzing elections data:

While we can explore relationships between different factors that elections offices collect information on, such as voter age and turnout, we have to be very careful not to assume causality between them.

For example, we can say “On average, younger voters turnout at lower rates than older voters,” but we cannot say “People in East Portland vote at lower rates because they are younger.” Any variety of other factors may be at play.

In November when we are able to aggregate a new year of elections data into our database — the first under Portland’s new voting system — we must interpret any comparisons to previous electoral trends cautiously. We will not be able to say with certainty that the new system caused any of the potential changes we might see in voter turnout or participation. This is because many things beyond the electoral system can affect voters’ choices of whether or not to vote — from how informed and/or passionate voters feel about particular candidates or issues on the the ballot to how easy it is to vote. Even the weather can impact voter turnout!

In an experiment, researchers can control all of these factors, and can therefore be more confident that the one factor they change is the thing causing the change. Researchers can also replicate experiments over time to see if they get the same results. But we live in an ever-evolving world where many factors outside of our control may shift or change to impact whether and how people vote in local and national elections.

Survey Data

Survey data, unlike elections data, do not necessarily capture information from the whole community. Survey data is often used to explore public opinion and collect information about behavior and choices.

As with other types of data on this website, one should be careful not to draw causal conclusions from survey data (i.e., that one aspect about a person causes a certain effect).

Census Data: The U.S. Census is a complete count of the U.S. population. Other surveys rely on the Census to establish representative samples of survey respondents for the populations they hope to study. We use Census data (as well as American Community Survey data) as the demographic backdrop of our analyses.

Public opinion data: Surveys or “polls” of public opinion seek to understand public sentiment from a representative sample of the population. They use different tools to estimate who is “representative” and also to gain access to respondents. Common techniques include creating a longitudinal panel, and random phone or text surveys. Pew Research provides a deep dive on polling techniques and best practices here.

Throughout this website, we refer to public opinion research conducted by researchers who have taken care to ensure representative sampling. For example, the Oregon Values and Beliefs Center’s Portland: Your City, Your Choice survey, was weighted to be representative in terms of race, gender, age, political party, education, and Portland’s four geographic districts.