Who creates the surveys?

This survey is created by the polling design team of the Polling and Open Data Initiative at the University of Washington, a nonpartisan registered student organization organization dedicated to better public data in the University community and Pacific Northwest. After creation, it is reviewed by the organization. Larger polls will be reviewed by academic advisers. Our goal is to collect data in an accurate and unbiased manner. Prior to publication, questions are reviewed extensively for neutrality and objectivity.

 

How do we aim to write questions?

We aim to employ simple, unambiguous and concrete language to minimize wording bias. Furthermore, multiple choice answer options are exhaustive to anticipate a wide variety of responses. Some surveys will allow respondents to choose which questions they wish to answer. This will be indicated clearly before the text. 

Survey questions are ordered by topic. All surveys start with basic demographic information for use in analysis. By default, university-wide surveys will ask respondents to provide their class and gender. 

Multiple choice answer options are exhaustive to anticipate a wide range of responses.

 

Who gets polled?

We aim to collect samples that accurately represent the target population. To achieve this, we will take a random sample of the entire target population, whenever possible. Sample size will vary based on target population size, but will be large enough to account for chance variation. 

In our analyses, we will overview the sample size and margins of error.

 

Who has access to responses?

When responses are submitted, they will then be visible to designated team members of the Polling and Open Data Initiative at the University of Washington.

How are surveys conducted?

Randomly selected individuals will be emailed a link to our survey. Surveys are either conducted through the Catalyst WebQ interface, which requires a UW NetID login, or Qualtrics. From February 2021 onward, new polls are all conducted in Qualtrics.

 

What limitations do we recognize?

Due to potential nonresponse bias and chance variation, we recognize that our samples are limited representations of our target populations. Additionally, we recognize that our sample will not include students who choose not to provide their email in a mailing list or directory. Other barriers, such as little to no access to the internet, may also impact our ability to contact a randomly sampled individual and/or receive a response.

 

Data Use and Privacy

All current surveys are conducted through Qualtrics. Some legacy surveys are conducted through the University’s Catalyst WebQ interface and are secured using a UW NetID login and the confidential survey option. For some university surveys, we require all respondents to log in with their UW NetID to ensure that they are an enrolled UW student or faculty member and to prevent the same respondent from responding more than once. 

After data is collected, anonymized and/or summary data will be shared on our website at https://www.poddata.org. Our data visualization and publication team aim to report the data in a way that is truthful, fair, and in the best interest of informing the community. Respondents’ identifying information will never be included in the visualizations or publications. If we decide to release the raw data for public use, demographic information will be removed to protect respondent privacy.

 

Weighting

If the collected sample does not accurately reflect the demographic balance, data weighting may be implemented. By default, university-wide surveys will be weighted by race, sex, class, and residency (see “How are Questions Written?” for how these are classified). 

We will perform data weighting by splitting the sample into strata that have known population proportions and are of significant size. The responses within these strata will then be assigned reasonable weights so that the weighted proportions reflect the population distribution. Individual responses will never be weighed more than 200% or less than 50% in order to maintain the integrity of the data.