Data collection

Type of texts

Vaccination expectations may be conveyed in various formats of texts. It could be vaccination FAQs, usually on the webpages of the departmemts of health and wellness, or news articles.

We focus on one single type of text, community message, to control for the effects from the type of documents on the expression of vaccination expectations.

Occasionally, when a community message is not clear enough about the university’s vaccination expectation, we use additional text to validate our understanding (e.g. news article).

Sampling

We take a sample from all community messages following the procedures below.

Sampling frame: a list of U.S. universities from the Times Higher Education World University Rankings

Goal: selecting equal number of universities in “red” and “blue” states from the top 50 universities

Sampling strategy: selecting all 10 universities in “red” states from the top 50 and randomly sampling 10 universities in “blue” states from the remaining 40 universities

Note: Among the top 50 universities, we only got 10 universities residing in “red” states. Therefore, the sample is not balanced.

Sample size: resulting number of universities to be investigated (20)

Document level information

One university can have more than one community messages because a university may separately send messages to students and employees, or update the previous document. The unit of analysis in this workshop is document, not university, which is each community message.

Therefore, we also collected information associated with each community message, including

• university name
• global and national ranks of the university
• city, county, state
• political orientation
• published date, retrieved date of the community message

These information could help us contextualize our analysis.

---