Research question framing

The research question is the foundation of the Policy Data Brief. Everything else follows from it. A poorly framed question leads to an analysis that is either unanswerable, too vague to act on, or answered with the wrong data. A well-framed question is practically a map to the analysis you need to run.

The three-part formula

A good policy research question has three parts:

1. Outcome: What are you measuring? (mail ballot rejection rate, voter turnout, minority share of a district's CVAP)
2. Comparison: Compared to what, or compared across what groups or time periods?
3. Jurisdiction + time: Where and when?

Example structure: "[Outcome] [in Group A / in Year X] compared to [Group B / Year Y] [in Georgia]?"

Practice exercise — evaluate these questions

Which of the following is well-formed?

• Question A: "Does Georgia discriminate against Black voters?" — Too broad, not answerable with a single dataset, infers intent.
• Question B: "Did mail ballot rejection rates in Georgia counties with majority-Black active registrant populations differ from counties with majority-white populations in the 2022 general election?" — Well-formed. Specific outcome, explicit comparison, named election.
• Question C: "How many people voted in Georgia in 2024?" — Answerable but not a policy question. No comparison, no disparity, no actionable angle.
• Question D: "After the dissolution of majority-Black districts in Georgia following Callais v. Landry, did voter turnout in those former districts decline compared to prior cycles?" — Well-formed and directly relevant to the post-Callais landscape. Outcome: turnout. Comparison: before/after Callais. Jurisdiction: former majority-Black districts in Georgia.

How to test your question

Ask yourself: (1) What data would I need to answer this question? (2) Does that data exist and is it accessible? (3) If I found the expected answer, what policy action would it support? If you can answer all three, your question is ready for the brief.