How to Choose a Research Question (That You Can Actually Finish)
How to Choose a Research Question (That You Can Actually Finish)
Most students don’t fail at research because they’re not smart enough.
They fail because the question they picked was:
- too vague,
- too ambitious,
- impossible to support with evidence,
- or quietly unanswerable.
A good research question isn’t “interesting.” It’s finishable.
The goal of a research question
A research question should do three things:
1) Tell you what to measure
2) Tell you what method is plausible
3) Tell you what evidence you need
If your question doesn’t imply a method and evidence, it isn’t ready yet.
The 4-Part Stress Test (use this every time)
A research question is viable only if it is:
1) Specific
Can you write it as one sentence?
Bad:
- “How does social media affect teenagers?”
Better:
- “How does TikTok use correlate with sleep duration among U.S. high school students?”
2) Answerable
Can you name a method that could answer it?
Examples:
- experiment / quasi-experiment
- regression / time-series
- event study
- content analysis
- structured interviews (ethics required)
If your “method” is “I’ll research and explain,” that’s usually an essay, not research.
3) Sourcable
Can you actually get the evidence within weeks?
Ask:
- What dataset will you use?
- What variables exist?
- What’s your sample size?
- Is the data public, accessible, and legal to use?
Great topics die here because the data is trapped behind paywalls, private companies, or unrealistic access.
4) Defensible
Can you explain why it matters in 2 sentences?
Not “it’s cool.”
More like:
- it affects policy decisions,
- it changes behaviour in a measurable way,
- it addresses a debated claim,
- it has real-world implications.
A simple template that forces clarity
Use this:
> “To what extent does X affect Y for Z over T?”
Where:
- X = the input/cause/variable
- Y = outcome you can measure
- Z = population/setting
- T = timeframe
If you can’t fill these in, you don’t have a research question yet.
Common traps (and how to fix them)
Trap 1: The topic is a theme, not a question
Theme: “Climate change”
Question: “How do heatwaves affect electricity demand in California counties?”
Fix: force a measurable outcome.
Trap 2: You’re trying to solve the whole world
Bad:
- “How do interest rates affect the economy?”
Better:
- “How do Fed rate hikes affect mortgage application volume in the U.S.?”
Fix: narrow the outcome and the window.
Trap 3: Your question implies impossible access
Bad:
- “How does Instagram’s algorithm affect political opinions?”
Better:
- “How does exposure to short political clips affect recall/attitudes in a small controlled sample?”
Fix: replace hidden systems with observable measures.
Three examples (with the implied method)
Example 1 — Economics
Question:
- “Do surprise inflation prints move sector ETFs differently on the day of release?”
Method:
- event study
Evidence:
- CPI release timestamps + ETF returns
Example 2 — STEM/health
Question:
- “Does caffeine intake correlate with self-reported anxiety scores in high school students?”
Method:
- survey + correlation (with careful limitations)
Evidence:
- structured survey data
Example 3 — Humanities/social science
Question:
- “How did news framing change before vs after a major policy announcement?”
Method:
- content analysis
Evidence:
- articles + coding rubric
What you should do next (30 minutes)
1) Draft 5 questions using the X→Y→Z→T template.
2) Run the stress test.
3) For the top 2 candidates, write:
- a one-paragraph method
- where the evidence comes from
- one limitation
If you can do that, you’re ready to start.
---
If you want the full structure for finishing (week-by-week), download our free 8‑Week Research Roadmap + Proposal Template.
And if you want 1:1 mentorship through the whole process, that’s what the Core Research Fellowship is built for.