Markdown version of the slide deck "Choosing and Crafting a Research Question." NBER Economics of Health Training Workshop, 2026. Sebastian Tello-Trillo, University of Virginia. Slide text is verbatim from the deck, including original wording. Speaker notes appear as quotes under each slide; screenshots are described in brackets. Where a slide has no title, a descriptive heading was added and marked.

• Slide 1 — Welcome
• Slide 2 — Everything has already been solved. Right?
• Slide 3 — Lay out of talk
• Part 1 — How to come up with research ideas (Slides 4–18)
• Slide 4 — Five doors to a research idea
• Slide 5 — The world
• Slide 6 — The literature door has three branches
• Slide 7 — Example: a review paper (JEP) (slide has no title; heading added)
• Slide 8 — Example: the review paper's abstract (slide has no title; heading added)
• Slide 9 — Open questions generated from the review paper (slide has no title; heading added)
• Slide 10 — Data access
• Slide 11 — Nonprofit reports: a menu of policies
• Slide 12 — Example: a nonprofit report (slide has no title; heading added)
• Slide 13 — Demo output: policies, datasets and empirical strategies (slide has no title; heading added)
• Slide 14 — Policy
• Slide 15 — Other anecdotes from friends
• Slide 16 — Other anecdotes from friends:
• Slide 17 — Other anecdotes from friends:
• Slide 18 — Other anecdotes from friends:
• Part 2 — How to turn research ideas into research questions (Slides 19–24)
• Slide 19 — Lay out of talk (roadmap)
• Slide 20 — A "research" idea is not yet a research question
• Slide 21 — Narrowing an idea into a research question
• Slide 22 — An idea is not yet a question
• Slide 23 — Descriptive, predictive, or causal?
• Slide 24 — Exercise: sort these ideas
• Part 3–4 — Evaluating contribution and tradeoffs (Slides 25–28)
• Slide 25 — Lay out of talk (roadmap)
• Slide 26 — Big question or small question?
• Slide 27 — Worksheet: from idea to question
• Slide 28 — Pair share: attack the question
• Slide 29 — Four takeaways

Slide 1 — Welcome

Choosing and Crafting a Research Question

NBER Economics of Health Training Workshop, 2026 Sebastian Tello-Trillo, University of Virginia

Notes: Welcome. Goal for the hour: leave with one research question you wrote yourself, classified as descriptive, predictive or causal, and stress-tested. About 25 minutes of talk and demo, the rest is hands-on.

Slide 2 — Everything has already been solved. Right?

In graduate school I thought so.

It was partly true and partly false. There is still a great deal we do not know, and questions come in every size.

The hard part is not finding questions. It is telling which ones are good.

Quick poll: raise your hand

1. I have a question I am working on
2. I have a topic, not yet a question
3. I have neither

Notes: Personal story: in grad school I thought everything had been solved. That is partly true, partly false. Run the poll by show of hands; it tells you how much of the room needs the idea-generation part versus the refinement part.

Slide 3 — Lay out of talk

1. How to come up with research ideas?
2. How to turn research ideas into research questions?
3. How to evaluate the contribution of your RQ?
4. Thinking about tradeoffs of what to pursue?
1. Data/co-author/difficulty/waiting/
2. Have many stoves
3. Returns to experience
5. How to pitch your idea

Quick poll: raise your hand

1. I have a question I am working on
2. I have a topic, not yet a question
3. I have neither

Part 1 — How to come up with research ideas (Slides 4–18)

Slide 4 — Five doors to a research idea

[Five icons, one per door.]

These give you ideas, not research questions. Every idea still has to be worked through.

Notes: Keep 'a policy you find interesting' as its own door. The fifth door (your own data or institutional knowledge) is my addition; it is often the most defensible source of a question because nobody else can ask it. Emphasize the caution at the bottom.

Slide 5 — The world

[Image: a stack of newspapers.]

• Social scientist – to an extent- are trying to understand and explain the world they live in
• How to know the world and why is working/not working?
• Newspapers
• Books
• Podcasts

Slide 6 — The literature door has three branches

Habit to build:

• after reading any paper, write down the one question it raises but does not answer.
• Print the papers
• Check Journal of Economic Perspective and Journal of Economic Literature

Notes: The three branches from my notes: a review paper, recent literature, and the broader journal literature. The point about journals: a lot of strong ideas were published and then simply stopped being read. That is ripe for mining.

Slide 7 — Example: a review paper (JEP) (slide has no title; heading added)

[Screenshot of the Journal of Economic Perspectives page for "America's Continuing Struggle with Mental Illnesses: Economic Considerations", Richard G. Frank and Sherry A. Glied, JEP Vol. 37, No. 2, Spring 2023 (pp. 153–78).]

Slide 8 — Example: the review paper's abstract (slide has no title; heading added)

[Screenshot of the abstract and citation.]

Mental illnesses affect roughly 20 percent of the US population. Like other health conditions, mental illnesses impose costs on individuals; they also generate costs that extend to family members and the larger society. Care for mental illnesses has evolved quite differently from the rest of health care sector. While medical care in general has seen major advances in the technology of treatment this has not been the case to the same extent for mental illnesses. Relative to other health care, the cost of care for mental illnesses has grown more slowly and the social cost of illness has grown more rapidly. In this essay we offer evidence about the forces underpinning these patterns and emphasize the challenges stemming from the heterogeneity of mental illnesses. We examine institutions and rationing mechanisms that affect the ability to make appropriate matches between clinical problems and treatments. We conclude with a review of implications for policy and economic research.

Citation: Frank, Richard G., and Sherry A. Glied. 2023. "America's Continuing Struggle with Mental Illnesses: Economic Considerations." Journal of Economic Perspectives 37 (2): 153–78. DOI: 10.1257/jep.37.2.153

Slide 9 — Open questions generated from the review paper (slide has no title; heading added)

[Three boxes.]

1. Dynamic economics of mental illness — how do mental-health shocks propagate through education to employment to earnings to family to health over decades, and when can intervention break that propagation?
2. Optimal allocation across medical and nonmedical interventions — should the marginal dollar go to therapy, drugs, housing, income, employment support, or case management, and how complementary are these inputs?
3. Cross-system returns to mental-health policy — what happens to total social welfare and government spending when an intervention affects Medicaid + SSI/SSDI + housing + criminal justice + earnings simultaneously?

Slide 10 — Data access

Data access

• Phonics paper: "I had access to amazing data with decet DUA. I listened to a podcast and was like "whoa someone should study that formula. Wait?! That could be me!""
• "I had access to this interesting dataset, and kept reading on papers until I realize there was one idea that I could do with these data"

Slide 11 — Nonprofit reports: a menu of policies

1. Read the report: progress claimed, ideas proposed
2. List the policies that could be evaluated
3. Ask: treatment, comparison group, outcome
4. Ask: where can I get the data?

There is no shortage of places to get ideas. The work is turning an idea into something you can measure.

Notes: If you are interested in policy, nonprofits are a goldmine: they publish what they think has worked and what they want to try. Read that as a list of candidate evaluations. Then immediately ask where the data would come from. This sets up the demo.

Slide 12 — Example: a nonprofit report (slide has no title; heading added)

[Screenshot of the Inseparable website: "138 Policy Wins For Mental Health."]

Notes: https://www.inseparable.us/wins/

Slide 13 — Demo output: policies, datasets and empirical strategies (slide has no title; heading added)

[Three screenshots of a generated report that lists candidate policies and, for each, the policy details, available datasets and empirical strategies. Visible policy headings: "1. School-Based Mental Health Staffing & Coordinators," "3. Annual Mental Health Wellness Checks / Screenings," and "4. Mental Health Parity Enforcement & Commercial Insurance Reform." Each has sections for Policy, Datasets, and Empirical strategies, with a feasibility score.]

Slide 14 — Policy

• What is the policy about?
• Do we already know if this policy works or not?
• What does it mean that the policy "works"? (Is this measurable? If so, how)
• What are the alternatives?
• If we were to know the effects of this policy, how does this advance economic knowledge?
• What's the thing that we learned that we otherwise didn't know?

Slide 15 — Other anecdotes from friends

[Image: piles of paperwork.]

• Stacked DID paper: Basically, he lunches with a co-author. He says yes with me, pushing occasionally back.
• I went to a NASCAR race and was like, "This seems like bad air pollution." I texted my buddy about it. Lots of papers done from that for us
• Was always complaining about people not realizing air pollution moves across counties. Wrote a paper where we modeled that.
• When a grad student stumbles on some data, they utilize it without knowing whether it was useful or not. Later on, they get a decent idea of what to do with it, but slow burn.

Slide 16 — Other anecdotes from friends:

• "For example in my newest project it came about because i noticed something weird in a dataset i had for a different project."
• "For another, it's a project that forever i kept saying "someone is going to write this paper. someone is going to write this paper." and no one was doing it."
• "Most of my better ideas come from seeing a conference presentation and getting inspired by a dataset or a methodology i hadn't thought of before."

Slide 17 — Other anecdotes from friends:

• "My hep c project was pitched to me by a coauthor, who heard about the program on npr I think"
• "My parents losing job project is a follow up to a project about postpartum job loss, which was just based on the idea that losing a job while you have a new baby seems very hard to recover from"
• "My gpa project was based on my coauthor and I talking about the fact that Norway has these standardized random high-stakes exams that count as much as a course grade, and how sometimes people get assigned topics they are good in and sometimes topics they are less good in"

Slide 18 — Other anecdotes from friends:

• "I am applying for data to try to do something about the Black Panther Party having free breakfast program, which I read about or saw somewhere (I dont remember how I originally learned about it) but it seemed like there was no data way to do it. but then the hep c coauthor found the black Panther news paper which announced when locations opened, and knew census was link old and new data"

Part 2 — How to turn research ideas into research questions (Slides 19–24)

Slide 19 — Lay out of talk (roadmap)

1. How to come up with research ideas?
2. How to turn research ideas into research questions?
3. How to evaluate the contribution of your RQ?
4. Thinking about tradeoffs of what to pursue?
• 4.1 Data/co-author/difficulty/waiting/
• 4.2 Have many stoves
• 4.3 Returns to experience
5. How to pitch your idea

Slide 20 — A "research" idea is not yet a research question

For example:

• "I would like to explore how if Medicaid affects people's health" (Idea)
• "I would like to know why there is disparities in take-up between black and whites individuals in Medicaid"
• "I would like to know if there is a way to know which are the rural hospitals that are more likely to close so we can help them"

Slide 21 — Narrowing an idea into a research question

In order to turn this ideas into research questions we need to specify things, here are some questions that help us narrow down to a research question:

1. Identify if the nature is: causal, descriptive, or predictive in nature
2. For who?
3. When?
4. How to measure the main outcome?
5. What is it about the policy that you care about?

Let's apply some of these questions to these statements:

• "I would like to explore how if Medicaid affects people's health" (Idea)
• "I would like to know why there is disparities in take-up between black and whites individuals in Medicaid"
• "I would like to know if there is a way to know which are the rural hospitals that are more likely to close so we can help them"

Slide 22 — An idea is not yet a question

A question names a population, a factor or treatment, an outcome, and a comparison.

Notes: These examples are illustrative. The same topic (Medicaid, hospitals, disparities) can become three different kinds of question depending on how it is worded. Swap in examples from your own work if you prefer.

Slide 23 — Descriptive, predictive, or causal?

A descriptive project can contain causal exercises. Just know which one it is in spirit, and say so.

Notes: My own example: a project on racial disparities in medication use. However you slice it, in spirit it is descriptive: documenting trends and what explains them. You can bolt causal exercises onto it and that is fine on the page. Just be aware. The classification determines what evidence convinces a reader and what you can claim.

Slide 24 — Exercise: sort these ideas

In pairs, 8 minutes. Label each D, P or C and justify it in one sentence.

1. Racial disparities in the prescribing of a given medication
2. The effect of Medicaid expansion on mortality
3. Which patients are most likely to be readmitted within 30 days
4. The growth of private equity ownership of physician practices
5. Whether a crisis hotline launch changes suicide deaths
6. Why prices for the same procedure differ across hospitals
7. Which rural hospitals are likely to close in the next five years
8. The rise in telehealth use after 2020

Notes: The handout has the same eight cards with a D/P/C column. Teaching point: the wording of the question, not the topic, determines the type. (The facilitator answer key from the speaker notes is omitted from this page.)

Part 3–4 — Evaluating contribution and tradeoffs (Slides 25–28)

Slide 25 — Lay out of talk (roadmap)

1. How to come up with research ideas?
2. How to turn research ideas into research questions?
3. How to evaluate the contribution of your RQ?
4. Thinking about tradeoffs of what to pursue?
• 4.1 Data/co-author/difficulty/waiting/
• 4.2 Have many stoves
• 4.3 Returns to experience
5. How to pitch your idea

Slide 26 — Big question or small question?

You may not have a good thermometer yet.

• Sometimes a question is simple, important, and simply unnoticed. Sometimes the reverse.
• Train yourself by doing research. That experience is how you learn what matters.
• Talking to people helps. Presenting helps.
• Example: TennCare Disenrollment

Until then, three filters:

1. So what? If the answer is X versus Y, who decides something differently?
2. Known already? Can you name the two closest papers and what they do not answer?
3. Feasible? Is there data and variation to answer it within a year?

Notes: Do not pretend there is a formula for big versus small. The honest advice is that judgment comes from doing research. In the meantime, give them the three-part filter as a scaffold; the worksheet uses it in Part C.

Slide 27 — Worksheet: from idea to question

12 minutes, on your own

Notes: Hand out the one-page worksheet (the full text is on the Notion page). Ask them to pick the idea they care most about, even if it is messy. Walk around during the 12 minutes. If someone has no idea at all, point them to one of the five doors and tell them to start with the demo output.

Slide 28 — Pair share: attack the question

5 minutes. One minute to read your question aloud, then your partner asks only two things:

1. "Is that descriptive, predictive, or causal?"
2. "So what?"

Then swap. Revise your question with whatever you heard.

Notes: Limit the partner to those two prompts. Otherwise the conversation turns into advice about methods before the question is even clear. Whoever is listening should not offer solutions yet.

Slide 29 — Four takeaways

1. Ideas are cheap and everywhere. Questions take work.
2. Know which type your question is: descriptive, predictive or causal.
3. Think about level of contribution
4. Test the question before you fall in love with it.

Later today

• 11:40 Finding and acquiring data. Bring your question.
• 1:30 Econometrics workshop: designs for causal questions.
• 4:10 Office hours: refine your question.

Notes: Close by bridging forward: the next session is about data, so their worksheet question is the input. The afternoon workshop covers the designs that answer causal questions, and office hours are for refining what they wrote today.