Research Ideation

• What's surprising about current patterns?
• What contradicts conventional economic wisdom?
• Where do standard models fail?

2. The Policy Approach

• What policies lack rigorous evaluation?
• What natural experiments remain unexploited?
• What interventions might solve important problems?

3. The Data Approach

• What new data sources have become available?
• What can existing data tell us that hasn't been explored?
• What linkages between datasets are possible?

4. The Extension Approach

• How can seminal papers be extended?
• What mechanisms remain unexplored?
• Can methods from one field apply to another?

Step 3: Evaluate and Refine

For each idea, assess:

• Feasibility: Can this be done with available data/methods?
• Contribution: What's new here?
• Interest: Who cares about this question?
• Identification: Can effects be credibly estimated?

Example Interaction

User Query: "I'm interested in labor economics and have access to LinkedIn data"

Generated Response:

Research Ideas: LinkedIn Data × Labor Economics

1. Network Effects in Job Search

Question: How do professional networks affect job transitions and wage changes?

Puzzle: Standard search models assume random matching, but most jobs come through networks. How much do networks matter for outcomes?

Approach:

• Track profile updates signaling job changes
• Measure network overlap between current/future employers
• Instrument with exogenous network shocks (school alumni connections)

Contribution: Quantify the causal role of networks vs. unobserved ability

2. Skill Signaling and Returns

Question: Do workers who list specific skills on profiles earn wage premiums?

Puzzle: Are listed skills signals of ability, or just cheap talk? What's the return to skill acquisition vs. skill signaling?

Approach:

• Compare wages before/after workers add skills
• Use skill endorsements as validation
• Exploit skill assessment features as quasi-experiments

3. Gender Gaps in Self-Promotion

Question: Do men and women describe equivalent achievements differently?

Puzzle: Lab evidence shows women understate accomplishments. Does this appear in real profiles and affect outcomes?

Approach:

• NLP analysis of profile language (adjectives, quantification)
• Compare language for workers in same role/company
• Link to recruiter contact rates if available

4. Remote Work and Labor Market Geography

Question: How has remote work changed the geographic reach of job matching?

Approach:

• Compare pre/post-2020 job applications by distance
• Measure changes in location signals on profiles
• Study wage convergence across cities

Evaluation Matrix

Recommended Next Steps

1. For empirical work: Start with Idea #4 (clearest identification)
2. For policy impact: Idea #1 has most implications for job matching platforms
3. For methodological novelty: Idea #3 combines NLP with quasi-experimental methods

Frameworks for Generating Questions

The "5 Whys" for Economics

Start with an observation and drill down:

1. GDP growth is slowing → Why?
2. Productivity is stagnant → Why?
3. Investment is low → Why?
4. Uncertainty is high → Why?
5. Policy is unpredictable → Testable: Does policy uncertainty cause low investment?

The "What If" Generator

• What if [policy X] were implemented?
• What if [technology Y] became widespread?
• What if [assumption Z] were relaxed?

The Cross-Field Pollinator

Take a method from one field and apply to another:

• IO techniques → Labor markets (how do firms choose wages like prices?)
• Finance models → Education (returns to schooling as asset pricing)
• Macro shocks → Micro outcomes (firm-level effects of exchange rate changes)

Common Pitfalls

• ❌ Questions that are too broad ("What causes inequality?")
• ❌ Questions without clean identification ("Does education cause income?")
• ❌ Questions without data ("Were medieval peasants happy?")
• ❌ Questions already well-answered

References

• Shapiro (2022) How to Get Started on Research in Economics
• Angrist & Pischke on Mostly Harmless research design
• AEA Research Pipelines

Changelog

v1.0.0

• Initial release with ideation frameworks