brycewang-stanford/research-ideation

Research Ideation

Generate structured research questions, testable hypotheses, and empirical strategies from a topic, phenomenon, or dataset.

Input: $ARGUMENTS — a topic, phenomenon, dataset description, or extensions to brainstorm extensions of the current paper.

---

Project-Specific Context

Available Data

• EDM data (2021-2024+): Event-level sewage spill records for all overflow sites in England
• Land Registry: Universe of property transactions (prices, dates, postcodes)
• Zoopla: Rental listings with prices and characteristics
• Met Office: Daily rainfall at LSOA level
• River networks: PostGIS topology of English waterways
• Annual returns: Water company infrastructure data (sewer capacity, population served)
• Geographic boundaries: LSOAs, MSOAs, constituencies, water company areas

Existing Identification Strategies

Hedonic, repeat sales, long difference, DiD with media coverage, upstream/downstream, dry spills, hydraulic capacity IV

Related Literature Areas

Environmental disamenity capitalisation, water quality and property values, UK water industry regulation, infrastructure and house prices, information and price discovery

---

Workflow

Step 1: Understand the Input

Read $ARGUMENTS and any referenced files.

• If extensions: read current manuscript sections and analysis scripts to understand what's been done
• If a topic: ground it in the available data and methods

Step 2: Generate 3-5 Research Questions

Order from descriptive to causal:

• Descriptive: What are the patterns? (e.g. "How do spill frequencies vary across water companies?")
• Correlational: What factors are associated? (e.g. "Is sewer age correlated with spill frequency after controlling for rainfall?")
• Causal: What is the effect? (e.g. "What is the causal effect of media coverage of sewage spills on house prices?")
• Mechanism: Why does the effect exist? (e.g. "Do dry spills affect prices through stigma or physical damage?")
• Policy: What are the implications? (e.g. "Would mandated real-time spill disclosure affect property markets?")

Step 3: Develop Each Question

For each:

• Hypothesis: Testable prediction with expected sign/magnitude
• Identification strategy: How to establish causality
• Data requirements: What data is needed — is it available in the project?
• Key assumptions: What must hold
• Potential pitfalls: Threats to identification
• Related literature: 2-3 papers using similar approaches

Step 4: Rank by Feasibility and Contribution

| RQ | Feasibility | Contribution | Priority | |----|-------------|-------------|----------| | 1 | High/Med/Low | High/Med/Low | ... |

Step 5: Save Output

# Research Ideation: [Topic]
**Date:** YYYY-MM-DD

## Overview
[1-2 paragraphs situating the topic]

## Research Questions

### RQ1: [Question] (Feasibility: High/Medium/Low)
**Type:** Descriptive / Correlational / Causal / Mechanism / Policy
**Hypothesis:** [Prediction]
**Identification:** [Method + key assumption]
**Data:** [What's needed and what's available]
**Pitfalls:** [Top 2 threats]
**Related work:** [2-3 papers]

## Ranking
[Table]

## Suggested Next Steps
1. [Most promising direction]
2. [Data to obtain or analysis to run]
3. [Literature to review]

Save to output/log/research_ideation_[topic].md.

---

Principles

• Be creative but grounded. Every suggestion must be empirically feasible with available or obtainable data.
• Think like a referee. For each causal question, immediately identify the identification challenge.
• Leverage existing infrastructure. The project already has spatial matching, spill aggregation, and river network tools — build on them.
• Consider data availability. Flag when data is already available vs needs to be obtained.
• Extensions are valuable. For extensions, focus on questions that use the same data infrastructure but ask different questions.