Finance Modeling Research

When: User introduces a new finance or economics concept to study

Steps:

1. Identify topic category: Classify the topic (e.g., monetary-policy, commodities, credit-derivatives)
2. Create folder structure: Auto-create study_notes/[topic-category]/
3. Initialize framework files:
   • 01-[topic]-fundamentals.md - Core concepts and theory
   • 02-[topic]-market-structure.md - How the market works (if applicable)
   • 03-[topic]-modeling-approaches.md - Analytical methods
4. Update INDEX.md: Add new topic to _core/INDEX.md
5. Fill in progressively: Add content as discussion develops

Example:

User: "I want to study yield curve dynamics"

Actions:
1. Create: study_notes/interest-rates/yield-curves/
2. Create baseline files with framework templates
3. Update _core/INDEX.md with new topic entry
4. Begin documenting term structure theories as discussed

Workflow 2: Identify Modeling Approach

When: User needs to analyze data or build a model for a finance problem

Steps:

1. Understand the question: What economic relationship needs to be analyzed?
2. Map causal relationships: Document cause-and-effect chains
3. Identify model types: Match problem to appropriate methodology
   • Econometric: Time series (ARIMA, VAR), regression (OLS, panel data)
   • Statistical: Factor models, state-space, regime-switching
   • Machine Learning: Classification, regression, clustering
   • Financial: Asset pricing, option pricing, risk models
4. Document model specifications:
   • Purpose and assumptions
   • Data requirements (frequency, variables, time range)
   • Expected interpretation
   • Known limitations
5. Capture in study notes: Save to 03-[topic]-modeling-approaches.md

Example:

User: "How do we model Fed rate impact on stocks?"

Document:
- Economic theory: Higher rates → higher discount rate → lower valuations
- Model type: Vector Autoregression (VAR)
- Variables: Fed Funds Rate, S&P 500, GDP Growth, CPI
- Specification: 4-variable VAR with 4 quarterly lags
- Data needs: Quarterly data, 25+ years, stationary series
- Interpretation: Impulse response functions show dynamic effects

Workflow 3: Specify Data Requirements

When: Modeling approach identified, need to define data inputs

Steps:

1. Document WHAT data is needed:

   • Variable names (e.g., "Fed Funds Target Rate", "S&P 500 Index")
   • Frequency (daily, monthly, quarterly, annual)
   • Time range (start date, end date, history needed)
   • Geography (US, EU, global, country-specific)
   • Format (levels, changes, growth rates, indices)

2. Explain WHY this data is needed:

   • Theoretical justification (how does it fit the model?)
   • Variable role (dependent, independent, control)
   • Expected relationship (positive, negative, non-linear)

3. Outline HOW to obtain data:

   • Potential sources (FRED, World Bank, Bloomberg, vendor APIs)
   • Mark for technical-research skill handoff

4. Save specification: Add to modeling approach document

Example:

WHAT:
- Fed Funds Target Rate (monthly, 1990-present)
- S&P 500 Index (monthly, 1990-present)
- GDP Growth (quarterly, interpolate to monthly)
- CPI Inflation (monthly)

WHY:
- Fed Funds: Independent variable (policy tool)
- S&P 500: Dependent variable (market response)
- GDP: Control variable (economic activity)
- CPI: Control variable (inflation environment)

HOW:
- FRED API likely has all series
→ technical-research skill: Verify FRED API coverage and document endpoints

Workflow 4: Maintain Study Note Organization

When: Continuously throughout research sessions

Folder Structure:

study_notes/
├── _core/
│   ├── 00-learning-roadmap.md
│   ├── 01-macroeconomic-fundamentals.md
│   ├── INDEX.md
│   └── QUICK-REFERENCE.md
│
├── [topic-category]/
│   ├── 01-[topic]-overview.md
│   ├── 02-[topic]-advanced.md
│   └── 03-[topic]-modeling.md

Maintenance Tasks:

1. Create topic folders when new subjects arise
2. Update INDEX.md to maintain navigation
3. Fill framework files progressively during conversations
4. Cross-reference related topics for integrated understanding
5. Keep notes concise - detailed content goes in reference files

Workflow 5: Prepare Handoff Summaries

When: Research phase complete, ready for next skill (requirements-gathering or technical-research)

For requirements-gathering skill:

## Study Summary: [Topic]

### Concepts Studied:
- [Key concept 1]
- [Key concept 2]
- [Key concept 3]

### Modeling Approach:
- [Model type and purpose]
- [Key assumptions]
- [Expected outputs]

### Data Requirements:
- [High-level data needs]

### Next Steps for requirements-gathering:
1. [Clarification question 1]
2. [Scoping consideration 2]
3. [User requirement to confirm]

For technical-research skill:

## Data Research Request: [Topic]

### Required Data:
1. [Variable 1] - [frequency], [time range]
2. [Variable 2] - [frequency], [time range]

### Potential Sources:
- [Source 1]: [What it might have]
- [Source 2]: [What it might have]

### Technical Research Tasks:
- Research [API/service] for [data type]
- Document data availability and quality
- Capture sample API responses
- Check free vs. paid access

Study Note Best Practices

Start with Theory First

Always establish economic foundation before jumping to models:

## Economic Theory

[Explain the underlying economic mechanism]

### Key Principles:
1. [Principle 1]: [Explanation]
2. [Principle 2]: [Explanation]

### Why This Matters:
- [Practical implication 1]
- [Practical implication 2]

NOW we can discuss modeling approaches...

Map Relationships Clearly

Use causal chains to show interconnections:

Fed raises rates → 
  Bond prices fall (duration effect) → 
    Stock valuations decline (higher discount rate) →
      Dollar strengthens (capital inflows) →
        Commodity prices fall (dollar-denominated)

Be Specific About Models

Don't just name models - specify them completely:

## Proposed Model: [Model Name]

**Purpose**: [What question does it answer?]

**Variables**: [List all variables with roles]

**Specification**:
- [Technical detail 1]
- [Technical detail 2]
- [Parameter choices with justification]

**Why This Model**:
- [Advantage 1]
- [Advantage 2]

**Data Requirements**:
- [Specific frequency and time range]
- [Preprocessing needs]

**Interpretation**:
- [How to read results]
- [What outputs mean]

**Limitations**:
- [Assumption 1 and when it might break]
- [Known weakness and mitigation]

Think About Implementation

Always consider feasibility:

• Is the model achievable with available data?
• Are assumptions realistic for this application?
• Can results be interpreted clearly by users?
• What are the practical limitations?

Skill Integration

This skill focuses on theory, concepts, and model design. It hands off implementation details to other skills:

finance-modeling-research: "Need Fed Funds Rate, monthly, 1990-present"
    ↓
technical-research: "FRED series FEDFUNDS, check API limits, document code"
    ↓
requirements-gathering: "Confirm user needs this timeframe, finalize scope"

Examples

Example 1: Central Bank Communication Analysis

User: "I want to study how central bank communications affect markets"

Process:

1. Create folder: study_notes/monetary-policy/central-bank-communications/

2. Document theory in 01-communications-framework.md:

## Policy Transmission Mechanisms:
- Interest rate channel (traditional tool)
- Expectations channel (forward guidance)
- Confidence channel (credibility effects)

## Communication Types:
- Policy statements (FOMC)
- Press conferences (Q&A)
- Speeches (by officials)
- Meeting minutes (deliberations)

## Market Response Mechanisms:
- Surprise vs. expected components
- Hawkish vs. dovish interpretation
- Credibility and consistency effects

3. Identify modeling approaches in 03-communications-modeling.md:

## Approach 1: Event Study
- Measure abnormal returns around announcements
- Isolate surprise component (Fed Funds futures)
- Window: [-1 day, +1 day] around announcement

## Approach 2: Text Analysis
- NLP on policy statements
- Sentiment scoring (hawkish/dovish scale)
- Topic modeling (which issues emphasized)

4. Specify data requirements:

WHAT:
- Fed policy statements (text, 2000-present)
- Fed Funds target rate changes
- Fed Funds futures (for expectations)
- Asset prices (S&P 500, 10Y yield, DXY, VIX)

WHY:
- Statements: Measure communication tone
- Futures: Separate surprise from expected
- Asset prices: Measure market response

HOW:
- Fed website (may need scraping)
→ technical-research skill: Research Fed statement access
→ technical-research skill: Check CME futures data APIs

5. Update INDEX.md and prepare handoff

Example 2: Commodity Price Modeling (Iron Ore)

User: "How do we model iron ore prices?"

Process:

1. Create folder: study_notes/commodities/iron-ore/

2. Document market structure:

## Supply Side:
- Oligopolistic (BHP, Rio Tinto, FMG, Vale)
- High barriers to entry (capital intensive)
- Quality differences (Fe content)

## Demand Side:
- China dominates (70%+ of seaborne trade)
- Steel production driver
- Infrastructure and construction cycles

## Key Price Drivers:
- Chinese steel production (demand)
- Vale production disruptions (supply shocks)
- Shipping costs (freight rates)
- Port inventory levels (supply indicator)

3. Identify models:

## Model 1: Supply-Demand Balance
Iron Ore Price = f(Steel Production, Inventories, Supply Shocks, Freight)

- Method: OLS regression with lagged variables
- Y: Iron ore spot price (62% Fe CFR China)
- X: Steel production (+), inventories (-), supply shocks (+), freight (+)

## Model 2: Steel Spread
Iron Ore Price = f(Steel Price, Steel Margins, Coking Coal)

- Method: Cointegration analysis
- Logic: Steelmakers pay for ore based on profitability

4. Specify data:

WHAT:
- Iron ore prices: Spot (Platts 62% Fe), daily, 10 years
- China steel production: Monthly
- Port inventories: Weekly
- Freight rates: Capesize daily
- Steel prices: Rebar and HRC in China

WHY:
- Prices: Dependent variable
- Production: Demand indicator
- Inventories: Supply-demand balance
- Freight: Cost component

HOW:
→ technical-research skill: Research commodity data APIs (Quandl, Bloomberg)
→ technical-research skill: Check China data availability (Wind, Reuters)
→ technical-research skill: Investigate Baltic Exchange freight data

5. Prepare handoff summary

Advanced Features

Model Type Reference

See references/modeling-frameworks.md for:

• Detailed specifications for common model types
• When to use each model class
• Data requirements for each methodology
• Interpretation guidelines
• Common pitfalls and solutions

Economic Theory Primer

See references/economic-foundations.md for:

• Core economic principles relevant to modeling
• Common economic relationships and mechanisms
• Glossary of key terms
• Reading recommendations

Activity Logging

Log research activities via the agent-logging skill in agent_log/activity/YYYYMMDD.md.

Templates

• topic-framework-template.md — new topic folders
• modeling-spec-template.md — model approach docs
• data-requirements-template.md — data need specs