Build Trade Model

---
name: build-trade-model
description: "Builds a trade model: gravity and structural gravity estimation, quantitative trade general equilibrium, and global value chain accounting."
argument-hint: "[model name]"
user-invocable: true
allowed-tools: Read, Glob, Grep, Write, Edit, Bash, AskUserQuestion, Task
---

# /build-trade-model — Build Trade Model

**Category:** Build · **Primary agent:** `trade-policy-simulation-specialist`

Builds a trade model: gravity and structural gravity estimation, quantitative trade general equilibrium, and global value chain accounting.

Arguments: `$ARGUMENTS`

## Purpose
Build a trade model fit for the question: reduced-form or structural gravity estimation of trade costs and policy effects, a quantitative general-equilibrium trade model solved in changes, or value-chain accounting. The output reproduces baseline trade flows, makes the trade-balance treatment explicit, and reports welfare and sectoral effects across the plausible range of trade elasticities.

**Use it when** the question involves tariffs, trade agreements, trade costs, retaliation, value chains or trade exposure across partners and sectors.

**Do not use it when** domestic tax and factor-market detail dominates (`/build-cge`) or the question is a supply disruption through production networks (`/supply-chain-shock`).

## Before you start
1. **Check required inputs.** For each input below record `FOUND` or `ABSENT` (never 'assumed'):
   - Bilateral trade data
   - Tariff and trade-cost data
   - Global input-output tables (optional)
   If a required input is `ABSENT`, that part is **`NOT ASSESSED — NO DATA`**: do not estimate or invent it; name what is missing and which skill produces it. If all required inputs are absent, stop and say so.
2. **Check confidentiality.** Identify the data classification level of the inputs (`.claude/docs/confidentiality-and-ai-use.md`). Do not proceed with restricted or confidential data in a way the policy does not allow.
3. **Collaboration.** Follow `.claude/docs/collaboration-protocol.md`: ask questions first, present options, get approval before writing files, and update `production/session-state/active.md` after each approved section.

## Agents
- `trade-policy-simulation-specialist` — primary agent; coordinates the work and integrates results
- `trade-gvc-specialist` — Models trade and global value chains
- `production-networks-specialist` — Models production networks and supply chains
- `cge-modeling-specialist` — Builds computable general equilibrium models

Run these agents as subagents (the `Task` tool), in parallel where the steps are independent. Subagents only write to the paths named in your instructions.

## Key design decisions (present options, let the user choose)
| Decision | Typical options | What it changes |
|---|---|---|
| Model type | reduced-form gravity, structural gravity, quantitative GE (Armington, Eaton–Kortum), CGE | Counterfactual capability |
| Sectors | single, multi-sector, with input-output links | Amplification through intermediates |
| Trade balance | balanced trade, fixed deficits | Counterfactual results |
| Elasticities | estimated, sourced from literature, ranges | Welfare magnitudes |
| Data | product-level trade, global IO tables | Coverage and detail |

## Workflow
1. Define the question and choose between reduced-form gravity, structural gravity, quantitative GE trade or CGE
2. Estimate trade elasticities and trade costs with PPML and high-dimensional fixed effects
   - Exporter–year, importer–year and pair fixed effects; treatment of zeros
3. Calibrate the GE model and solve counterfactuals in changes
   - Use observed trade shares; state the solution algorithm and convergence
4. Check the model reproduces baseline trade flows and the trade balance treatment is explicit
5. Report welfare and sectoral effects with elasticity sensitivity
   - Decompose welfare; compare results with and without input-output links
6. Register the model as a lego set (`/lego-map`): a baseplate plus bricks with declared studs (variables, units, timing) and stubs, so components can later be attached, detached or swapped

## Checkpoints with the user
- After choosing the model type: confirm sectors and partners
- After estimation: confirm elasticities and ranges
- Before reporting: confirm the trade-balance treatment

## Common pitfalls
- OLS gravity on logs with zeros dropped
- Single-sector models for policies hitting intermediates
- Trade-flow changes reported as welfare
- Elasticities from different aggregation levels

## Output
`models/trade/<name>/ (+ model.lego.json)`

The folder holds data, estimation scripts and results, the GE model and counterfactual scripts, the elasticity table with sources, results with welfare decomposition and the lego manifest. Every run saves its results to `output/<run-id>/` in the model folder, with a run log (`.claude/rules/model-code.md`).

## Gate / verdict
- Baseline trade flows reproduced
- Elasticity sources and ranges documented
- Trade balance closure stated
- Lego manifest passes `python tools/lego.py check`
- Gravity estimates use PPML with appropriate fixed effects
- Welfare is decomposed into its sources

## Suggested next steps
- `/trade-policy-analysis`
- `/validate-model`
- `/lego-map`