# Construction Cost Leading-Indicator Model — Data Inventory
**rateauthority.org | PolicyChat**
*Generated: 2026-05-22*

---

## 1. FRED / BLS Series — Available via API Key

| Friendly Name | FRED Series ID | Frequency | Coverage | Public? | Notes |
|---|---|---|---|---|---|
| PPI Lumber & Wood Products | WPU081 | Monthly | 1947–present | Yes (free) | Primary lumber input signal |
| PPI Copper & Copper Alloys | WPU102501 | Monthly | 1947–present | Yes (free) | Wiring, plumbing cost driver |
| PPI Gypsum Products | WPU1321 | Monthly | 1947–present | Yes (free) | Drywall / wallboard |
| PPI Asphalt & Roofing Materials | WPU0591 | Monthly | 1947–present | Yes (free) | Shingles, roofing |
| PPI Construction Materials (composite) | WPUIP2311001 | Monthly | 1985–present | Yes (free) | Aggregate input-price index |
| PPI Inputs to Residential Construction | PPIIDC | Quarterly | 2005–present | Yes (free) | Most direct construction-cost proxy |
| Employment Cost Index — Construction wages | CIU1010000120000A | Quarterly | 1980–present | Yes (free) | Labor cost component ~40% of rebuild |
| FHFA Purchase-Only HPI (national) | HPIPONM226S | Monthly | 1991–present | Yes (free) | Closest free proxy for dwelling value trends |
| Median Sales Price of Houses Sold | MSPUS | Quarterly | 1963–present | Yes (free) | Dwelling-value anchor |

**Access:** All via `https://api.stlouisfed.org/fred/series/observations?series_id=<ID>&api_key=<KEY>`
**Key:** loaded from `FRED_API_KEY` environment variable
**Fetch script:** `build_model.py` in this directory

---

## 2. Lumber / Copper Futures — Yahoo Finance

| Instrument | Yahoo Symbol | Exchange | Coverage | Notes |
|---|---|---|---|---|
| CME Lumber Futures (front month) | LBS=F | CME | ~2000–present | Thin history before 2015 on YF; use FRED WPU081 as backstop |
| CME Copper Futures (front month) | HG=F | COMEX | 1988–present | Good liquidity; monthly close available |

**Access:** `yfinance` Python package (`pip install yfinance`); no API key required.
**Limitation:** Yahoo Finance futures history can be truncated for older contracts; WPU102501 (FRED) is the more reliable long-run copper price series.

---

## 3. Marshall-Swift / CoreLogic Reconstruction Cost Estimator Index

| Aspect | Detail |
|---|---|
| Publisher | CoreLogic (acquired Marshall-Swift-Boeckh) |
| Frequency | Quarterly |
| Geography | National + regional (9 census divisions) |
| Public availability | **Licensed** — full quarterly index requires CoreLogic data subscription (~$5k–$20k/yr for research license) |
| Free summary | Published annually in the III Insurance Information Institute *Insurance Fact Book* (aggregate year-over-year change only) |
| NAIC reference | NAIC *Homeowners Insurance Report* cites MSB/CoreLogic for dwelling replacement-cost trends |
| Closest free proxy | FRED `HPIPONM226S` (FHFA HPI) or `MSPUS` (median home price) — correlated but not identical to rebuild cost |
| Published excerpts | III *Fact Book* 2023–2024 reports rebuild-cost inflation of +11.4% (2021), +12.1% (2022), +7.3% (2023), +3.8% (2024e) |
| Academic proxy | Sirmans & Macpherson (2003) use PPI inputs composite as MSB proxy; correlation r ≈ 0.91 in-sample |

**Recommended path for live model:** License CoreLogic quarterly index OR use III annual excerpts as calibration checkpoints against the PPI composite. The PPI Inputs to Residential Construction (PPIIDC) is the best single free substitute.

---

## 4. NAIC Home Insurance Premiums — Local DB

| Aspect | Detail |
|---|---|
| Source | NAIC *State Average Expenditures & Premiums* 2023 report |
| Location | Rate Authority rate ledger (SQLite) → `rate_records` table |
| Coverage | 51 state/DC cells, product=`home`, year=2023 |
| Granularity | Annual average monthly premium (×12 = annual) |
| Multi-year panel | **Not in DB** — only 2023 cross-section. NAIC publishes historical tables (2013–2023) in PDF format; structured ingest would require `ingest_naic.py` extension |
| Historical NAIC | Available in NAIC Homeowners Annual Data publications (request or scrape); III Fact Book carries 10-year time series |

---

## 5. Coverage Gaps and Mitigations

| Gap | Impact | Mitigation |
|---|---|---|
| Marshall-Swift quarterly index (unlicensed) | Cannot directly validate model against official rebuild-cost index | Use PPIIDC as primary target; cite III annual excerpts as calibration |
| NAIC multi-year panel (single year = n=51) | Cannot estimate state-year panel model; lag distribution is underpowered | Use national time series (FRED, 2005–2024, n=20 annual observations) for lag estimation; cross-section for level-correlation |
| State-level construction cost variation | National PPI series do not capture regional cost differentials (e.g., CA vs MS have 2–3× labor cost differences) | RSMeans regional cost modifiers (published annually, some free) could be integrated as state multipliers |
| Lumber futures thin history (pre-2015) | Short usable window for futures-vs-PPI lead analysis | WPU081 monthly back to 1947 is the operational series; futures used only for real-time forward-looking component |

---

## 6. Recommended Data Acquisition Roadmap

1. **Immediate (free):** Ingest all FRED series listed above via `build_model.py`. Run quarterly.
2. **Short-term (~30 days):** Extend `ingest_naic.py` to load NAIC 2013–2022 homeowners data from NAIC annual reports (PDFs available free). Adds ~450 state-year cells for panel regression.
3. **Medium-term (~90 days):** License CoreLogic quarterly MSB index (or obtain through insurance-industry partner) for direct model validation.
4. **Optional:** Integrate RSMeans city cost indices (annual, licensed) for state-level construction cost modifiers.