Materials Monitor: methodology
This page is public. It names every source, states every rule, and explains what the record can and cannot show.
What the publication reports
Market evidence about construction-material prices in the United States, from a small set of public sources, arranged so that a reader can see which material groups rose, which fell, over what period, and what the available evidence can explain. One edition a month. Each edition has the same parts: the read, the comparison table, historical charts, material notes, context, uncertainty, and the record.
The publication cannot establish a contractor's actual costs, validate an invoice, determine a fair markup, or decide whether a contractual adjustment is permitted. It does not forecast and gives no purchasing instruction. It does not assume that a contractor or supplier is exaggerating; the record can support a reasonable explanation as readily as it can show an unsupported one.
The six core measures
One producer price index per material group, from the U.S. Bureau of Labor Statistics. All are PPI commodity indexes, not seasonally adjusted. Each has its own base period, so index levels are not comparable across rows; a higher index does not mean a more expensive material. Changes are comparable.
| Material | Series | Base period | What BLS prices |
|---|---|---|---|
| Softwood lumber | WPU0811 | 1982 = 100 | Softwood lumber as it leaves the sawmill: not edge worked, and made from purchased lumber, cut stock and dimension |
| Copper | WPU10260314 | December 1986 = 100 | Copper wire and cable, a manufactured product |
| Steel | WPU1017 | 1982 = 100 | Steel mill products, by product and grade, as they leave the mill |
| Aluminum | WPU102501 | 1982 = 100 | Aluminum mill shapes: semifinished sheet, plate, foil and extrusions |
| Gypsum products | WPU137 | 1982 = 100 | Gypsum products, including gypsum building materials and other gypsum products |
| Ready-mix concrete | WPU1333 | 1982 = 100 | Ready-mix concrete delivered from the producer's plant |
A producer price index measures what domestic producers receive for their output. It is not a delivered price, not a quote, and not any supplier's cost.
Context measures
Each is here to answer a stated question. None is a core measure and none enters a comparison table as if it were.
| Measure | Series | Question it answers |
|---|---|---|
| PPI, cement (hydraulic) | WPU1322 (BLS, 1982 = 100) | Did concrete move with or without its priced input? |
| PPI, gypsum building materials | WPU13710102 (BLS, December 1993 = 100) | Did the narrower wallboard-containing index move with the broad gypsum index? |
| LME copper, monthly average settlement price | World Bank Pink Sheet, U.S. dollars per metric ton | Did copper wire and cable follow the raw metal? |
| LME aluminum, monthly average settlement price | World Bank Pink Sheet, U.S. dollars per metric ton | Did aluminum mill shapes follow the raw metal? |
| Portland and blended cement shipments, U.S. plants | USGS Mineral Industry Surveys, metric tons, monthly | Was the market behind concrete busy or slack? |
| Housing starts, total and single-family | Census New Residential Construction, seasonally adjusted annual rate, thousands | Was residential demand rising or falling? (Not all construction demand.) |
Sources
Official releases and one benchmark compilation, fetched by script with no browser and retained byte for byte:
| Source | What it gives | Cadence and lag |
|---|---|---|
| U.S. Bureau of Labor Statistics, public API v1 | The eight PPI series above, ten years per request, with BLS's own preliminary flags | Monthly, released about two weeks after the reference month. The most recent four months are preliminary and may be revised. |
| World Bank Commodity Price Data (the Pink Sheet), monthly workbook | Monthly average LME settlement prices for copper (grade A cathode) and aluminum (high-grade primary ingot), nominal U.S. dollars per metric ton | Monthly, early in the following month. The workbook link on the landing page changes each release, so the page is captured and the current link followed at run time. |
| U.S. Geological Survey, Mineral Industry Surveys: Cement, monthly workbook | Portland and blended cement shipments from U.S. plants by month, current and prior year, with reported imports; survey figures marked estimated and preliminary by USGS | Monthly, about four to five months after the reference month. The newest workbook is the greatest-dated file name on the page. |
| U.S. Census Bureau, New Residential Construction, starts time-series workbook | Housing units started, total and single-family, seasonally adjusted annual rate and not seasonally adjusted, monthly since 1959 | Monthly, about three weeks after the reference month. Recent months are revised. |
FRED (the Federal Reserve Bank of St. Louis) republishes the BLS series and is a convenient place to view them. It is not fetched here, because a republication is not an independent source and would not corroborate the original.
Rules
- Compare like with like: producer index with producer index, benchmark price with benchmark price. A producer index, an exchange benchmark and a local delivered price are three different things and are never treated as one.
- Keep the sources' own distinctions: raw copper apart from wire and cable, raw aluminum apart from mill shapes, cement apart from ready-mix concrete, the broad gypsum index apart from the narrower building-materials index.
- Preserve exact coverage, geography, units, seasonal-adjustment status, base period and observation date. Display each series' actual latest period rather than implying all figures describe the same month.
- Month over month is against the prior month. Year over year is against the same calendar month a year earlier. Both are rounded to one decimal. Nothing is seasonally adjusted by Field Assembly.
- Do not compare index levels across series. When series are rebased to a common starting month for a comparison chart, that transformation is stated beside the chart and in the data dictionary, and the original values stay in the downloads.
- No composite. The six core series are not averaged into an overall construction-cost figure; they are selected indicators, not a weighted project budget.
- A source that fails to fetch is recorded as failed. Nothing is substituted from another source. A parser that cannot find what it expects fails the source rather than guessing. Rows are never edited or deleted; a corrected reading is appended and the earlier row stays.
- A gap between two series' changes is reported as an observation about two series. It is not a margin, a markup or a cost.
- Correlation, plausible explanation and demonstrated causation are kept apart in the text. Where the sources do not support an explanation, none is forced.
- A measure is added only when it answers a reader's question, and a supplementary source that is hard to maintain or contributes little is omitted and the omission explained.
Revisions and versions
Each edition quotes the values the source published at capture time and marks preliminary and estimated values as the source marked them. A later revision by a source is not applied to an earlier edition; it appears in the next edition's data, and that edition notes it in its uncertainty section when the revision changes a previously quoted comparison by 0.5 points or more. A correction to Field Assembly's own text or calculation is dated on the affected page, given a new version, listed at /corrections/, and leaves the original text visible. New editions never replace earlier ones.
History and contemporaneous capture
The BLS request returns ten years of each series and the Census workbook returns history to 1959, so a first edition can show five-year charts without waiting months to accumulate observations. Every row in the record carries an observation type. The latest period a series held at capture time is CONTEMPORANEOUS_CAPTURE: the figure as first seen, before the source could revise it. Every earlier period is RETROSPECTIVE_HISTORY: history as the source published it on the capture date, which may already include the source's revisions. Collecting public government history creates no exclusive dataset; what accumulates here is the coherent, reproducible record of what was published when, and the interpretation written from it.
Evidence and timestamps
Each edition directory holds numbers.json (every figure the edition quotes, regenerated from the data files), evidence.json (every capture file with URL, retrieval time and SHA-256), exceptions.json (what the run flagged for the editor) and the edition as delivered. Capture files are hashed into an append-only chain that can be anchored externally by RFC 3161 timestamp tokens and OpenTimestamps proofs. An edition page states whether an anchor entry covers its files; none is claimed that has not been made.
An external timestamp shows that a captured file existed unchanged by the anchor date. It does not show that the source's figure was correct.
Editorial review
The collection step fetches the sources, retains them, parses them, computes the comparisons and flags exceptions. A person writes the edition from the computed set; every figure in the text must appear in numbers.json, which a mechanical check enforces before publication. An edition is a draft until a named reviewer has read the evidence and the interpretation and the owner marks it published. Machine-generated drafts are never published as editions.
Corrections
Corrections are made by appending a dated note to the affected edition and listing it at /corrections/. The original text stays visible.