Aluminum Data Provider: Production, Trade, Prices, Scrap and Recycling · Head-to-head
USGS Mineral Yearbook - Aluminum and Bauxite chapters vs Hugging Face global_aluminum_price_daily
Which aluminum data provider: production, trade, prices, scrap and recycling data fits your job: USGS Mineral Yearbook, or Hugging Face global_aluminum_price_daily. API, files, or your warehouse. Daily, weekly, or hourly.
USGS Mineral Yearbook
Hugging Face global_aluminum_price_daily
Where the fields line up
No shared field names. These two answer different questions.
| Field | USGS Mineral Yearbook | Hugging Face global_aluminum_price_daily |
|---|---|---|
Primary production: Quantity | documented | not in this set |
Primary production: Value | documented | not in this set |
Price, average, U.S. market, spot | documented | not in this set |
Stocks, December 31: Aluminum industry | documented | not in this set |
Secondary recovery: New scrap / Old scrap | documented | not in this set |
Exports, crude, semicrude, and scrap | documented | not in this set |
Imports for consumption, crude and semicrude | documented | not in this set |
World, production | documented | not in this set |
TABLE 2 - Primary annual production capacity by company | documented | not in this set |
TABLE 8 - Aluminum prices | documented | not in this set |
TABLE 10/12 - Exports/imports by country or locality | documented | not in this set |
TABLE 13 - World production by country | documented | not in this set |
Coverage, side by side
| USGS Mineral Yearbook | Hugging Face global_aluminum_price_daily | |
|---|---|---|
| Geographic | United States in depth; world production by country for all major producers; Canada inside the end-use shipment tables | Global only - a single world-level series, location code WLD |
| Temporal | Editions from 2016 to 2023, the aluminum chapter itself present for 2019, 2020 and 2021, each carrying roughly five years of salient statistics | Daily observations from 2014-05-06 to 2026-07-28, dated in both Gregorian and Persian Jalali calendars |
What each contains
Pick by fit, not by loyalty.
| USGS Mineral Yearbook | Hugging Face global_aluminum_price_daily | |
|---|---|---|
| Publisher | USGS, the U.S. Geological Survey | Hugging Face, the dataset card published by user Farmaanaa |
| Subject lens | Industry structure: production, smelter capacity by company, scrap classes, trade by partner, end-use shipments - with a sibling chapter for bauxite and alumina | Market price: one global aluminum series rendered as daily OHLC bars in two subsets |
| Geographic coverage | United States in depth; world production by country for all major producers; Canada inside the end-use shipment tables | Global only - a single world-level series, location code WLD |
| Temporal coverage | Editions from 2016 to 2023, the aluminum chapter itself present for 2019, 2020 and 2021, each carrying roughly five years of salient statistics | Daily observations from 2014-05-06 to 2026-07-28, dated in both Gregorian and Persian Jalali calendars |
| Detail level | National annual totals down to facility-level US capacity and bilateral country trade flows | One row per trading day, four prices deep, in wide mart and long observations layouts |
| Formats | PDF chapters with matching spreadsheets | Parquet tables |
| Scale | Roughly twenty pages and thirteen tables per aluminum chapter; twelve documented fields | About 3,038 rows per subset, 6,076 across the two; ten documented fields |
| Best for | Supplier screens, trade-flow mapping, industry structure and capacity studies | Backtesting, volatility research and feature engineering on the world price |
What each does better
USGS Mineral Yearbook
If you need to know which reduction capacity actually exists and in whose hands, this is the only table in the pairing that answers - see primary aluminium production.
Trade texture by partner and class. Separate export and import tables break crude, semicrude and scrap categories out by partner country - the bilateral grain that turns 'the US imported 4,830 thousand tons' into a mappable flow, the substance of bilateral trade flows.
Scrap and secondary recovery, properly classified. Purchased new and old scrap are recovered and reported by class, with secondary alloy production by independent smelters broken out separately - the accounting layer behind any serious look at recycling scrap supply.
World context with rank and share. The closing table sets US output against world production by country, China's 58 percent share included, so the domestic story always arrives with its denominator attached.
Hugging Face global_aluminum_price_daily
Resolution, by three orders of magnitude. Roughly 3,038 daily rows per subset - about twelve years of trading days from 2014-05-06 to 2026-07-28 - where the yearbook offers one annual average per year. Any volatility, drawdown or regime question needs the daily spine; annual averages erase exactly the variation such questions exist to study.
The 2014 opening stretch reads like a tape: 2,182.75 to 2,205.75 to a 2,172.75 close on day one, then straight down to a 2,149 close the next.
A double calendar, kept honestly. Every row is dated twice - date_greg in Gregorian notation and period in the Persian Jalali calendar (1393-02-16 alongside 2014-05-06) - so calendar-aligned regional workflows need no external conversion table.
Typed, self-describing columns. unit declared as usd_per_ton, freq as D, scope coded WLD, series identifiers carried as data. That shape drops straight into pandas or Polars as-is, which is why it is the slice's natural modeling table.
Where they're equivalent
More than their shapes suggest. Both field dictionaries were verified during research, checked against the artifacts themselves rather than inferred from descriptions. Both are strictly about aluminum - not one commodity column inside a sixty-commodity spreadsheet - so neither forces you to filter your way to the metal. Both declare their units explicitly, cents per pound and thousand metric tons on one side, usd_per_ton on the other. Both fix every row to a stated geography, named countries versus the WLD world code. And both are tabular with stable meanings: thirteen tables of consistent layout per chapter on one side, two tidy subsets on the other, each ready to become a country-year time series or its daily equivalent. Where they stop being equivalent is everything downstream of the schema: one accumulates structure slowly, the other ticks every trading day.
The verdict
Verdict: sample both, pick by fit - they are different instruments pointed at the same industry.
Take USGS Mineral Yearbook if your question contains an institution. Which smelters remain standing and at what capacity, who imports American scrap and who sells the US its crude metal, how end-use shipments split between construction, transport and packaging, where the US ranks in world production - anything answered by a structure, a ranking or a flow. Accept annual grain: the numbers arrive as national totals, not sessions.
Accept one geography: the world, undifferentiated.
Sample both, pick by fit. See USGS Mineral Yearbook · See Hugging Face global_aluminum_price_daily
Or take both in one feed
A defensible workflow: read the environment from HF - the daily close series in dollars per tonne - then read the physical base from USGS: which smelters stand, at what capacity, shipping what to whom.
Two cautions from the records. First, there is no shared key: USGS keys to years, countries, companies and commodity classes while HF keys to dates and series identifiers, so the bridge is a deliberate mapping - annual averages computed from the daily bars are the natural join artifact. Second, mind the vintages: each yearbook chapter states the cut-off date behind its tables, and figures are rounded to three significant digits, so stamp which edition fed any merged output rather than assuming the latest.
Browse the rest of the shelf at the aluminum data hub and the ranked best aluminum datasets. Datadory ships either record alone or both merged onto one calendar, delivered daily, weekly, or hourly - your call. Or take both in one feed.
API, files, or your warehouse. Daily, weekly, or hourly.
Fair questions
Do the two datasets cover the same ground?
Everything else diverges - tonnage, capacity, trade and end-use shipments on one side; a single undifferentiated world series on the other. They are complements, not substitutes.
Which dataset reaches further back?
Depends which axis you mean. By calendar span, HF wins: a continuous daily series from 2014-05-06 to 2026-07-28. By archival depth, the yearbook goes further - editions exist from 2016 through 2023 and each chapter reissues roughly five years of salient statistics - but the aluminum chapter itself appears only for 2019, 2020 and 2021 in the current run, and never below annual grain. For twelve years of days, take HF; for half-century context assembled edition by edition, take the yearbook.
Which should a quant backtesting aluminum strategies sample first?
Start with HF global_aluminum_price_daily - daily OHLC bars in a typed, self-describing schema load straight into pandas or Polars, which is exactly what a backtest consumes.