Access the 'Anthropic Economic Index' Dataset

Provides clean, tidy access to the 'Anthropic Economic Index' (AEI) dataset hosted on 'Hugging Face' < https://huggingface.co/datasets/Anthropic/EconomicIndex>. The AEI is a recurring release from 'Anthropic' that maps usage of the 'Claude' family of large language models to occupations and tasks using the 'O*NET' taxonomy and the 'Standard Occupational Classification' system, following the methodology of Handa et al. (2025) and the privacy-preserving system 'Clio' of Tamkin et al. (2024) . Functions list available releases, fetch raw and enriched usage tables, retrieve task statements, request hierarchies, country-level breakdowns, and the standalone labor market impacts tables (job exposure and task penetration), compare two releases, join the index to user-supplied data on a shared key, and compute usage-concentration metrics (Herfindahl-Hirschman Index, top-N concentration ratios, Shannon entropy). Data is cached locally for subsequent calls. Reproducibility helpers produce 'BibTeX' or plain-text citations that include the methodological source paper. This product uses the 'Anthropic Economic Index' data (released under CC-BY by 'Anthropic') but is not endorsed or certified by 'Anthropic'.

aieconindex

Tidy R access to the Anthropic Economic Index dataset.

Background

The Anthropic Economic Index (AEI) is a recurring open dataset that maps real Claude conversations to occupations and tasks. Anthropic classifies millions of conversations against the U.S. Department of Labor's O*NET task taxonomy and the Standard Occupational Classification (SOC) system, and publishes the resulting usage shares on Hugging Face under CC-BY-4.0. Each release also splits conversations into automation-style interactions (the user delegates to Claude) and augmentation-style interactions (the user works through a task with Claude). From the September 2025 release onwards, the data is broken down by country and US state. Methodology is documented in Handa et al. (2025); the privacy-preserving classification pipeline is described in Tamkin et al. (2024).

Six releases have shipped between February 2025 and June 2026, covering Claude 3.5 Sonnet through the current model family. From June 2026 the index moved to calendar-month aggregates with country and subregion breakdowns, and Anthropic also publishes standalone labor market impact tables (job exposure and task penetration). aieconindex lists releases, fetches usage tables, retrieves task statements and request hierarchies, exposes country and subregion slices, fetches the labor market tables, caches downloads, and produces ready-made citations. Schema differences across releases are handled internally, and releases published after this package version was built still resolve via the live Hugging Face listing. Three runtime dependencies (cli, httr2, jsonlite) plus base R. No API key needed.

Table of contents

• Background
• Installation
• Quick start
• Function reference
  • Discovery
  • Download
  • Structured access
  • Analysis
  • Reproducibility
  • Cache management
• The aei_tbl class
• Common workflows
• Releases covered
• Caching
• Relationship to the Anthropic Python notebooks
• Related work
• Citation
• Contributing
• Licensing and attribution

Installation

install.packages("aieconindex")

# or the development version
# install.packages("remotes")
remotes::install_github("charlescoverdale/aieconindex")

R 4.1.0 or later.

Quick start

library(aieconindex)

# 1. See what's available
aei_releases()
#> # AEI: releases · 6 rows
#>           release_id release_date               model
#> 1 release_2026_06_26   2026-06-26   All Claude models
#> 2 release_2026_03_24   2026-03-24 Claude Opus 4.5/4.6
#> 3 release_2026_01_15   2026-01-15   Claude Sonnet 4.5
#> 4 release_2025_09_15   2025-09-15     Claude Sonnet 4
#> 5 release_2025_03_27   2025-03-27   Claude 3.7 Sonnet
#> 6 release_2025_02_10   2025-02-10   Claude 3.5 Sonnet
#> ...

# 2. Look inside a release
aei_files("2025-09-15", recursive = TRUE)

# 3. Fetch the canonical usage table
df <- aei_index("2025-09-15", source = "claude_ai", variant = "enriched")

# 4. Slice to a country
uk <- aei_geography("2025-09-15", country = "GBR")

# 5. Cite the dataset
aei_cite("2025-09-15", format = "bibtex")

Function reference

Discovery

aei_releases(live = FALSE)        # offline-safe (uses bundled metadata)
aei_files("latest")               # tree of the most recent release
aei_files("2025-03-27", recursive = FALSE)  # top-level only

Download

aei_index() locates the canonical usage CSV by file-pattern matching. Arguments:

• source: "claude_ai" (consumer product traffic) or "1p_api" (first-party API). Not all releases include both.
• variant: "raw" (counts and percentages from Anthropic's pipeline) or "enriched" (joined to O*NET / SOC metadata, with derived per-capita and tier metrics). Older releases may only ship one variant. Ignored for releases from 2026-06-26 onwards, which ship a single file per source.

df_monthly  <- aei_index("2026-06-26", source = "claude_ai")
df_raw      <- aei_index("2026-03-24", source = "claude_ai", variant = "raw")
df_enriched <- aei_index("2025-09-15", source = "claude_ai", variant = "enriched")
df_api      <- aei_index("2026-06-26", source = "1p_api")

aei_download() fetches any path returned by aei_files():

soc       <- aei_download("2025-03-27", "SOC_Structure.csv")
hierarchy <- aei_download("2025-09-15",
                          "data/output/request_hierarchy_tree_claude_ai.json")
report    <- aei_download("2026-01-15", "aei_v4_appendix.pdf")  # returns local path

Structured access

# Clio-derived request hierarchy (from 2025-09-15 onwards)
tree <- aei_clusters("2025-09-15", source = "claude_ai")

# Bundled O*NET task statements (ships in 2025-03-27)
tasks <- aei_tasks("2025-03-27")

# UK country slice (geographic facets ship from 2025-09-15 onwards)
uk <- aei_geography("2025-09-15", country = "GBR")

# Australia country slice
au <- aei_geography("2025-09-15", country = "AUS")

# US state-level breakdown
us_states <- aei_geography("2025-09-15", geography = "state_us")

# All subregions in the monthly schema (2026-06-26 onwards)
subs <- aei_geography("2026-06-26", geography = "subregion")

# Standalone labor market impacts tables
exposure    <- aei_labor_market("job_exposure")
penetration <- aei_labor_market("task_penetration")

Country codes are ISO-3 ("GBR", "AUS", "USA"). In the monthly schema (2026-06-26 onwards) subregions carry ISO 3166-2 codes, so US states appear as "US-CA", "US-NY", and so on. Releases before 2025-09-15 have no geographic data; the function errors informatively.

Analysis

# How did the cluster shares move between Sept 2025 and March 2026?
diff <- aei_compare("2025-09-15", "2026-03-24")
head(diff[order(-abs(diff$delta)), ])

# Splice AEI country shares to your own GDP-per-capita table
overlay <- data.frame(
  geo_id = c("GBR", "AUS", "USA"),
  gdp_pc = c(48000, 65000, 80000)
)
joined <- aei_link(aei_geography("2025-09-15"), overlay, by = "geo_id")

# How concentrated is UK Claude.ai usage across O*NET tasks?
uk <- aei_geography("2025-09-15", country = "GBR")
uk_tasks <- uk[uk$facet == "onet_task" & uk$variable == "onet_task_pct", ]
aei_concentration(uk_tasks)

aei_link() is a thin wrapper over base::merge() that preserves the aei_tbl class and provenance metadata, supports left / inner / full joins, and warns when a join produces zero rows. Use it to attach occupational crosswalks (SOC, ANZSCO, ISCO, SOC2020 UK), national labour-force data (ONS, BLS OEWS, ABS), or anything else keyed on country code or task identifier.

Reproducibility

By default aei_cite() returns both the dataset citation and Handa et al. (2025). Set method = FALSE for the dataset only.

aei_cite()                                         # text, project-wide, with paper
aei_cite("2025-09-15", format = "bibtex")          # BibTeX, both refs
aei_cite("2026-03-24", format = "bibentry")        # bibentry object (multi-entry)
aei_cite(format = "text", method = FALSE)          # dataset only

Cache management

The aei_tbl class

All data-returning functions emit an aei_tbl: a data.frame subclass with provenance metadata stored in the aei_query attribute. The metadata carries endpoint, the resolved release identifier, the source URL, and the fetch timestamp; it is preserved across row and column subsetting.

df <- aei_index("2025-09-15")
attr(df, "aei_query")
#> $endpoint   "index"
#> $release    "release_2025_09_15"
#> $facet      "raw/claude_ai"
#> $source_url "https://huggingface.co/datasets/Anthropic/EconomicIndex/.../aei_raw_claude_ai_*.csv"
#> $fetched_at "2026-04-28 18:34:00 BST"

# Custom print method shows the provenance header
print(df)
#> # AEI: index · release=release_2025_09_15 · facet=raw/claude_ai · 12345 rows
#> ...

# Subsetting preserves the class and attribute
sub <- df[df$value > 1, ]
class(sub)
#> [1] "aei_tbl" "data.frame"

The class inherits from data.frame, so any function that takes a data frame works without conversion. Drop the class with as.data.frame() if you need a plain frame.

Common workflows

Pin a release for production. Default release = "latest" resolves to the most recent release at call time, which is fine for exploration but unsuitable for reproducible pipelines. Pin a release identifier explicitly:

RELEASE <- "2025-09-15"  # or "release_2025_09_15"
df <- aei_index(RELEASE, source = "claude_ai", variant = "enriched")

Replicate an Anthropic figure. Anthropic ships Python replication notebooks (v2_report_replication.ipynb) inside several releases. To replicate the augmentation-vs-automation headline figure in R:

df <- aei_download("2025-03-27", "automation_vs_augmentation_v2.csv")
df$family <- ifelse(df$interaction_type %in% c("directive", "feedback loop"),
                    "Automation", "Augmentation")

Country exposure ranking. Top O*NET tasks for the UK by share of Claude.ai usage:

uk <- aei_geography("2025-09-15", country = "GBR")
top <- subset(uk, facet == "onet_task" & variable == "onet_task_pct")
top <- top[order(-top$value), ][1:15, c("cluster_name", "value")]

Cross-country comparison. Per-capita usage index for selected economies:

df <- aei_index("2025-09-15", source = "claude_ai", variant = "enriched")
country_overall <- subset(df,
  geography == "country" &
  variable  == "usage_per_capita_index" &
  cluster_name == "not_classified" &
  level == 0
)
country_overall <- country_overall[order(-country_overall$value), ]

Cite in a paper. Drop the BibTeX form straight into your .bib:

cat(aei_cite("2025-09-15", format = "bibtex"), file = "refs.bib", append = TRUE)

Releases covered

The package recognises every release published to Hugging Face up to 2026-06-26. Releases published after this package version was built are discovered automatically via the Hugging Face tree API: they appear in aei_releases(), resolve in every data function (including "latest"), and default to the newest known schema.

Each release ships its own data_documentation.md on Hugging Face. The package's aei_releases() blends bundled metadata (model, report URL) with a live Hugging Face listing. Anthropic also publishes a standalone labor_market_impacts directory (job exposure and task penetration tables), available via aei_labor_market().

Caching

Downloaded files are cached under the path returned by aei_cache_dir(), which defaults to tools::R_user_dir("aieconindex", "cache"). Override before the first call:

options(aieconindex.cache_dir = "/your/preferred/path")

Cache is keyed by release identifier and relative path, so re-downloads are byte-identical to the original.

aei_cache_info()
#> $dir         "/Users/.../aieconindex/cache"
#> $n_files     3
#> $size_bytes  126839425
#> $size_human  "121.0 MB"
#> $files       <data.frame: 3 rows>

aei_cache_clear()  # removes all cached files

The latest release usage CSVs are large (the June 2026 claude_ai file exceeds 200 MB), so the first call to a fresh release is bandwidth-heavy; aei_index() reports the file size before starting any download over 50 MB. Subsequent calls are served from disk.

Relationship to the Anthropic Python notebooks

Anthropic ships its own replication code as Jupyter notebooks inside several releases (e.g. release_2025_03_27/v2_report_replication.ipynb). For exact figure replication, use those. aieconindex is the R-side equivalent of Hugging Face's Python datasets loader: typed, cached access to the same source files, with downstream analysis left to you.

Related work

• Anthropic/EconomicIndex on Hugging Face: the upstream dataset.
• Handa et al. (2025), arXiv:2503.04761: the methodological source paper.
• Tamkin et al. (2024), arXiv:2412.13678: the Clio system paper.
• Felten, Raj, and Seamans (2021): the AI Occupational Exposure (AIOE) measure, on the same O*NET task taxonomy.
• Acemoglu and Restrepo (2020): the canonical automation-and-jobs paper referenced by AEI's interaction-type framework.
• O*NET Database: the U.S. Department of Labor's task taxonomy.
• Standard Occupational Classification (SOC): the U.S. Bureau of Labor Statistics' occupational classification system.
• Anthropic Economic Futures: Anthropic's broader economic research programme.

Related packages

Citation

Cite both the package and the underlying dataset:

citation("aieconindex")
aei_cite("2025-09-15", format = "bibtex")

aei_cite() returns the dataset citation alongside Handa et al. (2025), the methodological source paper.

Contributing

Issues and pull requests welcome at https://github.com/charlescoverdale/aieconindex/issues. Useful contributions for v0.3 include:

• Occupational crosswalks (SOC ↔ ANZSCO ↔ ISCO ↔ SOC2010_UK)
• Joins to ONS Labour Force Survey, BLS OEWS, ABS Labour Force
• Replication helpers for specific Anthropic figures
• Vignettes demonstrating multi-release comparisons

For Anthropic-introduced schema changes that break aei_index() or aei_geography(), please open an issue with a sample of the new file structure (output of aei_files(<new_release>)).

Licensing and attribution

This package is released under the MIT License.

The underlying Anthropic Economic Index dataset is released by Anthropic under Creative Commons Attribution 4.0 International (CC-BY-4.0). When using this package to retrieve or redistribute that data, attribution to Anthropic and to Handa et al. (2025) is required. Use aei_cite() for ready-made citation strings.

The bundled O*NET and SOC reference data (when accessed through the AEI) inherit their respective licences. See the O*NET licensing page and the BLS Standard Occupational Classification documentation.

This product uses the Anthropic Economic Index data but is not endorsed or certified by Anthropic.