This is a mirror of all data files archived from Data.gov. The data can be browsed and searched at Data.gov Archive Search.

This repository is maintained by the Harvard Law School Library Innovation Lab as part of our project to preserve U.S. federal public data. For more information, see the Public Data Project.

Collection Format

Each dataset on Data.gov has a unique slug known as its name. We store each dataset in this repository as:

collections/data_gov/<name>/<version>.zip

We also store a metadata file for each dataset in the metadata directory:

metadata/data_gov/<name>/<version>.json

<version> is a v followed by the number of times we have downloaded the dataset (v1, v2, etc.)

For example, the Data.gov dataset https://catalog.data.gov/dataset/fruit-and-vegetable-prices is stored in this repository as:

• collections/data_gov/fruit-and-vegetable-prices/v1.zip
• metadata/data_gov/fruit-and-vegetable-prices/v1.json

Dataset Format

Each dataset zip file is a BagIt package created by our bag-nabit tool.

BagIt is a simple file format, established by the Library of Congress, consisting of a folder of metadata and text files. Our BagIt files follow this directory structure:

• data/
  • files/:
    • ...: these are the actual files you likely want to use as a researcher, downloaded from the Data.gov listing.
  • headers.warc: request and response headers from HTTP fetches for files in files/
  • signed-metadata.json: metadata including Data.gov's API description of the dataset

The bags also contain these files, which are useful for authenticating the provenance of the data:

• bagit.txt: standard BagIt file
• bag-info.txt: standard BagIt file
• manifest-sha256.txt: standard BagIt file
• tagmanifest-sha256.txt: standard BagIt file
• signatures/: directory of signature files

Metadata File Format

Each metadata JSON file contains three main attributes:

1. bag_info: Contains the BagIt metadata including:

   • Bag-Software-Agent: The version of nabit used to create the archive
   • Bagging-Date: When the archive was created

2. signed_metadata: Contains detailed information about the dataset including:

   • id: A UUID for this specific archive
   • url: The Data.gov URL for the dataset
   • description: A brief description including the dataset title and creating organization
   • data_gov_metadata: The complete metadata from Data.gov's API, including:
     • Dataset details (title, description, etc.)
     • Organization information
     • Resource listings
     • Tags and other metadata
   • collection_tasks: Records of the HTTP requests made to collect the dataset

3. zip_entries: Listing of each entry in the collection zip file, which can be used to fetch individual files from the zip file via range request without downloading the entire archive.

Rollup files

There are several rollup files at the top level to help with finding datasets of interest:

• metadata.csv.zip: CSV listing the name, organization, title, date, metadata path, and collection path for each dataset
• metadata.jsonl.zip: JSON lines file with complete metadata for each dataset, including the signed_metadata and zip_entries sections (equivalent to downloading the metadata/ directory as a single file)
• file_listing.jsonl.zip: zipped JSON lines file showing the s3 listing of all files in the repository

Downloading data

To download an individual dataset by name you can construct its URL, such as:

To download large numbers of files, we recommend the aws or rclone command line tools:

aws s3 cp s3://us-west-2.opendata.source.coop/harvard-lil/gov-data/collections/data_gov/<name>/v1.zip --no-sign-request

Data Limitations

Data.gov includes many kinds of datasets, including some that link to actual data files, such as CSV files, and some that link to separate websites. Our process runs a "shallow crawl" that collects only the directly linked files. Datasets that link only to a separate website will need to be collected separately.

Source code

The source code used to generate this and other repositories is available at https://github.com/harvard-lil/data-vault.

We welcome conversation and collaboration in the issue tracker for that project.

Collection Dates and Update Schedule

Files in this repository were collected intermittently between 2024-11-19 and 2025-02-06.

In the future, we hope to update the repository daily.