Skip to main content

Updating WiNDCNational with the Latest US Data

Mitch Phillipson November 14, 2025

julia data

The United States released 2024 updates to Supply/Use tables a few months ago. I have been working on a larger project and have not yet had time to view the updates and incorporate them into WiNDCNational. This past week I spent some time refining WiNDCNational and updating the documentation.

Changes to the SUT Data

Fortunately, there were no structural changes to the SUT data, no added commodities, sectors or other data. However, the BEA did update the location of the data download and the file names. Luckily, they simplified both. Previously, the file names included the year of the data, e.g., Supply_Tables_1997-2023.xlsx. While this was helpful to distinguish between years, it meant I had to use a regular expression to match the file names. This was less than ideal as users may not be familiar with regex, which just creates another barrier to entry. The new file names are simply Supply_Summary.xlsx and Use_Summary.xlsx, which are much easier to work with.

To load this new data into WiNDCNational, the only modification needed is updating the YAML file to update these names.

download_url_common: "https://apps.bea.gov/industry/release/zip/SUPPLY-USE.zip"
download:
supply:
    pattern: "^Supply_Summary.xlsx$"
use:
    pattern: "^Use_Summary.xlsx$"

I’ve left the ^ and $ anchors in place to ensure that only the exact file names are matched, and to hint at the regex ability. This should make it easier for users to understand the matching process without needing to dive deep into regex syntax.

The Goal of WiNDCNational

The system I am designing to make it easy to load and use economic data. At the national level, most countries provide the same data, but the formats and structures vary widely. WiNDCNational aims to abstract away these differences and provide a consistent interface for users to access and analyze national economic data, regardless of the source country.

This has already been done for Australia. The Australian data is structured very similarly to the US data, so adding it was fairly straightforward. The biggest issue being that Australia releases their data as XLSB files, which Julia does not currently support. I had to convert the files to XLSX format before processing them.

However, there are a few assumptions that WiNDCNational makes based on the US data:

  1. Supply and Use are in separate files
  2. Each year is in a separate sheet
  3. The data is in wide format (i.e., commodities as rows, sectors as columns)

Let’s look at the Canadian IO tables. Canada provides both XLSX and CSV formats. The CSV file is a long table with all provinces, margins, and taxes included. This is fantastic for analysis, but they combine the descriptions with the codes in a single column, which makes it (slightly) more difficult to parse. The XLSX files are in wide format, but there are separate files for each province and each province has two files, one for Margins and Taxes and one for the Supply/Use data. The Supply/Use files have the tables on different sheets, but also include full summary level breakdowns of the taxes and margins.

The Canadian data is fantastic. I want to load this data into WiNDCNational and build a model using the full tax and margin structure. However, this is going to require some work to accommodate the different data structure. Unfortunately, this is going to require time which is at a premium right now.

The Canadian data is motivation to create similar data for the United States. A full table of taxes by commodity and sector would be extremely useful for building more accurate models. However, this is not currently available from the BEA. It would be interesting to explore options for estimating this data.