Is it just simple extrapolation? $419M in the period from July - Oct 2025 portends $1.676B in the period from July 2025 - July 2026?

Those "Change From..." columns are just simple deltas.

Looks like we're in good shape. This simple roll-up of the budget line-items from the Open NYC data source appears to match the Agency Summary table in the Expense Budget PDF.

Success!

I made a table showing how these columns relate between the two sources.

Note that the Expense Budget PDF has these "Change From..." columns that don't have an analog in the columns from Open NYC data. We'll have to calculate those ones ourselves

The next step is to figure out how the columns from the Open NYC data map to the columns in the PDF.

While there is taxonomical overlap (adopted & modified) in the column names, the PDF table columns refer to previous fiscal years. But we've just filtered to the current FY...

The publication dates for each fiscal year align roughly with the milestones of the budget cycle (a series of proposals, reviews, and responses between the Mayor and the City Council):

January: Preliminary Budget
April: Executive Budget
June: Approved Budget

The dataset has columns for `fiscal_year` and `publication_date`, implying that the dataset contains more than just the current fiscal year's budget data.

We can see there are 3 budgets per fiscal year, each having 30-40K rows.

This implies that the tables in the official PDF can be recreated with the spreadsheet. Let's test this out by reconstructing the table that summarizes authorized Expense spending for FY 2026 by agency.

Let's take the Expense Budget as an example. The Expense Budget exists in two forms:

1) a 1000-page PDF
2) a spreadsheet with 1,000,000 rows

Ostensibly, the spreadsheet is raw data product that is used to create the PDF