# Boston Property Assessment Rolls, FY2004 to FY2026

## Facts

- Provider: City of Boston (Assessing Department)
- License: Open Data Commons Public Domain Dedication and License (PDDL) 1.0
- Format: Parquet (in sampled files)
- Stored at: Amazon S3, s3://scigantic-boston-open-data/property-assessment (us-east-1)
- Access: public, readable without an account
- Category: Urban & Civic
- Source: https://data.boston.gov/dataset/property-assessment
- Page: https://scigantic.com/archive/7677e08e-5504-574b-b1d3-a1879a7605bc
- Keywords: Urban & Civic, Parquet, AWS Open Data

## About

The City of Boston Assessing Department's annual parcel-level assessment roll, published for fiscal years 2004 through 2026: 3,673,709 parcel-year rows across 23 DataStore-loaded files. One row is one parcel (or one condominium unit) in one fiscal year. The modern files carry the parcel identifier and GIS id, the street address and ZIP, land use code and description, building type, an owner-occupancy flag, the owner name and mailing address, land area and gross and living area, land value, building value and total assessed value, gross tax, year built and year remodeled, and a long tail of structural and condition attributes: roof structure and cover, interior and exterior wall finish, interior, exterior and overall condition, bedroom, bathroom and kitchen counts with per-fixture style codes, heating type and system, air conditioning, fireplaces, parking count, orientation, view and corner-unit flags. Joined across years on the parcel identifier, this is a 23-year longitudinal record of assessed value, physical condition and ownership for essentially every property in the city, which supports work on valuation trends, gentrification and displacement, absentee and corporate ownership, the housing stock's age and condition, and the geography of the tax base. Three facts govern any multi-year analysis and are documented in full in the schema card. The column set is not stable: it runs 23 columns in the FY2004 to FY2007 'lite' text files, 55 to 57 in FY2008 to FY2015, 76 to 78 in FY2016 to FY2021, and 61 to 68 in FY2022 to FY2026, and the early files embed the fiscal year in the column names themselves (FY2004_TOTAL, OWNER FY04) alongside genuine typos that must be reproduced exactly to select them ('LIVING _AREA', 'FY200_ LAND'). The FY2014 resource is listed as DataStore-active but holds zero rows, so a naive loop over the resource list silently drops that year; the Scigantic mirror recovers it, because the published FY2014 CSV is intact at 164,091 rows and only the portal's copy of it is empty. And the portal's own dataset notes state that PID, CM_ID, GIS_ID, ZIPCODE and MAIL_ZIP_CODE carry a trailing underscore to preserve leading zeros, which is no longer true of either the DataStore columns or the published FY2026 CSV; that documentation is stale and code written against it selects columns that do not exist. Monetary and area columns are text with embedded commas, dollar signs and padding, and the raw CSV header ships GROSS_TAX with a leading and trailing space while the DataStore serves it trimmed, so the same column has two different names depending on which access path is used. Published by the Assessing Department under PDDL 1.0. This archive is served from a Scigantic mirror rather than live from the portal: the fiscal-year files are immutable once published, so s3://scigantic-boston-open-data/property-assessment/ carries a byte-verbatim copy under raw/ alongside a Snappy Parquet copy under parquet/, FUSE-mounted read-only at ~/archive. The Parquet layer keeps every column as a string, matching the DataStore and preserving the zero-padded parcel and ZIP identifiers that type inference would destroy.

## Load the data

```python
import os
import pandas as pd
df = pd.read_csv(os.path.join(os.environ['SCIGANTIC_MOUNT_PATH'], 'raw/fy2026-property-assessment-data_rev.csv'), nrows=1000)
df.head()
```

Run in a Scigantic notebook on 2026-09-14 against `raw/fy2026-property-assessment-data_rev.csv`: CSV/TSV: columns ['PID', 'CM_ID', 'GIS_ID', 'ST_NUM', 'ST_ALPHA', 'ST_NUM2', 'ST_NAME', 'UNIT_NUM', 'CITY', 'ZIP_CODE', 'BLDG_SEQ', 'NUM_BLDGS'], dtypes ['int64', 'float64', 'int64', 'float64', 'str', 'float64', 'str', 'str', 'str', 'int64', 'int64', 'int64'], first 1000 rows read.

From any machine:

```python
import boto3
from botocore import UNSIGNED
from botocore.client import Config
s3 = boto3.client('s3', region_name='us-east-1', config=Config(signature_version=UNSIGNED))
s3.list_objects_v2(Bucket='scigantic-boston-open-data', Prefix='property-assessment/', Delimiter='/')
```

Lists the files with anonymous access. Runs on any machine, with no cloud account.

## Data structure

File types in sampled files: Parquet

Example file: `property-assessment/parquet/ast2018full.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/7677e08e-5504-574b-b1d3-a1879a7605bc
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/7677e08e-5504-574b-b1d3-a1879a7605bc
- Over MCP: `get_schema_card` with id `7677e08e-5504-574b-b1d3-a1879a7605bc` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Urban & Civic datasets](https://scigantic.com/topics/urban-and-civic)
