Parquet vs CSV for Property Data

Use Parquet if you're analyzing data in code (pandas, DuckDB, Polars, Spark) or you want the full schema including price and tax history. Use CSV if you want to open it in a spreadsheet or feed a tool that only speaks CSV.

• Parquet — all 32 columns, including the nested price_history and tax_history arrays. A columnar, compressed binary format.
• CSV — the same records as a flat table, minus the two nested history columns (a flat file can't hold arrays cleanly).

Parquet is columnar and compressed, so the same data is often several times smaller on disk than CSV, and queries that touch a few columns read only those columns — much faster on large files. CSV is row-oriented plain text: universally readable, but bigger and slower to scan.

Parquet: pd.read_parquet("file.parquet") in pandas, or SELECT * FROM 'file.parquet' in DuckDB — no import step. CSV opens in Excel, Google Sheets, or any language's CSV reader. Full walkthrough in how to download property data.