National Bureau of Economic Research

What You're Getting

NBER Research Data provides access to datasets assembled for academic economic research — historical US economic data, demographic microdata, patent citation networks, mortality statistics, and other specialized datasets. Notable collections: NBER Macrohistory Database (long historical series 1867+), CPS Extract (Current Population Survey extracts), Patent Data Project (USPTO patent records), and NBER-CES Manufacturing Industry Database. Data is researcher-compiled, not official statistics.

Ingestion Strategy

NBER data is distributed as downloadable files — no API. Each dataset has its own structure and format. The Macrohistory Database offers consistent annual/monthly time series as CSV/XLSX. For complex microdata (CPS, patent), SAS/Stata input scripts are often provided. Plan for per-dataset schema variation.

import requests, io, pandas as pd

# NBER Macrohistory database (annual series)
url = "https://www.nber.org/research/data/nber-macrohistory-database"
# Data files are linked per series; download the master index first
index_url = "https://data.nber.org/data/macro-historical/narrative-database/bci/"
# Typical file: CSV with columns: date, series_value

Rate Limits & API Details

• No API; all access via direct file download
• No authentication for most public datasets
• Some datasets require NBER restricted data access (institutional researchers only)
• Files hosted at data.nber.org — standard HTTPS

Schema Stability

Each NBER dataset has its own schema — there is no uniform structure. Schemas are set when the dataset is published and rarely change, but new versions of datasets may restructure columns. Always check the codebook PDF for variable definitions. Dates are inconsistently formatted across datasets (YYYYMM, YYYY, or full date strings).

Data Quality Gotchas

• Academic, not operational: NBER data is research-grade — may have methodological peculiarities documented in working papers
• Historical data: Macrohistory series were reconstructed from historical sources and may not match other historical databases
• Infrequent updates: many NBER datasets are one-time publications; do not expect regular refresh
• Format heterogeneity: CSV, XLSX, Stata .dta, SAS .sas7bdat, fixed-width — build format-specific parsers per dataset

Tool Compatibility

• Python: pandas for CSV/XLSX; pyreadstat for Stata/SAS files
• R: native for Stata/SAS formats
• No standard ETL connector support
• Snowflake/BigQuery: use manual file upload or cloud storage staging

Benchmarking Approach

Macrohistory Database: US GNP series should start from the 1870s. Verify a known historical value (e.g. 1929 GNP) against published historical economic history sources. For patent datasets: total US patents should match USPTO annual report figures. Set alerts for failed downloads (NBER occasionally reorganizes file paths).