Census of Manufactures - Mobile Header
Census of Manufactures

About the Data

This project assembles and harmonizes historical manufacturing data from the Census of Manufactures, utilizing all surviving establishment-level manuscript schedules from 1850, 1860, 1870, and 1880.

The resulting data enable researchers to analyze industrial change, geographic patterns, and firm dynamics throughout the mid-19th century in the United States:

This is an expansion of samples of CMF microdata gathered for the paper “The Regional Diffusion and Adoption of the Steam Engine in American Manufacturing” by Jeremy Atack, Fred Bateman, and Thomas Weiss.

These data reflect all known surviving establishment-level manuscripts, but do not cover particular counties and industries in some decades, and so aggregating these data by geography or industry could give a wildly misleading impression of total manufacturing activity by place and industry. We discuss the data coverage in detail, on this website and in the documentation that describes the data cleaning process in detail. Our general approach to data collection and data cleaning has been to digitize all establishments, and reflect the underlying manuscripts, but there will be errors in reflecting the economics of individual establishments both due to digitization errors and original collection errors by the Census enumerators. We have not imposed particular economic structure on the data, such as what might seem plausible from the perspective of a production function, but individual users of the data should take care when considering the presence of outliers and other features of the data that might drive spurious results (or imposing further data cleaning assumptions that themselves might drive spurious results).

So, in general, users should beware to understand the intricacies of the data coverage and variables.

Click the following link to download a PDF of the complete set of documentation for this project:

Download Documentation
CMF Manuscript Navigation Overview of how to interpret the CMF images.
Prepared CMF Data Structure Available Variables, with discussion of differences by year.
Coverage The surviving establishment manuscripts omit some states completely, in particular decades, or portions of some states. We provide detailed discussion of dataset coverage over time, including the industry-specific special schedules and differences from the published CMF tabulations.
FAQs Common questions and important notes about data limitations.
Methodology Full cleaning and preparation procedures, including sub-pages for each processing step.
Change Log Ongoing record of dataset revisions.

Citation

Publications using any data from this website should cite the following paper:

"Gaining Steam: Technology Diffusion with Recurring Lock-in," Richard Hornbeck, Shanon Hsuan-Ming Hsu, Anders Humlum, and Martin Rotemberg, September 2025.

Using our data?

We would love to hear about your work using the data from the Census of Manufacturing Project. Tell us about it here!

Data Use Survey
← Back