This project is ongoing. This page is just a demo.
The Dataset of U.S. Lynching Reports (DUSLR)
Figure 1: A Map of 4482 Documented Lynchings in America (1865 to 1921) – Compiled from the Tolnay-Beck-Bailey Inventory and the Seguin-Rigby Dataset. Locations are narrowed to the county, not the precise locations.
THE FOLLOWING TEXT IS OUTDATED AND NEEDS TO BE REVISED
Newspaper reports about lynchings and data about lynchings in the United States have long been mutual components in each other’s production. When Ida B. Wells first presented her computation of lynching statistics in A Red Record (1895), she drew her data from lynching reports published by the Chicago Tribune. When Monroe Work compiled lynching data in the early 1900s, he did so with a goal of distributing the data to more newspapers across the country. Without newspaper reports about lynchings, we would have very little data on the subject. Without the dissemination of lynching data, the painful legacy of lynchings in America would be harder to examine and respond to–both culturally and legislatively, since lynching data has long been used as evidence of the pervasiveness of the issue and as proof of the necessity for broader anti-lynching legislation.
As part of the Virality of Racial Terror project, we reassessed this relationship between newspaper reports and data about lynchings by text-mining the Chronicling America digitized newspaper archive for known lynching victims. We compiled their names from two data sources–the Tolnay-Beck Inventory and the Seguin-Rigby dataset. Both datasets built on the previous work of the aforementioned scholars and activists, and numerous other sources, containing details like names, places, alleged offenses, and more. Using these data in a text-mining pipeline that included multi-step database queries, numerous preprocessing steps, and fine-tuning a BERT-base language model, we were able to identify tens of thousands of newspaper reports about lynchings between 1865 and 1921. For a detailed breakdown of these computational processes, see our GitHub Repository.
This new dataset–the Dataset of U.S. Lynching Reports–is deeply indebted to the tradition of scholarship and activism dating back to the nineteenth century. It is our hope that this new dataset foregrounds further research on the painful legacy of lynchings in the United States. It is a relatively large sample of the newspaper reports that have always played a fundamental role in our nation’s understanding of lynchings–reports that were read by Americans across the country contemporaneously with these horrific events, reports that were used to verify and compile data on the issue, reports that were subsequently fundamental to anti-lynching efforts despite the fact that so many of these reports are not sympathetic to lynching victims or the anti-lynching cause.
Below we provide the data in a dashboard with maps showing where the reports were printed as well as tables that allow users to review individual clippings. Before reviewing the dashboard, however, we must warn users that these clippings are often offensive, cruel, and violent in detail. Some are brief with just one or two telegraphed sentences that identify place, victim, and alleged offense. Others are more detailed in their reporting of the brutality and violence of these events. Many focus on the victim’s alleged offense and clearly have a racial bias that seeks to excuse lynchings. Please be aware that this data has not been censored in any way. It is composed entirely of newspapers from the era which are often racist, offensive, and inaccurate or biased in their reporting.
Statistical Overview
A detailed breakdown of our data curation processes–including code–can be viewed here. But to summarize the contents of the Dataset of U.S. Lynching Reports, here are some basic facts about the data:
- 51,907 clippings with a ‘high’ probability that they reference a lynching or lynchings
- clusters of clippings that reference 3,084 Black victims of lynchings
- clusters of clippings that reference lynchings that occurred in 32 states between 1865 and 1921
- 2,300 newspapers total–newspapers from every state in the country–are included in the data
It must also be noted that this data is not comprehensive of lynchings that occurred during the period–not even close. It is limited by several factors, including 1) the limited number of American newspapers that have been digitized and added to the Chronicling America archive, 2) the relative notoriety of cases, meaning that it only includes lynchings that were reported in newspapers, and 3) the cases that occurred within the period of 1865 and 1921. The data also only represents lynchings where the victim’s race was identified as Black. We chose to focus on lynchings of Black victims because, in these cases, lynchings were often mechanisms of racial terror, meant to reinforce white supremacy in the United States, making them particularly important for the larger Virality of Racial Terror project.
Reviewing the Data
The following dashboard includes a dropdown menu of victim names. Once a victim name is selected, it will present an interactive map of the newspaper reports we’ve identified as likely having to do with the victim’s lynching. The dashboard also presents a table view of these reports, allowing users to review individual clippings by newspaper and date of publication.
The dashboard is designed to emphasize the relationship between newspaper reports and data. We also hope that it can facilitate preliminary research on both individual cases and the pervasiveness of lynchings on a national scale. If you’d prefer to download the data in bulk and review it in other ways, please contact us.
Reviewing Cases
Reviewing Cases by State
The table below is another way to review the newspaper reports. Click any row to view an approximated clipping of the corresponding report. Please note these reports were computationally identified. In rare instances, they may not correctly reference the given case. The image versions are also computationally identified through a different process. They estimate the boundaries of the newspaper clippings. This means they occasionally appear in difficult formats to read in this dashboard. Use the “Full page URL” below the image to open the complete newspaper page in Chronicling America. This will allow you to review the full page where partial or unreadable clipping images were pulled from.