National Crimes: A New National Data Set of Lynchings in the United States,1883 to 1941

Abstract

Historians are increasingly studying lynching outside of the American Southeast, but sociologists have been slow to follow. We introduce a new public data set that extends existing data on lynching victims to cover the contiguous United States from 1883 to 1941. These data confirm that lynching was a heterogeneous practice across the United States. We differentiate between three different regimes over this period: a Wild West regime, characterized mostly by the lynching of whites in areas with weak state penetration; a slavery regime, found in former slave states, characterized mostly by the lynching of blacks; and a third minor regime, characterized by the lynching of Mexican nationals mostly along the Texas-Mexico border. We also note great variability at the county level in the extent of lynching. By contrast, we find very little state-level variability in lynching once local and regional regimes are considered. We discuss the implications of local and regional heterogeneity for quantitative lynching research using these data.

On June 1, 1909, Ida B. Wells spoke in front of the National Negro Conference, a forerunner of the National Association for the Advancement of Colored People (NAACP). She began by outlining the three major points she would make. “First,” she noted, “lynching is color-line murder. Second: Crimes against women is the excuse, not the cause. Third: It is a national crime and requires a national remedy” (Wells 1909). In the 110 years since Wells’s speech, sociologists have confirmed and extended both of her first two points. Historians have increasingly embraced Wells’s third point, that lynching was a national, and not simply Southern, problem. Sociologists, however, have not kept pace, and the development of a national-level lynching data set has been one missing link (for a discussion, see Pfeifer 2014:842–43).

Our goal for this article is to introduce and describe our newly collected and publicly available data, which extend previous data covering the Southeast, to cover the contiguous United States from 1883 to 1941. Following Tolnay and Beck (1995), we began our data collection with contemporaneous lynching inventories from the NAACP and Chicago Tribune. We also drew on published lists of lynching victims from historians (Barrow 2005; Campney 2008, 2018; Carrigan 2004; Carrigan and Webb 2013; Downey 2013; Downey and Hyser 2011; Frazier 2015; Pfeifer 2011, 2013; Williams 2001). Our data include every lynching victim we were able to locate and confirm with local newspapers from 1883 to 1941 in the 38 contiguous U.S. states not included in the original Tolnay-Beck data. We coded each lynching for the size of the mob, the race and gender of the victim, the alleged offense claimed to have incited the mob, and the method by which the mob murdered the victim. These data are available on the first author’s website, www.charlieseguin.com. Below, we first discuss our data collection methodology, describe our data, and discuss the promises and pitfalls of using these data.

Data Collection

There is a long sociological tradition of studying lynching in the South, dating back to the lynching era itself (Cutler [1903] 1969; Raper 1933). Ida B. Wells was the first to use lynching statistics, collected by the Chicago Tribune, to show that, even taking the stated motivations of lynch mobs at face value, rape and sexual assault were not the primary motivations for lynching, as lynching apologists had claimed (Wells-Barnett 1895). Monroe Work (1913) took Wells’s analysis a step further, showing that although around 20 to 25 percent of lynchings of black men were ostensibly about rape, only about 2 percent of blacks incarcerated for “major offenses” were charged with rape, and many groups of European immigrants were incarcerated for rape at much higher levels, suggesting that whites were afforded the protections of legal system when accused of rape whereas blacks were not (Work 1913:75–76). Researchers at the Tuskegee Institute, and the early NAACP under W. E. B. DuBois, also began to collect lynching statistics in the early 1900s (see Figure 1).

Figure 1.

National Association for the Advancement of Colored People (NAACP) lynching map 1889 to 1921.

Until Tolnay and Beck (1995) undertook their project to validate lynching statistics for the American Southeast, the NAACP, Chicago Tribune, and Tuskegee Institute were the best sources of data, both nationally and for the Southeast. Tolnay and Beck showed that these original data sets had significant errors and created a new data set consisting only of lynchings that they could verify with local newspapers. This data set has since been used in a diverse array of research on lynching in the American South (e.g., Messner, Baller, and Zevenbergen 2005; Stovel 2001; Tolnay, Deane, and Beck 1996). Our data collection is meant to extend the Tolnay-Beck data set beyond the 10 original Southern states (Cook 2012:59) in the hopes of spurring lynching research in other regions and at a national scale. ¹

To ensure similarity to Tolnay and Beck’s (1995) data, we follow the common academic definition of lynching as (1) an extrajudicial killing, (2) motivated or justified by reference to “justice” or tradition, and (3) committed by three or more people. Loosely following Tolnay and Beck’s original strategy, we began with the Chicago Tribune and NAACP inventories and attempted to confirm those lynchings with local papers. Unlike Tolnay and Beck, we did not have access to the Tuskegee database. Also unlike Tolnay and Beck, we supplemented these contemporaneous data sets with those published by historians (Brundage 1993; Capeci 1998; Carrigan 2004; Carrigan and Webb 2013; Frazier 2015; Gilbreath 2002; Gonzales-Day 2006; Leonard 2002; Pfeifer 2013; Segrave 2010; Tórrez 2008; Yost 1933). We then used nine searchable online newspaper archives to locate local news accounts that could confirm or disconfirm the lynchings in these data sets. ²

For each lynching recorded in any of our source data sets, we searched online newspaper archives for local papers to confirm the lynching. ³ If we could not find a paper within the same state reporting the lynching, we considered it unconfirmed, and it does not appear in our data set. ⁴ In many cases when we were not able to confirm a lynching, we were able to find a paper trail that suggests how the erroneous lynching event made it into the original lynching inventories. For example, both the Chicago Tribune and the NAACP list a lynching of a black man of unknown name in Millersburgh, Ohio, on April 1, 1982. We found, however, that

[t]he Marysville County Union Journal, had reported a lynching occurring on April 1 in Millersburgh of the “hardest looking man” that the county had ever seen. Two weeks later, on April 14, it was explained by the Mansfield Weekly News that the original lynching report was an April fool’s joke spurred by the “lynching” of an iron effigy, and that a Columbus correspondent had “bit” on the ruse and published a “lynching yarn.” (Anonymous, Mansfield Weekly News, April 14, 1892)

Once all of the lynchings in our source data had been confirmed or disconfirmed, the second author repeated the process to double-check all decisions.

Data

We combed through 2,850 records of lynching victims, ultimately confirming 1,319 victims and finding 15 lynching victims not recorded in any data set, for a total of 1,334 victims of lynching. Lynching spanned 33 states over this period, with only Maine, Massachusetts, New Hampshire, Rhode Island, and Vermont totally untouched. ⁵ While searching, we also found three lynchings within the Tolnay-Beck states, one of which was previously unrecorded and two of which were incorrectly recorded by the NAACP and Chicago Tribune as having occurred outside the South. There were also many cases for which our improved data allowed us to identify and name a victim who was lynched at a similar time and place as a victim listed by the NAACP or Chicago Tribune inventories as “unknown.”

When combined with the new Tolnay-Beck data (Beck 2015), we record 4,467 total victims of lynching from 1883 to 1941. Of these victims, 4,027 were men, 99 were women, and 341 were unknown gender (although likely male); 3,265 were black, 1,082 were white, ⁶ 71 were Mexican or of Mexican descent, 38 were American Indian, 10 were Chinese, and 1 was Japanese (see Figure 2).

Figure 2.

Map of lynching in the United States by race of victim. Interactive map available at http://davidrigbysociology.com/lynching_dot_map/.

Comparability with the Tolnay-Beck Data

Although we followed a data collection strategy similar to Tolnay and Beck (1995), comparing our data with their original data necessitates caution. First, even within our data, quality is uneven across states and regions. In some regions, the data quality is much better than in others because we were able to draw on detailed local histories. In Iowa, for example, we found six additional lynchings because they were included in Pfeifer’s (2013) regional inventory. In other places, our data collection suffered owing to a low density of newspapers, such as North Dakota. In places with few newspapers, lynchings are likely to be more undercounted both because lynchings would have been less likely to make it into the contemporaneous data sets that we drew upon and because we were less likely to find sources sufficient to confirm the lynchings that were in contemporaneous inventories.

Second, it’s likely that data quality differs systematically from the original Tolnay-Beck data because Tolnay and Beck (1995) relied on physical copies of local newspapers, and we used digital archives. It is not entirely clear to us whether our data would be more complete, or less complete, as a result of this difference. On the one hand, we were able to easily cast a wide net of all digitized papers in an area and simultaneously search nonlocal papers, which could help lead us back to local sources. On the other hand, we were limited to only those papers that had been digitized and made accessible. Thus, although we are not sure exactly what differences this might cause between our two data sets, it seems likely it has introduced some differences.

Third, Tolnay and Beck (1995) focused on 10 states that were more or less homogenous relative to the states within our data. Within the Tolnay-Beck data, it is reasonable to assume that the process through which a lynching enters the historical record is similar between, say, Alabama and Mississippi. In our data, however, it is unlikely that a lynching in Chicago, Illinois, has nearly as many barriers to entering the historical record as one in rural Georgia. It is also unlikely that the kinds of frontier lynchings in Montana entered the historical record at the same rate as the lynchings of Mexican nationals along the Texas-Mexico border.

Finally, Tolnay and Beck’s (Beck 2015) new data set includes not just online searches for specific lynchings found in the original historical data sets but also keyword searches for lynching in general. Because of their more expansive methodology used to find lynchings, these data likely have more complete coverage than ours. Users of our data will have to bear in mind these limitations. Whether the data-generating processes are similar enough to facilitate statistical comparisons will depend on a number of factors specific to research designs and questions.

Comparison with the Tuskegee Inventory

Ultimately the usefulness of this data set depends on how well it captures the underlying reality of lynching relative to other, alternative data sets. At the national level, there are three data sets that have attempted to be comprehensive: the Chicago Tribune, NAACP, and Tuskegee Institute inventories. We started our data collection with the Chicago Tribune and NAACP inventories and found many new lynchings that were not in either data set as well as uncovered documents disconfirming many of the lynchings in those inventories. As such, we are confident that our data set provides a superior alternative to the NAACP and Chicago Tribune inventories. Since we did not include the Tuskegee inventory in our original data collection, this provides us with an opportunity to compare our data set with one of the contemporaneous data sets.

We compare our data with the Tuskegee Institute’s lynching inventory for the state of Texas. Texas provides an excellent point of comparison for two reasons. First, Texas has the most cases, with 438 lynching victims in our data. Second, Texas is a particularly complex state in its lynching history since it contained lynchings of blacks in the cotton-planting region, lynchings of whites in the western frontier region, and lynchings of Mexicans on the Texas-Mexico border. Thus, short of comparing the entire data sets, Texas provides a more thorough comparison than any of the other states in our data.

Overall, we find that our data compare well with Tuskegee’s. We found that our data contain 44 confirmed lynchings that are not in the Tuskegee data. Moreover, we find that the Tuskegee inventory contains 50 lynchings where we have documentation strongly suggesting that these were not, in fact, lynching victims. For instance, both the NAACP and the Chicago Tribune list an entry for a “William Black” lynched in Texas. However, we found a newspaper article showing that the “lynching” was a rumor beginning with a newspaper publishing a racist joke about “Billy Black,” a billy goat that had supposedly been lynched (see Figure 3). Additionally, the Tuskegee data contain five lynchings recorded as occurring in Texas but that our documents suggest actually occurred in Kentucky or Tennessee. ⁷

Figure 3.

Clipping from the Austin Weekly Statesman, December 3, 1891.

The comparison also pointed to limitations and ways to improve the data set in the future. First, the Tuskegee inventory contains six lynchings that our new records show clearly did occur but that we originally failed to confirm. These lynchings were also in the NAACP or Chicago Tribune records, so we had searched for them but did not find documentation sufficient to confirm them. Although we cannot rule out that we simply erred in our attempts to confirm these lynchings, we suspect this difference is due at least in part to the substantial expansion of available searchable records between when we first started data collection (September 2014) and when we conducted this last check (March 2019). Our most-used and most-effective source of local papers, Newspapers.com, for instance, had 77 million archived pages of newspapers when we began and 475 million when we finished and went from around 3,200 to 11,700 newspapers, respectively, representing a sixfold increase in pages and a 3.5-fold increase in newspapers. These six lynchings represent roughly 1.4 percent of the total 438 lynchings in Texas, so taking these results at face value, we might expect to find something in the neighborhood of 12 or 13 additional lynchings in the states other than Texas (1.4 percent of the non-Texas lynchings in our data) if we were to make a third pass at checking our data at the time of this writing. Given that records are continuing to expand, however, doing so would likely be far more profitable 5 or 10 years from now. Given that all the missing lynchings were also in the NAACP or Chicago Tribune records, comparison with Tuskegee does not appear to be the path forward here.

Finally, the Tuskegee data contain an additional 58 lynching records that we did not include because it was not clear from our searches whether these lynchings occurred or, if they did, whether they met our definition of lynching. These were lynchings where reports are based on rumor, there are too few details to confirm, reports seem to be redundant with other records, or we were unable to find any local reporting to confirm the lynching. As more historical records become available, we may be able to settle many of these lynchings; for the present, however, this discrepancy points to an important limitation of data collection when historical records are limited and to our practice of being conservative in confirming lynchings when in doubt.

Lynching Regimes

Over a century of scholarship has shown us that American lynching was a different practice in different times and places. In 1909, for example, Ida B. Wells sketched this variation in the broadest strokes, outlining two different lynching “regimes,” one in the South and the other in the West:

[W]hile frontier lynch law existed, [lynchings] showed a majority of white victims. Later, however, as law courts and authorized judiciary extended into the far West, lynch law rapidly abated and its white victims became few and far between. Just as the lynch law regime came to a close in the West, a new mob movement started in the South.

In the time since Wells, scholars have collectively tackled much of the variation in the practice of lynching across time and place in a piecemeal fashion, focusing on one or two regions (e.g., Carrigan 2004; Rigby and Seguin 2018; Tolnay and Beck 1995) or states (e.g., Brundage 1993; Campney 2018), or have taken the broad sweep of lynching in qualitative terms (e.g., Dray 2007; Waldrep 2002). Taking these studies as our starting point, here we use our data to sketch the broad outlines of variation in the practice of lynching in the United States from 1883 to 1941, drawing primarily from visualizations of our data over time and space.

Regional Lynching Regimes

Our data show that lynching was characterized by two major regional regimes and one minor regime. The first major lynching regime was in the “Wild West,” was caused by weakness of the state, and declined as a function of the increasing penetration of state law enforcement (Pfeifer 2011). The majority of victims subject to the Wild West lynching regime were white. This regime can be seen in Figure 2, mostly through the Dakotas, Nebraska, Kansas, New Mexico, and the states to the west of these. These lynchings declined faster over our period, as can be seen in Figures 4 through 6: In 1883, there were marginally more white victims than black, but lynchings of whites declined rapidly beginning in 1894, becoming a small fraction of total victims by 1900. Seventy-six percent (251) of the lynching victims in this frontier area were white, 11 percent (36) were black, 6 percent (19) were American Indian, 4 percent (13) were Mexican nationals or of Mexican descent, and 3 percent were Asian (8 Chinese and 1 Japanese victim).

Figure 4.

Map of black lynching, 1883 to 1941, and slavery in 1860. Interactive map available at http://davidrigbysociology.com/lynchings_of_black_victims_and_slavery/.

Figure 5.

Lynchings in the United States by race and year.

Figure 6.

Lynchings in the United States by region and year.

The second major regime was in the former slave states. In contrast to the decline in the lynchings of whites in the Wild West regime, lynchings of blacks grew more common until 1893 before beginning to decline. After roughly 1900, and increasingly so over time, lynchings were almost exclusively of black victims and occurred in the former slave states. Thus, this lynching regime was, as Wells termed it, “color-line murder,” where most of the victims were black and where black victims were sometimes subject to heinous tortures, such as being burned alive in a “gruesome tribute . . . to the color line” (Wells 1909). Over the entire period, 60 percent of nonblack lynchings and 97 percent of black lynchings occurred in states that had slavery at the outset of the Civil War, accounting for 88 percent of total lynching victims. Overlaying the map of black lynching victims with the extent of slavery in 1860 reveals the basic contours of this regime (see Figure 4). Black lynchings cluster in areas that had high rates of slavery in 1860. Much of this regime followed from the spread of cotton agriculture. The section of southeast Texas where black lynchings are clustered, for instance, marks the western edge of the expansion of cotton planting up to this period. This regime was not entirely confined to cotton agriculture, however: black lynchings in Missouri clustered among the counties known as “Little Dixie” along the Missouri River watershed, where hemp and tobacco were the common crops and where slavery was also prevalent. Campney (2019:chap. 5) shows how racial violence in Little Dixie, including lynching and whippings, were practices borrowed directly from the region’s history of slavery. Oklahoma was not a state in 1860, so census data on the extent of slavery were not collected at the county level (O’Connell 2012:722), but there was slavery in Oklahoma, and the southeastern corner where black lynchings are clustered was also the edge of the cotton belt. It is not surprising that lynching followed the contours of slavery for any number of reasons, especially given that even as late as 1910, 89 percent of blacks lived in former slave states, but it is striking to note that almost all—97 percent—of lynchings of black victims occurred in slave states, and even within those states, lynchings appear to be most common in places with a larger slave population in 1860.

Together, the Wild West and slave-state regimes encompass the vast majority of lynchings. An additional, minor regime was the lynching of Mexican nationals on the Mexico-Texas border. We confirm 438 lynching victims in Texas during our period. This state total is third in the nation, trailing only Georgia (471) and Mississippi (606). Of those lynched in Texas, 70 percent (307) were black, 17 percent (75) were white, and 13 percent (56) were Mexican. Of the 56 Mexican lynching victims in Texas, 40 were lynched in five counties along the Mexican border. Of these 40 lynchings, 25 occurred in 1915, during the Mexican Revolution, amid concern by Anglo-Texans over a series of raids by Mexican rebels. In addition to these lynchings, during this period, numerous massacres of Texans of Mexican descent were carried out by the Texas Rangers and the U.S. military (Carrigan and Webb 2013; Martinez 2018); since they were committed by agents of the state, these murders do not fit our criteria for lynching.

Local Lynching Regimes

Broad differences between regions explain much of the variation in lynching throughout the United States; however, these regions also contain a large amount of variation within them. Forty-one percent of counties (648) in slave states, for instance, did not witness a single lynching in our data, whereas multiple black victims were lynched in places like Duluth and Chicago. Among slave-state counties, the total number of victims ranges from 0 to 37 (in Bossier Parish, Louisiana). Overall, less than 18 percent of counties (553) accounted for 80 percent of lynching victims, and less than 7 percent of counties accounted for half of all lynching victims.

Differences between counties occurred for a number of reasons. In the U.S. South, lynching was more common in counties that planted cotton (Tolnay and Beck 1995) and was more common when cotton prices were low (Beck and Tolnay 1990). County-level lynchings were also affected by the structure of the farming economy, such as the number of whites engaged in tenant farming, but that relationship appears to be complex and time varying (Tolnay et al. 1996:808). Southern counties with more religious diversity, black-controlled churches, and more racially segregated churches all experienced more lynchings (Bailey and Snedker 2011). The incidence of public-torture lynchings was a function of white racial solidarity at the county level in Georgia and Louisiana (Smångs 2016). Southern lynchings were more common in counties where Populists challenged existing political orders (Inverarity 1976; Olzak 1990; although see Soule 1992). Finally, lynching was less common, net of other factors, when lynchings had recently occurred in bordering counties (Tolnay et al. 1996).

Lynching was less common in counties where local law enforcement was more proactive in suppressing lynch mobs (Beck 2015; Hagen, Makovi, and Bearman 2013). In some cases, differences in law enforcement between counties can be reduced to minute historical contingencies, such as the personality or skill of the county sheriff or other actors (Franzosi, De Fazio, and Vicari 2012; Griffin 1993:1111–12), making lynching to some extent an irreducibly local phenomenon. Nevertheless, law enforcement and other actors were also subject to systematically varying political pressures. Specifically, in places where the local economy was affected by its national and international reputation, local elites were more responsive to the “bad advertising” resulting from media coverage of local lynchings (Bernstein 2005; Campney 2013; Wood 2009). Moreover, places that were more linked to the national economy were more likely to receive negative media attention to lynchings in the first place (Weaver 2019). Accordingly, it was in the places most integrated with the national economy that lynching declined the fastest (Beck, Tolnay, and Bailey 2016). Given this long list of both idiosyncratic and systematic causes of lynching at the county level, it is perhaps not surprising that there is so much variance in lynching at the county level.

State-level Lynching Regimes?

In contrast to local (county) and regional (slavery/Wild West) lynching regimes, the data do not show much evidence of meaningful state-level variation. One reason we believe we do not observe great differences between states is that bordering states are generally similar to one another along the important dimensions. For example, the 14 states with the most total lynchings over this period form a single contiguous cluster in the Southeast, defined by Texas, Oklahoma, Missouri, Kentucky, Virginia, and all states to their south and east. This cluster includes only slave states and is missing only the small slave states: West Virginia, Delaware, and Maryland. Where one does find dramatic differences between nearby states, such as Iowa and Missouri, or between Virginia and Pennsylvania, the difference is slavery rather than state-specific differences. Moreover, within states, multiple regional and local logics operated. Eastern Texas, where cotton and slavery had spread before the Civil War, looks much like the rest of the Southeast, while south Texas along the Mexican border had a very different dynamic (Carrigan 2004; Carrigan and Webb 2013).

An implication is that state-level lynching counts should not generally be used as an independent or dependent variable. Consider, for instance, Nebraska and Maryland: These states each had 27 confirmed lynchings. But in Nebraska, 22 victims were white and 3 were black, and in Maryland, 25 were black and 2 were white. An obvious way to deal with this would be to restrict measurement to only the lynchings of whites or blacks; however, because 97 percent of black lynchings occurred in slave states, it is unlikely that the count of black lynching victims can be meaningfully differentiated from the state’s experience with slavery. Finally, any state-level count will gloss over an incredible amount of local-level variation. Naturally, there will be exceptions, but our general recommendation is to avoid the use of state-level lynching counts as a variable.

Discussion and Conclusion

As historians have increasingly moved the study of lynching “beyond Dixie,” sociologists have been late to the study of lynching outside of the South. We hope that the data presented here will spur more such work. At the same time, sociologists should be careful when using these data and not uncritically consider these data to represent a unitary phenomenon across the country.

Ultimately, we will never know the full and exact extent of lynching in the United States, as many lynchings are undoubtedly lost entirely to the historical record. As such, our data set remains a work in progress. A number of extensions to these data would be welcome. First, extending our lynching data both further into the past and closer to the present would be most useful. It is clear that lynching began to be common, particularly in the West, long before the Civil War and shortly after emancipation in the U.S. South (see Pfeifer 2014 on the need for additional data). The data we have now tends, then, to miss how the practice initially developed in the West. Lynching also continued past 1941, into the civil rights era. While there were far fewer lynchings then than at the peak of the lynching era, much might be learned about the places where the practice survived the longest. Second, taking an inductive approach to searching for lynchings, by reading articles turned up from keyword searches, would likely turn up additional lynchings. Third, given that digital historical records are continuing to expand at a rapid pace, simply repeating this data collection at a later date would likely also turn up additional lynchings. Finally, scholars have recently noted that for questions regarding the motivation or cause of lynching mob formation, attempted lynchings, rather than only completed lynchings as we collected here, may be the variable of interest and thus worth extending these data to include (Beck 2015; Hagen et al. 2013; Makovi, Hagen, and Bearman 2016; Tolnay and Beck 2018).