klan feb 21 2011 Hatred and Profits: Under the Hood of the Ku Klux Klan* Roland G. Fryer, Jr. Harvard University and NBER and Steven D. Levitt University of Chicago and NBER February 2011 * We are grateful Edward Glaeser, Claudia Goldin, Lawrence Katz, Robert Margo, and Peter Temin for helpful comments and suggestions. The authors thank Bowling Green University, Clark Historical Library, Colorado Historical Society, Emory University, University of Georgia, Harvard University Libraries, University of Kentucky, the Indiana Historical Society, the Montana Historical Society, and the Pennsylvania State Archives for their help in constructing our database. Paras Bhayani, Mandy Ching, Steve Cicala, Sean Garborg, Ethan Lieber, Alipo Melusi, Carter Mundell, and Lisandra Rickards provided exceptional research assistance. Correspondence can be addressed to either of the authors at (Fryer) 1805 Cambridge Street, Littauer Center – North Yard, Harvard University, Cambridge MA 02138 [rfryer@fas.harvard.edu]; (Levitt) 1126 East 59th Street, University of Chicago, Chicago IL 60637 [slevitt@uchicago.edu]. Abstract The Ku Klux Klan reached its heyday in the mid-1920s, claiming millions of members. In this paper, we analyze the 1920s Klan, those who joined it, and the social and political impact that it had. We utilize a wide range of archival data sources including information from Klan membership roles, applications, robe-order forms, an internal audit of the Klan by Ernst and Ernst, and a census that the Klan conducted after an internal scandal. Combining these sources with data from the 1920 and 1930 U.S. Censuses, we find that individuals who joined the Klan in some cities were better educated and more likely to hold professional jobs than the typical American. Surprisingly, we find few tangible social or political impacts of the Klan. There is little evidence that the Klan had an effect on black or foreign born residential mobility, or on lynching patterns. Historians have argued that the Klan was successful in getting candidates they favored elected. Statistical analysis, however, suggests that any direct impact of the Klan was likely to be small. Furthermore, those who were elected had little discernible effect on legislation passed. Rather than a terrorist organization, the 1920s Klan is best described as a social organization with a wildly successful multi-level marketing structure fueled by an army of highly-incentivized sales agents selling hatred, religious intolerance, and fraternity in a time and place where there was tremendous demand. 1 I. Introduction The Ku Klux Klan (KKK) is the most prominent hate-based organization in American history. Founded in the aftermath of the Civil War as a whimsical social club, the Klan quickly transformed into a terrorist organization aimed at subjugating newly freed blacks and driving out moderate whites that attempted to improve the plight of Freedmen in the Reconstruction South. Legal and military action undermined the effectiveness of the Klan, and by the 1870s it had largely disappeared. The Klan was revived in 1915. Its popularity and influence grew, first slowly, and then more rapidly, to a peak in the mid-1920s. By that time it claimed four million members, was credited with engineering the election of politicians across the country, and included in its membership some of the most powerful men in America.1 Marred by a high profile sex scandal, the Klan’s membership rolls decreased dramatically in the latter half of the 1920s and the Klan has since remained a shadow of its former self. In this paper, we analyze the rise and fall of the Klan during its peak years in the 1920s, attempting to answer a wide-ranging and ambitious set of questions concerning the organization: what explains the stunning rise and precipitous fall of the Klan? Was it a hate-based terrorist group, or merely a social club like the Elks or Veterans of Foreign Wars? Who joined the Klan, and why did they do so? To what extent did the Klan’s rise and fall have real effects, for instance, in the form of increased violence, migration of Blacks and Catholics away from Klan strongholds, or on electoral/legislative outcomes? These are difficult questions to answer, even under ideal circumstances. In this particular setting, the task is made more challenging both because nearly a century has elapsed, and also, the Klan was a highly secretive organization 1 It is alleged that President Warren Harding was inducted into the Klan in a ceremony performed at the White House (Wade 1987), though there is disagreement among historians on this point (Newton 2006). 2 about which information is scarce. Consequently, the answers that we offer are in many cases suggestive rather than conclusive. The key input into our analysis is a unique dataset on membership in the Ku Klux Klan in the 1920s that we have constructed based on archival research. The datasets contains data on nearly 60,000 actual Klan members whose names we gleaned from internal Klan documents including membership lists, completed application forms, meeting attendance sheets, signed correspondence between members and officials, and order forms for robes and other paraphernalia. Our research uncovered Klan records of this type for a number of different areas, including membership records for the states of Pennsylvania and Colorado, as well as cities including Knoxville, Tennessee and Athens, Georgia. By merging these Klan records with individual-level data from the 1920 and 1930 censuses, we are able to address the question of who joined the Klan. Second, based both on the documents contained in these archival sources and the existing historical record, we are able to paint a more complete and data-driven picture of the economic structure of the Klan than was previously available. For instance, using credible estimates of the number of Klan members in Indiana between 1923 and 1925 from two sources (an internal audit performed by Ernst & Ernst and a census of Klan members), as well as information on dues and other sources of revenue, we estimate the revenues that high-ranking individuals within the Klan likely earned. Third, we use county-level data from Indiana and Pennsylvania, two Klan strongholds during the 1920s, on the number of Klansmen in a given county to analyze the extent to which variation in Klan membership influenced real social, economic, and political outcomes. A 3 dramatic historical accident – the sordid murder conviction of the leader of the Indiana Klan which precipitated the Klan’s implosion within two years – provides a natural experiment for exploring the causal impact of the Klan. It is unlikely that the nativist sentiments underlying the Klan dissipated as quickly as the organization itself, potentially allowing us to separately identify the role of the organization from the social forces that spawned it. The results we obtain using our new dataset on members of the Klan are, in some cases, quite surprising. In some cities, individuals who joined the Klan were better educated and more likely to hold professional jobs than the typical American. In other cities, however, Klan members were significantly less likely to hold professional jobs despite being better educated, providing mixed evidence for earlier findings of a number of historians on the subject (Goldberg 1981). Yet, despite the sophistication of its members and its enormous rolls, we find that the Klan had few tangible social or political impacts. There is little evidence that the Klan had an effect on black or foreign-born residential mobility or on lynching patterns. There is some evidence that the Klan was successful at getting candidates it favored elected, but the direct impact of the Klan in this domain appears small. Moreover, even when the Klan succeeded in the electoral process (e.g. Indiana in 1925 had a Klan-endorsed governor, as well as control of the state legislature), there is little evidence that the legislation passed effectively advanced the Klan’s mission (Jackson 1992, Goldberg 1981). Instead, the Klan’s true genius lay in its remarkable ability to raise revenue. We estimate that at the peak of the Klan, initiation fees, dues, and profits from robes in the state of Indiana alone generated nearly $4.4 million (in 2011 dollars) annually for the national Klan leader, $2.6 million for the head of the Indiana Klan, and over $330,000 each for the national head salesman 4 and the salesman responsible for Indiana. Per capita income in the United States at this time was roughly $8,700 in 2011 dollars. Rather than a terrorist organization, the 1920s Klan is better described as a wildly successful multi-level marketing entity fueled by an army of highly incentivised sales agents and an unprecedented interest in fraternal groups of all kinds (Skocpol 2003). Our findings fit squarely into Glaeser’s (2005) pioneering analysis of “entrepreneurs of hate,” in which political and business leaders create and encourage hatred for private benefit. With stunning rapidity, however, the Klan collapsed as quickly as it had grown. By 1930, its national membership had declined to perhaps one percent of that achieved at the peak (Moore 1991, Wade 1987). The implosion of the Klan stands in stark contrast to other fraternal groups of the time, many of which still function on a large scale nearly a century later (e.g Elks, Rotary club). While there were idiosyncratic factors that contributed to this collapse (most notably the murder conviction of a prominent Klan leader), we argue that even absent these extenuating circumstances, the Klan was likely to be unstable, both because of the structure of the incentives given to Klan recruiters and because of its hate-based rather than civic-minded nature. The next section provides a brief history of the Ku Klux Klan and explores the peculiar nature of the organization. Section III describes how we construct our database of Klan members from archival data, summarizes the dataset, and provides some basic social and economic demographics of Klan members. Section IV estimates the revenues that accrued to key Klan officials from the state of Indiana during the peak years of membership. Section V attempts to understand the rapid decline of the Klan in the late 1920s. Section VI estimates the effect of the Klan on lynching, black and foreign-born migration, and politics in Indiana and Pennsylvania. 5 Section VI concludes. There are two appendices: Appendix A provides details of how we constructed our database of Klan members and Appendix B is a guide to Klan terminology. II. A Brief History of the Ku Klux Klan In December 1865, six ex-Confederates from Pulaski, Tennessee (near the Alabama border) organized a social club in which the six, dressed in white sheets, rode around on horseback for amusement.2 They called themselves the Ku Klux Klan, merging the Greek for circle (kyklos) with “clan.” The founding members created the group as a secret fraternity replete with ridiculous names and costumes.3 The group soon discovered that their strange appearance frightened newly freed African-Americans (Chalmers 1987).4 The idea caught on among others, and similar autonomous Klans began to emerge across the south. Many of the Klan’s early activities were child-like mischief.5 In April 1867, representatives of the various—and heretofore independent—Klans met in Nashville, Tennessee (Chalmers 1987). The meeting resulted in the election of Nathan Bedford 2 Our discussion of the history of the Klan draws heavily on the excellent accounts provided in Chalmers (1987), Wade (1987), Moore (1991), Alexander (1965), Newton (2006), Newton (2001), Tucker (1991). 3 For instance, there was initially an elaborate initiation ceremony that consisted of blindfolding the candidate, subjecting him to lengthy oaths, and bringing him before a “royal altar” (a mirror) where he was invested with the “royal crown” (two large donkey ears). 4 Prior to the Klan, there had been only one similar organization in U.S. history: the American Party, better known as the “Know-Nothings.” The Know-Nothings were an actual political party whose beliefs were predicated on a strong opposition to immigration. The party was largely based on nativisim, not racism. The party arose durng a period of massive immigration and directed its fire at new immigrants from nations such as Ireland, which had experienced a horrendous Potato Famine in 1848. Because the subjugation of Blacks was not a concern before the Civil War—most were enslaved, after all—the Reconstruction Klan was the first major American organization whose primary objective was targeting Blacks. 5 In one favorite Klan tactic, a white-sheeted, masked Klansman would ride up to a Black home at night and demand water. When the well bucket was offered, the Klansman would gulp it down and demand more, having actually poured the water through a rubber tube that flowed into a leather bottle concealed beneath his robe. After draining several buckets, the Klansman would exclaim that he had not had a drink since he died on the battlefield at Shiloh, and gallop away. 6 Forrest, a former Confederate general from Memphis, as the Klan’s chief, or Grand Wizard. Less than two years later, however, Forrest ordered the dissolution of the Klan, arguing that it was “being perverted from its original honorable and patriotic purposes, becoming injurious instead of subservient to the public peace.” Though many local Klans did dissolve following Forrest’s orders, many others did not (Chalmers 1987). In response to continued Klan violence in the South, the Radical Republicans began a campaign to destroy the Klan, using U.S. Army troops to break up Klan activities and prosecuting Klansmen in federal courts. The Klan ceased operations by the mid-1870s. The Second Coming and the Height of Power For over forty years the Klan laid dormant. Around 1915, it is widely believed that the confluence of two events, the release of The Birth of a Nation and the lynching of Leo Frank led to the coming of the second Klan (Wade 1987, Newton 2006). In the summer of 1915, D.W. Griffith released a largely fictional documentary, The Birth of a Nation, which greatly enhanced the public opinion of the Klan. Griffith’s film was an adaptation of a book—The Clansmen: An Historical Romance of the Ku Klux Klan, by Thomas Dixon—that largely conformed to what was the commonly accepted historical record of the day (Wade 1987). The film’s second portion, which focuses on the Klan, portrays ex-slaves as criminals intent on raping white women and northern carpetbaggers as co-conspirators in Blacks’ attempt to subjugate white southerners. The film shows the Klan as an organization founded to empower and defend white southerners, particularly “white womanhood,” and willing to engage in “noble violence” to achieve its ends. The film, which some have claimed had the highest box 7 office gross of any silent movie, was endorsed by President Woodrow Wilson, a former historian and friend of Dixon. After a White House screening, Wilson commented: “My only regret is that it is all so terribly true” (Wade 1987). In September 1915, a Georgia physician named William J. Simmons saw Griffith’s film. Inspired, he led a group of 34 men, including two veterans of the first Klan, on a trip to Stone Mountain. On Thanksgiving Day, in what Simmons would later claim were sub-zero temperatures, the group inaugurated the second coming of the Klan. Simmons declared himself Grand Wizard (Newton 2006, Wade 1987). Within weeks, ninety-one new members joined Simmons’ Klan, bringing in $10 apiece in initiation fees and $6.50 apiece for robes and hoods; forty-two of them paid $53,000 worth of Klan life insurance. The insurance program, characteristic of most adult fraternal orders at the time, provides some anecdotal evidence that the Klan was designed, in part, as a money-making device (more on this in section IV). Simmons has been described as a “good lodge man” who held membership in six other fraternal organizations when he started the Klan (Alexander 1965). According to Simmons, the new Klan was designed to be respectful and harmless; Klansmen were to refrain from coercive activity except to frighten an occasional “uppity” Negro. In many respects, the Klan resembles the many other fraternal orders that were thriving at the time.6 Over the next several years, the Klan spread, though the overall membership remained relatively. In June 1920, Simmons struck a deal to convert the Southern Publicity Association – an organization that had conducted campaigns for the Salvation Army and others – into the “Propagation Department” of the Klan. The contract stipulated that managers of the Propagation 6 The Klan’s charter describes the group as a “…benevolent, ritualistic, social, and fraternal order” (Ku Klux Klan 1921). 8 Department, Edward Clarke and Elizabeth Tyler, would get $8.00 of every $10.00 initiation fee. They sent over 1,000 Kleagles into southern and southwestern states with the charge to make salient whatever prejudices – anti-Catholocism, anti-Semitism, racism, and so on – were most acute in the particular areas they were selling Klan memberships. Only then did membership truly accelerate, and at a mind boggling rate. By 1924, estimates of the Klan’s peak size range from 1.5 million (Jackson 1992) to as many as four million, or roughly 5-14% of the eligible population.7 The largest estimates of membership are based on claims of Klan leaders, which are likely to be inflated. Figure 1 depicts this dramatic rise.8 A simple adaptation of the pioneering work in Glaeser (2005) may help explain the rapid rise of the Klan. In Glaeser’s model, there is an interaction between politicians (or social entrepreneurs in the case of the Klan) supplying hatred against an out-group and the willingness of in-group members to demand such hatred. The size of the in-group is fixed. If we make the size of the in-group endogenous and allow entry decisions to depend on the costs and benefits of group membership, one can rationalize the rise in Klan membership. The costs of joining the Klan likely included membership fees and potential social sigma from some segments of society. Benefits likely included police and personal protection, networking, and so on. These benefits 7 While these self-reported Klan numbers are likely an exaggeration, they are not completely out of the realm of possibility. An analysis of data from an internal census of Indiana Klan membership in 1925 finds that a total number of Klansmen is 162,267 or 18.44% of the eligible Indiana population. Klan penetration into Indiana is reputed to have among the highest in the country, so extrapolating this percentage to the whole of the United States will yield a (potentially very loose) upper bound of 5.2 million Klan members; using the Colorado Klan as a benchmark yields even higher numbers. If instead we extrapolate from the membership rolls of the Pennsylvania we obtain estimates of approximately 400,000 8 While the Klan’s rapid rise and fall make it rare among membership organizations, the years in which it hit peak operations are almost exactly those in which similar cross-class, fraternal organizations reached their apex (Skocpol (2003)). Given that the Klan was part of a broader trend of civic engagement, one interesting question is why these groups thrived in these particular historical periods. Skocpol (2003) argues that the civil war and World War I had important impacts on views regarding mobilizing citizens under common civic causes. Of course the motivations of the founders of the Klan were not civic in the conventional sense—they were fueled not by a sense of progressivism but by what they considered affronts to southern and national dignity. 9 may have increased with the number of individuals joining the Klan (or the social stigma decreased), leading to a social multiplier effect. While we do not have definitive evidence, some combination of the sales force and profit motive of Klan leaders, nativist sentiment, and the explosion of fraternal orders may have lowered the costs of Klan membership and increased the benefits – spurring meteoric growth in the Klan in the early 1920s. The organizational structure of the Ku Klux Klan of the 1920s, designed by the Propagation Department, was a hybrid that combined features of other fraternal orders with a multi-level marketing firm, with two distinct sets of reporting hierarchies that operated more or less independently. One hierarchy was made up of the Klan’s members, from the lowliest rank and file to the highest leadership. This hierarchy corresponds to the social club aspect of the Klan—the arm that intimidated Blacks and foreigners and attempted to influence political outcomes. In addition, however, there was a nearly invisible, parallel hierarchy of Klan recruiters, organized like a modern multi-level marketing firm, which represents the financial arm of the Klan. This highly incentivised sales force was responsible for recruiting new members to the Klan, and almost all of the financial rewards accrued to either the handful of top leaders or the individuals in this auxiliary hierarchy. The structure of the main portion of the Klan is presented in Figure 2. The highest level of the organization was the Empire, which governed the Klan’s national operations. The Grand Wizard (or Emperor) served as the nominal chair of the body, with the Imperial Wizard acting as the chief executive and aided by a fifteen-member Imperial Kloncilium. These included the Klaliff (first vice president), the Klazik (second vice president), the Klokard (lecturer), the Kludd (chaplain), the Kligrapp (secretary), the Klabee (treasurer), the Kladd (conductor), the Klarago 10 (inner-guard), the Klexter (outer-guard), the Klonsel (general counsel), the Night Hawk (courier), and the four Klokann (auditors). These individuals were responsible for keeping the Klan’s books, providing in-house legal advice, and serving as a Klan cabinet. (Newton 2006) The next tier of the organization consisted of the Realms, or the states, which were overseen by Grand Dragons and their staffs. The Realms served as administrative centers that recorded and processed new members. They also coordinated political activities within the states; in places of great Klan strength, such as Indiana, the Grand Dragon often wielded great political power. Reporting to the Grand Dragon was a fifteen member cabinet composed of Hydras, who held the same responsibilities as the Imperial Kloncilium, but at the state level. The most important positions were the Kligrapp, akin to a chief operating officer, and the Klabee, the treasurer. These officials handled much of the actual administration of the Klan (Newton 2006, Wade 1987). In the bottom tier of the organization, individual Klansmen, or Ghouls, were organized into local Klaverns. Exalted Cyclopses presided over the Klaverns—which at the Klan’s peak had as many as 200 active members—and organized Klan activities, including monthly meetings. In addition to the main Klan hierarchy, there was an auxiliary structure (illustrated in Figure 3) that existed primarily as a source of income generation for those involved. Klan members generated a tremendous amount of revenue. Each Ghoul paid a $10 initiation fee (equivalent to $110 in 2011 dollars), $6.50 to buy an official Klan robe (which cost roughly $2 to make), an annual membership fee of $5, an imperial tax of $1.80, and was also encouraged to 11 purchase other Klan-sanctioned merchandise including swords, Bibles, helmets, dry-cleaning, and life insurance. Joining the Klan was not a cheap undertaking. Using the numbers above, the first year of membership costs $23.30 (roughly $250 in 2011 dollars) and subsequent years were $6.80 (approximately $75 in 2011 dollars). At its peak in 1924, the Klan conservatively generated annual revenues from all sources of at least $25 million – equivalent to $300 million in current dollars. Only a small portion of this revenue was required to fund basic operations (Alexander 1965). The period of the Klan’s membership explosion was also the period when their political power is believed to have reached its pinnacle, especially in Indiana, Tennessee, Colorado, Oklahoma, and Oregon, where the Klan controlled portions of the state governments.9 The Klan shifted its focus from opposing the rights of Freedmen to those of immigrants, while also becoming more populist. The Klan typically aligned itself with the Democratic Party (at least in the South), though a candidate’s religion was a better indicator of Klan support (Moore 1991, Tucker 1991). The second Klan fell as swiftly as it rose. Incredibly, between 1924 and 1930 its national membership declined from over 1 million to just 30,000, and it has remained on the periphery since that time.10 III. Constructing a Database of Members of the Ku Klux Klan 9 See, for instance, Jackson (1992), and Moore (1991). 10 We explore possible explanations for the Klan’s rapid demise in Section V. 12 We have developed an extensive database of members of the Klan during the 1920s, when the organization was at its peak of popularity. We searched archives, libraries, and historical societies for data relevant to the Klan, finding a number of collections with data on individual Klan members, minutes from Klavern meetings, expenses, applications, robe-order forms, etc.11 Our two most comprehensive datasets are from Pennsylvania and Colorado. The Pennsylvania data consists of thirty microfilms worth of material and 32,390 Klan members. The majority of the membership data comes from dues records, personal correspondence, and applications for membership. There is also a wealth of information on finances, including receipts, checks, and quarterly reports sent from the Klabee to the Realm office. The data from Colorado is contained on three microfilms, one of which is a copy of the membership and dues ledger from the Denver metro area Klans totaling 20,351 names, addresses, and records of dues paid. We have similar individual level data, though less expansive in terms of numbers of Klansmen or accompanying information, gathered from Athens, Georgia, Pond Creek, Kentucky; Newaygo and Mecosta County, Michigan; Harlowton, Montana; Knox County, Tennessee; and Wood County, Ohio. The total number of Klansmen in these data was over 3,000, for a grand total of over 55,000.12 11 Appendix A provides further details. 12 These data were found in the following archives: Emory University (Knox County Klan Number 14, Knoxville, Tennessee), the University of Georgia Special Collections Library (Athens Klan #5), the Pennsylvania State Archives (Census of Pennsylvania, Delaware, and New Jersey Klans), Bowling Green University (Ohio Knights of the KKK, Wood County), the Indiana Historical Society (Census of Indiana), the Montana Historical Society (Wheatland Klan Number 29, Harlowton Montana), and the Colorado Historical Society (Denver Area Klans). Other individual level data was available for Bayfield, Colorado, Tillamook, Oregon and Tulsa, Oklahoma. Unfortunately, a large fraction of the data only contains last name and first initial that makes matching to the census too unreliable. 13 Given the clandestine nature of Klan activities, it is impossible to know how representative the data set we have assembled is. We are encouraged, however, by the geographic diversity of the areas the data cover, and by the varied ways in which the different data sources originated (in one case Klan headquarters was burglarized and the stolen files dropped off at a police station; in another case old records were found years later stored in an attic). After finding and constructing our bank of Klan members, we linked these data to the 1920 and 1930 decennial censuses. The extent of information on individual Klan members varied widely from archive to archive.13 Matching individuals to their census entries provides a uniform (if somewhat lackluster) set of variables to be studied. Further, census data allows us to formally compare the characteristics of Klan members to a representative cross-section of the population. To link our database to the census, we used three approaches. For the smallest communities (Harlowton, Pond Creek, and Newaygo and Mecosta), it was feasible to manually keypunch all of the information from the 1920 decennial censuses using copies of the original hand-written census rolls that we downloaded from Ancestry.com. We then linked the data we have from Klan membership rolls directly to the census data. Thus, for these communities, we not only have a complete match of Klan members and census data, but also all non-Klan members in the data set. For Athens, Knox county and Wood county, we follow the same matching procedure using the original Census rolls, but found it impractical to type in the entire population of non- Klan members. Instead, for these three areas, our control group of non-Klan members is 13 For instance, Knoxville reports years of formal education while much of Pennsylvania consisted only of last name and first initial. 14 constructed from two complimentary sources. First, we include individuals from those counties who are included in the Integrated Public Use Microdata Series (IPUMS) from the University of Minnesota Population Center in 1920 and 1930. Unfortunately, the IPUMS is only a 1 percent sample in 1920 and a 0.5 percent sample in 1930, yielding relatively small control samples. To supplement this, we constructed a random sample of roughly 500 individuals in Athens and Knox County and 1,000 individuals in Wood County by randomly choosing PDF pages of the original hand-written census rolls for the 1920 census, and keypunching all of the information on white males twenty-one years or older. The Pennsylvania and Colorado data sets are too large to feasibly link individual Klan members to original census rolls by hand. Instead, we matched our Klan data to 1920 and 1930 IPUMS data using last names and first names, or in some cases, first initials, along with city or county of residence. The merge of our Klan data with IPUMS yields 475 matches on first name, last name, and county in Pennsylvania and 243 on first and last name in Colorado. We used the observations from the census data that did not merge with the Klan data as our controls. Two different types of errors can arise when merging IPUMS and Klan data this way. First, since IPUMS is only a 1 percent sample in 1920 and 0.5 percent sample in 1930, most Klan members will go unmatched. This reduces our usable sample size, but will not otherwise bias our findings. More problematic is the fact that if an individual’s name is at all common, there may be multiple people in the county who share that name. The IPUMS entry that matches the name in the Klan data base can be a different person who happens to share the same name. To deal with this problem, we cataloged the names of every individual included in the 1910, 1920, and 1930 IPUMS in states geographically proximate to Pennsylvania and Colorado 15 respectively.14 Based on the population of the county in question, we then used this large pool of names to estimate the expected number of people in the county who shared the same name as a Klan member whenever we found a match between IPUMS and the Klan data.15 The probability that the IPUMS record that matches the Klan data is a true match is the inverse of the expected number of people with that name, and that is the value we use as our measure of Klan involvement in the Pennsylvania and Colorado samples. For common names in large counties, the expected number of people sharing a name is not trivial, e.g. we estimate that there were approximately 500 John Smith’s in Philadelphia County in 1920. Most names, however, are relatively uncommon, making the probability high that the match we find is the correct one.16 Summary statistics for Klan members and our comparison group of males aged 21 and up who are not in the Klan members are displayed in Table 1 for each of our seven data sets. Klan members are shown in the odd-numbered columns, with non-Klan males in the even columns.17 The census information available during this time period is limited, unfortunately. As would be expected, members of the Klan are much more likely to be native born. They are also less likely to hold unskilled jobs (except in Athens) and more likely to be in service jobs or professionals 14 For Colorado, we used Arizona, Colorado, Idaho, Kansas, Montana, Nebraska, New Mexico, North Dakota, Oklahoma, South Dakota, Utah and Wyoming. For Pennsylvania, we used D.C., Delaware, Illinois, Indiana, Maine, Massachusetts, Michigan, New Hampshire, New York, Ohio, Pennsylvania, Rhode Island, West Virginia, Vermont and Virginia. 15 More precisely, we estimated the expectation of the number of people sharing the name of the Klan member in the county conditional on there being at least n people in the county with that name, where n is the number of people with that name in the county’s IPUMS data. Because n people are present in IPUMS with that name, there must be at least that many people in the county as a whole with that name. 16 Unfortunately, we are unable to use the matching approaches employed by Atack, Bateman, and Gregson (1992) or Ferrie (1996). In both of those papers, the match was across Censuses, whereas in our setting the matching is between Census data and non-Census Klan data. Moreover, because our data only contain, at most, first and last name we cannot employ other traditional matching techniques used in the literature. 17 In Colorado and Pennsylvania we match Klan members probabilistically. Any case with a positive match is classified as Klan members, with the entries in those columns weighted in inverse proportion to our estimate of the number of people sharing that name in the county. 16 (except in Athens and Pennsylvania), which stands in sharp contrast to the image of today’s Klansmen. Literacy rates for all individuals, Klan and non-Klan, are very high, but slightly higher within the Klan. In most locations, Klan members are less likely to never have been married. In both Colorado and Pennsylvania, where we have county-level data on the share of the population that is Black or foreign born, we see no systematic relationship between those variables and Klan membership. Table 2 presents regression estimates of the social, demographic, and economic predictors of Klan membership in each of our locations for which we have individual level data. The specifications estimated are of the form: i socialeconomic i membershipklan !""" +#+$+= 21 _ , where klan_membership is the probability that individual i is a Klan member. Except in Pennsylvania and Colorado, this variable takes a value of 0 or 1 (either we know you are in the Klan or we know you are not); in Pennsylvania and Colorado, any name in IPUMS that matches a name in our Klan data base will have a positive value, but that value is less than one because of the possibility that multiple people share the same name. The set of covariates available in both the Klan data sets and the census in these years is limited to indicator variables for whether an individual is native born, a measure of occupational status (which we have divided into three mutually exclusive categories: unskilled/trade jobs (e.g. manual laborers, carpenters), service workers (e.g. clerks, salesman, agents), and professional jobs (e.g. doctors, lawyers, engineers)). In our regressions, unskilled jobs is the omitted category.18 Other covariates include age (which 18 We also explored classifying occupations along the dimension of whether Klan membership might directly benefit an individual’s business interests (e.g. the owner of a dry-cleaning establishment, or someone who provides personal 17 we turn into a dichotomous variable corresponding to whether the person is above or below the median age in the sample to make the interpretation of the odds ratio straightforward), and whether the respondent is literate, owns his own home, has ever been married, is the head of household, and is a veteran. With the exception of Pennsylvania and Colorado, we estimate this equation using a logistic specification. The coefficient estimates are interpretable as odds ratios. Given the relative size of our Klan and IPUMS observations varies widely across data sets, odds ratios are more easily interpretable than are OLS coefficients.19 Recall, the dependent variable in Pennsylvania and Colorado is continuous, thus we cannot use the logistic specification. Instead, we use OLS, but transform these estimates into odds ratios, using the Delta Method to compute the appropriate standard errors. Because the Pennsylvania and Colorado data cover a broad geographic area, we are able to estimate an additional specification which controls for the percent Black and percent foreign in an individual’s county, which we dichotomize into above or below the median of counties in the state.20 The two standard error confidence interval is reported in parentheses. A number of patterns emerge from Table 2. As would be expected given the strong nativist sentiments of the Klan, being native born is positively associated with Klan membership in all of our samples, with odds ratios ranging from 1.38 in Harlowton (i.e. being native born raises the likelihood of being in the Klan by 38 percent) to 5.89 in Athens (being native born services that might be used by other Klan members), but we found these variables had no predictive value for Klan membership. 19 For instance, in Bowling Green we have the universe of Klan members, but only a small fraction of non-Klan residents in the county. The OLS parameter estimate will not have the usual interpretation in this setting. 20 For our other data sets, all of the individuals reside in just one county. 18 more than quintuplesthe likelihood of being in the Klan). For Knoxville all Klan members are native born, so the parameter can not be estimated. The nativity coefficient is statistically different from one for all columns except Harlowton. The relationship between Klan membership and job status is mixed. In Athens and Pennsylvania, few professionals are in the Klan; in the other areas professionals are overrepresented in the Klan, although the differences are not always statistically significant. Literacy is higher among Klansmen in all samples, though the estimates are not statistically significant in a few of the cases. These findings are consistent with a quote MacLean (1994, p. xii) attributes to an unnamed contemporary who described Klan members as “if not the ‘best people,’ at least the next best…the good, solid middle class citizens.” 21 There is no systematic relationship between Klan membership and age, likelihood of being married, or head of household status. In Colorado and Pennsylvania, where we can include the share of the population in the county that is Black or foreign, the only statistically significant result is that in Pennsylvania Klan membership is strongly positively related to the percent foreign in the regression. Separate from the regressions, we have attempted to identify the share of police officers that are in the Klan using the data we matched to the Census. It has often been suggested that there were close ties between the Klan and law enforcement (Chalmers 1987). We find that the overlap between the Klan and the police force varies widely across the communities for which we have data. In Bowling Green, a community of less than 7,000 people, seven Klansmen were on the police force. While we do not have data on the total size of Bowling Green’s police force, 21 In terms of having relatively high education, this pattern of Klan membership parallels that of modern day terrorist organizations, as reported by Krueger and Malecková (2003), or social movements more generally (Glaeser, Laibson, and Sacerdote 2002). 19 typically a town of that size would have fewer than ten police, suggesting that most or all of the police in that area were in the Klan. In Athens, there were 19 police officers according to official records; we are able to identify four of these officers as Klan members. In other areas, however, Klan representation on the police force was much lower. In Pennsylvania and Colorado, linking Klan records to Census data, we estimate that less than one percent of police officers were in the Klan. In Knoxville, between 5 and 10 percent of police officers appear to have been Klan members. IV. Hatred and Profits: The Fiscal Genius of the Klan In this section, we will show that the Klan was tremendously successful at making money by generating revenue from members in the form of dues, taxes, and product sales. Unlike a firm, which generates revenues by selling products to outsiders, the Klan derived all of its funds from its own members. Growing the membership was therefore critical to generating profits for those in leadership positions in the Klan. In order to fuel this growth, the Imperial Wizard and Imperial Kleagle created a sales force in 1921 whose job it was to enlist new members. The subsequent meteoric growth in the Klan appears to be directly attributable to the work of this sales force (Alexander 1965, Wade 1987). For most members of the social hierarchy, the promise of financial rewards was not an important enticement. Among the local leadership, only the Exalted Cyclops and the Kligrapp were paid for their services, but this pay was minimal and these individuals almost always held other jobs outside the Klan as well. Klan membership appears to be driven less by pecuniary opportunities and more by the factors that underlie social clubs of all kinds, namely shared 20 interests and ideologies, networking opportunities, a sense of belonging or “fraternity,” etc. (Alexander 1965). The one exception was the Grand Dragon in charge of a state (or “realm”).22 When the sales structure was initially put into place in 1921, it appears that none of the initiation fees went to the Grand Dragon who was in charge of the state, but by 1924, $2.50 of each initiation was funneled to the Grand Dragon. In addition, each member was required to pay an annual $1 realm tax which was paid to the Grand Dragon responsible for that state (Alexander 1965). Loucks (1936) also reports that 50 cents of the $6.50 cost of a robe went to the Grand Dragon. The strongest financial incentives, however, were for the sales force and the national leaders. The sales force was organized on a multi-level marketing principle, much like modern companies such as Amway and Avon. The U.S. was split into nine Domains with a Grand Goblin in charge of each. A Goblin would then hire a King Kleagle for each state under his control; the King Kleagle was responsible for the army of Kleagles, salesmen who were paid by commission, in his state. The Kleagles were the core of the financial structure, actively hawking memberships for $10 apiece, and pocketing $4 from each membership sold. As noted earlier, by 1924, $2.50 was directed to the Grand Dragon who ran the state. The remaining $3.50 was sent up the recruiting structure, with the person in charge of sales in the state (King Kleagle) taking $1, the regional sales overseer (Great Goblin) getting $.50, the national sales overseer (Imperial Kleagle) $1.25, and the two most powerful men in the klan (Imperial Wizard and Grand Wizard) splitting 75 cents. Kleagles were paid for recruiting new members; once someone joined, none of the ongoing revenues accrued to the sales force. 22 At his pinnacle, D.C. Stephenson was Grand Dragon of 23 states. 21 Robes and other Klan paraphernalia generated a second source of revenue. All Klan members were required to purchase an official Klan robe produced by an approved factory; Klan members were not allowed to make their own robes. These robes were sold initially for $6.50, with the price later reduced to $5.00 when the Klan constructed its own robe production factory (Jackson 1992). The robes cost only $2.00 to produce, generating large profits for the Klan leadership. The revenues from the sale of robes were split four ways. The Kleagle who recruited the purchaser, the King Kleagle, and the Grand Dragon who headed the realm each received 50 cents. The remaining $5 went to the national headquarters, which netted $3 in profit after paying the $2 production cost. In addition to being a source of direct revenue for the Klan national leaders, the requirement that all members purchase officially sanctioned robes presumably served a second important purpose. Individual salesmen had strong financial incentives to under report the number of new members they recruited, allowing them to keep the full $10 initiation fee rather than sharing it with the central leadership. Each member had to purchase a robe from the Klan leadership, providing the central office with a roster of all new members. That made it much more difficult for local Klaverns to underreport new members, keeping the full initiation fees for themselves. Besides the purchase of robes, members were encouraged to buy an array of other officially-sanctioned products ranging from life insurance sold through the Empire Mutual Life Insurance Company, robe dry-cleaning services, and even specially wrapped candies with the klan insignia on it (Alexander 1965). 22 The final source of income was the “Imperial Tax,” which was levied on all members of a Klavern after it had received its charter, usually when it had 100 members (Alexander 1965). The tax, which totaled $1.80 per Klansmen per year, was levied in four parts of $0.45 year. This revenue stream flowed directly to the top and was not shared by anyone outside the Imperial office.23 Reliable information on the total earnings of those in the financial side of the Klan is generally not available since the number of Klan members nationwide is so uncertain. There are, however, credible membership numbers for Indiana during the period 1923-25, allowing us to calculate revenues generated in Indiana during that time period. In 1923, the Klan commissioned Ernst & Ernst to conduct an internal audit in Indiana to investigate impropriety within the hierarchy. As of August 27, 1923, the report shows 117, 245 members. In 1925, as part of the D. C. Stephenson trial, the Klan took an internal census, reporting approximately 162,000 members. Based on these numbers, knowledge of initiation fees, membership dues, and purchases of robes and other paraphernalia, we are able to construct estimates of revenues accruing to the various levels of the hierarchy. It is important to stress, however, that this case study is far from representative – the Klan’s recruiting success in Indiana was unparalleled and the period studied represents the Klan’s peak there. In addition, a number of important caveats apply. First, there is anecdotal evidence that the Klan had some difficulty actually collecting dues from members, and at least in later years offered discounts on initiation fees (Kennedy 1990). This would lead us to overstate Klan revenues across the board. On the other hand, we make two assumptions that will bias our 23 There are also references to local taxes levied by individual Klaverns (see Alexander 1965), but we have been unable to determine how widespread or large such taxes might have been. 23 calculated revenues downwards: (1) our calculations on merchandise are limited to robes, excluding other paraphernalia, and (2) we compute new initiations as the increase in the size of the membership, ignoring the fact that there was exit as well. Since much of the revenue was generated from initiation fees, trading a new member for an existing member boosted revenues. Table 3 provides estimates of the distribution of the annual revenues generated by the Klan from the state of Indiana at its peak. The first four columns correspond to the four different sources of revenues: new initiations, robes, the realm tax, and the imperial tax. The final column is the sum of these sources. The values shown have been transformed into 2006 dollars using the consumer price index.24 Our estimates are based off an assumption of 140,000 total Klan members in Indiana and 22,511 new members per year, with the latter number derived from the increase between the two Klan internal audits. New initiation fees, shown in column 1, generated revenues of nearly $200,000 for the national headquarters, over $300,000 for the highest ranking person in the national sales structure (Imperial Kleagle), over $600,000 for D.C. Stephenson, the Grand Dragon of Indiana, $132,000 for the regional sales manager (Great Goblin), $265,000 for the sales manager in Indiana (King Kleagle), and over $1 million for the street-level salesmen (Kleagles). Because we have no reliable data on the number of Kleagles in Indiana, we are unable to calculate a per capita wage for the salesmen. The sale of robes generated almost $800,000 in profit for the national headquarters; these numbers are net of the costs of supplying the robes. The Grand Dragon, King Kleagle, and the Kleagles each received over $100,000 from robes. The Imperial Tax was the single greatest 24 Between 1924 and 2006, prices rose approximately 1200 percent. 24 source of revenue, contributing nearly $3 million to national headquarters. The Realm Tax provided $1.65 million to D.C. Stephenson. According to our estimates, the single state of Indiana generated nearly $4 million in revenues for the national headquarters.25 After some modest expenses, most of that revenue would go directly to the Imperial Wizard, with the Grand Wizard having claims on some of it. D.C. Stephenson, the head of the Indiana Klan, received nearly $2.5 million annually from the state’s operations. The head of the state sales hierarchy pocketed nearly $400,000 a year. To put these numbers into perspective, in current dollars, a typical full professor during this time earned $45,000 in current dollars (Bachman 1929), Babe Ruth earned $613,000, and President Calvin Coolidge earned $885,000.26 V. Understanding the Expeditious Collapse of the Klan As fast as the Klan grew in the early 1920s, it shrunk even more quickly. Its membership in 1930 had declined to an estimated one percent of that achieved at the peak just five years earlier (Moore 1991, Wade 1987). The nativist forces that underlay the Klan remained strong, even as the organization itself imploded. The question, then, is why the Klan failed in such a dramatic fashion, and whether there are broader lessons to be drawn from the collapse. The proximate cause of the Klan’s demise was the sordid murder trial of one of the movement’s leading figures, D. C. Stephenson. In March 1925, Stephenson kidnapped 25 Alexander (1965) reports that D.C. Stephenson remitted $641,475 (in 1923 dollars) to Klan national headquarters covering the period February 17-July 14, 1923. In current dollars, that is roughly $7 million for a five-month period. That number is roughly three times higher than our estimate and is inconsistent with the other numbers reported in Alexander (1965). 26 It was not until 1931 that Ruth famously demanded a salary of $80,000 -- $5,000 more than President Calvin Coolidge – with the explanation that “I had a better year than he did.” 25 Oberholtzer, forced her onto a train heading to Chicago, forced her to drink, raped her, and even bit off portions of her flesh. Oberholtzer died shortly thereafter, leaving a death bed statement detailing her treatment at Stephenson’s hands. In November 1925, Stephenson was tried and convicted of second-degree murder and sentenced to life in prison. Ed Jackson, the governor who had been Stephenson’s protégé, refused to pardon him. The trial received spectacular media attention. The details of the murder were so gory and contrasted so sharply with Stephenson’s support of temperance and his self-defined image as a defender of “Protestant womanhood,” that entire Klaverns deserted the Klan. On the other hand, Stephenson, while prominent in the Klan, was neither the founder nor the titular leader of the Klan revival. It seems far from inevitable that his misdeeds should be the undoing of a social movement with millions of members. There appear to be at least two other factors specific to the Klan that made it particularly unstable, allowing the Stephenson affair to trigger a collapse. The first explanation for the Klan’s instability is that the sales force of Kleagles was incentivized exclusively on acquisition of new members. Whether the new member remained active in the Klan for many years, or provided public goods to the organization, was immaterial to the Kleagle. This fueled the growth in membership during the ascent, but resulted in an organization in which the level of attachment of the typical member to the group was likely to be weak, compared to other social groups. Moreover, the majority of Klan members had been part of the organization for a few years or less when the scandal unfolded, and many of these individuals held memberships in other fraternal groups. Consequently, in response to a shock that reduced the utility of being in the Klan, the proportion of members who were pushed over the margin and exited the Klan was 26 large. Contrast, for example, the sensitivity of the Klan to scandal compared to the Catholic church’s resilience in the face of widespread pedophilia charges. Unlike the Klan, people have an allegiance to a single religious denomination, one that is typically from birth, making church membership much more stable. The second factor that may have contributed to the Klan’s rapid downfall is its hate-based ideology. One of the Klan’s primary weapons has always been fear and intimidation. Any individual that challenges the Klan risks violent retribution. The more critics of the Klan who are active, the less credible is the Klan’s threat of violence towards any one of them, i.e. there are positive spillovers across critics. Consequently, there may exist multiple equilibria: one in which no critics of the Klan come forwards and another in which many critics speak out. The Stephenson scandal may have served as a coordinating device that shifted critics from the silent equilibrium to the vocal one.27 Glaeser (2005) offers another channel through which the interaction of a hate-based ideology and the Stephenson affair could undo the Klan. In Glaeser’s framework, leaders foment hatred by telling stories about the targets of the hatred which may or may not be true. It is possible for followers to determine whether the stories told are true, but only at a cost. For some set of parameters in the model, it will not be optimal for group members to pay the cost of determining whether the hateful stories are true. The Stephenson scandal showed that the Klan leadership was not living the virtuous life that it espoused. This revelation presumably also reduced the perceived likelihood that the hate-mongering stories told by the Klan leaders were true, raising the benefits to the rank and file from paying the costs to learn the truth about the Klan’s enemies. To the extent that the enemies turned out to be far less 27 In the words of Glaeser (2005), it becomes much easier to generate “hatred of the haters” after the scandal. 27 dangerous than the Klan’s rhetoric would suggest, an exogenous shock that leads members to invest in learning the truth could have a cataclysmic impact on Klan membership.28 VI. Quantifying the Social and Political Impact of the Ku Klux Klan There are extensive anecdotal examples of both the terror that the Klan produced and its political clout, but there has been little in the way of systematic quantitative analysis to substantiate such claims. Complicating this analysis is the fact that even if one finds a positive correlation between Klan activity at the county level and an outcome such as hate crimes, it is unclear whether this is causal since the same factors that lead people in an area to commit hate crimes lead them to embrace the Klan. The Klan may worsen hate crimes, or it may simply thrive in places where hate crimes would have occurred regardless of their presence. Our attempts to differentiate these two stories hinge on the crucial role that D.C Stephenson played in the rise and fall of the Klan, as discussed above. Stephenson’s charisma and entrepreneurial skills were critical to the success of the Klan in Indiana. Although similar nativist sentiments were at work in nearby states such as Pennsylvania and Ohio, the Klan grew much more slowly in these places absent a Stephenson-like figure. Thus, one might argue that a comparison of Indiana to Pennsylvania over the first part of the 1920s provides a crude measure of an independent contribution of the Klan. With Stephenson’s conviction for murder, however, the Indiana Klan collapsed even more quickly than it grew. Assuming that underlying nativist beliefs were unlikely to dissipate simply because of the murder trial and the dissolution of an 28 The scenario of collapse described is not unique to hate-based groups, but rather could operate in any organization in which leadership motivates the rank and file through lies which eventually are revealed. 28 organization, the period immediately following his conviction in 1925 provides another avenue for identifying the causal impact of the Klan. In this section, we look for evidence of an impact of the Klan in the form of the number of lynchings, change in the population distribution of Blacks and foreigners, vote shares in congressional elections, and legislation passed. We tackle these various outcomes in turn. Klan membership and the number of lynchings Given the hate-based beliefs of the Klan and its notoriety for violence, it is plausible that Klan activity would be associated with the increased frequency of hate crimes. The only type of hate crime for which there is any form of systematic data during this time period is lynchings.29 The Tuskegee Institute has assembled an extensive catalog of documented lynchings in collaboration with Project HAL – a remarkable effort to accumulate a database of all the lynchings that have taken place in the United States.30 Figure 4 presents a time series of Black lynching from 1880-1930 using data from Project HAL. The time series pattern in lynchings is exactly opposite of what one would expect if the Klan had a major impact. The number of lynchings peaked after 1890—a time period in which the Klan did not exist—and decreases steadily afterward. Between 1915 and 1925, when Klan memberships grew at fever pace, lynching continued to decrease, hitting their lowest levels in the time-series precisely when Klan popularity is at its peak. There is not a single recorded Black lynching in either Pennsylvania or Indiana in the 1920s when the Klan was thriving there. 29 For instance, the Federal Bureau of Investigation’s uniform Crime Reports, for instance, are not available until 1929, though these data would not have provided any systematic insight into hate crimes anyway. 30 See http://people.uncw.edu/hinese/HAL/HAL%20Web%20Page.htm#HAL%20History for more information on the database, its origins, and purpose. 29 The relationship between Klan membership and residential location choices of Blacks and foreigners Even if there is little evidence of widespread lynching on the part of the 1920s incarnation of the Klan, it is nonetheless possible that the Klan engaged in less drastic forms of intimidation and persecution which made life difficult for Blacks and foreigners. If that is the case, than one would expect to observe a reduction in the number of Blacks and foreigners in areas where the Klan was most active due both to people moving away and fewer people migrating in. To test this hypothesis, we run first differenced county-level regressions of the form tctctctcctc ForeignBlackPopKlanBlack ,10,310,210,11924,, )ln()ln()ln()(%% !"""# ++++=$ %%% (1) where the dependent variable is the percent change in the Black (or in some cases foreign) population in the county from one decennial census to another.31 The key explanatory variable is the percent of men aged 21 and up in the county who were in the Klan in 1924. The sample used is all counties in Pennsylvania and all counties in Indiana for which we have data. We examine changes between the 1910-1920 censuses as a test of pre-existing trends in these outcomes that we do not expect to be strongly causally related to our measure of Klan activity both because the Klan was not particularly strong in this period, and because our Klan measure is for the later period. In contrast, we do think that population changes between 1920 31 Because our dependent variable is the share of the population that is black, this regression captures differential migration patterns across races, as well as any cross-race differences in fertility. 30 and 1930 are potentially influenced by Klan activity. In all cases, the unit of observation is a county with observations weighted by county population. Table 4a presents estimation results for Pennsylvania. Columns 1-4 correspond to changes between the 1910 and 1920 censuses; columns 5-8 are for changes between 1920 and 1930. In each case the dependent variable in the first two columns is the change in percent Black, with the change in percent foreign in the next two columns. We present regression coefficients on our Klan measure with no controls and including controls for the log of population and share Black and foreign at the start of the period.32 The results in Table 4a provide little evidence for a causal impact of Klan activity on migration patterns of Blacks or foreigners. The coefficients in the period before the Klan becomes active (columns 1-4) are generally similar in sign and magnitude to those when the Klan is at its peak (columns 5-8). Absent controls, for both these periods the Klan measure is associated with strong negative values on population flows of both groups. In other words, the places where the Klan will thrive/is thriving, are places that are experiencing relative declines in the share Black or foreign. Once one controls for county population and the initial shares, however, the results for Blacks disappear. Black populations are growing more quickly in the most populous counties (such as Philadelphia), which are also areas with relatively little Klan activity. The coefficients for foreigners persist with controls. The fact that the same pattern exists in the 1910-1920 period, however, suggests a common cause of foreign emigration and Klan activity, as opposed to a causal impact of the Klan itself. All results are robust to the exclusion of Philadelphia and Pittsburgh, the two urban centers of Pennsylvania. 32 Adding county-level controls for the fraction who own farm, fraction who own home, age, family size, fraction married, and fraction in the labor force does not alter the results. 31 Table 4b is identical in structure to Table 4a, except that the sample used is Indiana counties. The results largely parallel the findings from Pennsylvania. The Indiana results for Blacks provide a hint of evidence that the Klan had some impact. In the pre-period, there is no apparent relationship between future Klan intensity and Black migration patterns. Between 1920 and 1930, there is a statistically significant negative relationship between the Klan and Black migration when controls are absent. Once controls are included, both the magnitude of the coefficient and the statistical significance decline, but the point estimate remains negative. The coefficient implies that a 10 percentage point increase in Klan prevalence is associated with roughly a 0.3 percent reduction in the Black share of the population between 1920 and 1930. The baseline Black share of the population was 3 percent, so in percentage point terms, a 10 percentage point increase in the Klan reduces the Black share of the population from 3 percent to 2.91 percent. The data provide no support for a claim that Klan activity drove away the foreign born; in all four specifications presented, the coefficient on the Klan variable is positive, and the point estimates in the later period are larger than in the pre-period. The impact of the Klan on Voting Behavior and Legislation The Klan of the 1920s is well-known for its political influence. Numerous historians have documented widespread electoral victories for Klan-supported candidates across the nation (Jackson 1992). The Klan exercised its political might in other ways as well, such as marching 50,000 Klansmen down Pennsylvania Avenue in August 1925. There is anecdotal evidence of Klan-initiated voter intimidation; an example is depicted in figure 5. 32 Politicians of both parties throughout the south and Midwest were under the Klan’s sway. Five U.S. Senators and at least four governors were Klansmen (Jackson 1992, McVeigh 2001). In 1924, the Klan captured the Republican primaries in Colorado, elected a governor of one house of the legislature, several judges and sheriffs, and the Denver chief of police. (Goldberg 1981). In Alabama, it ended the career of veteran Senator Oscar Underwood, whom it denounced as the “Jew, jug, and Jesuit candidate,” and replaced him with Hugo Black, who accepted an engraved life membership in the KKK (Morison 1980). While the Klan had a national reach and was particularly strong throughout much of the Midwest and the South, most historians believe that it experienced the most political success in Indiana. Stephenson not only aggressively boosted Klan membership and used his wealth to sponsor favored candidates, but actively meddled in Republican Party politics to engineer the election of pro-Klan candidates. At the 1924 Republican convention, Stephenson was able to force a large number of anti-Klan Republicans off the party’s ticket, an event that foreshadowed a near sweep of the state legislature and the Indiana congressional delegation that November.33 Edward L. Jackson's election to the governorship was the Klan’s crowning achievement that year (Jackson 1992). Many historians have presented these anecdotes as prima facie evidence that the Klan had an important impact on the political scene. Here we attempt a more rigorous, quantitative approach. Again using county level data on Klan membership in Indiana and Pennsylvania, we 33 The Klan turned Republican around this time as the Republican Party became more conservative and the Democrats turned more liberal. At least in Indiana, the Klan focused its efforts on knocking “old-guard” liberal Republicans out of the party, as best seen by the defeat of Merrill Moores, a veteran congressman from Indianapolis, at the 1924 Republican convention. 33 estimate the Klan’s effect on the change in republican vote share during the Klan's rise and fall . To do this, we estimate first differences county-level regressions similar to equation (1): ! %" Republican Vote c,t = #(%Klan c,1924 ) + $ 1 ln(pop c,1920 ) + $ 2 ln(blackpop c,1920 ) + $ 3 ln( foreignpop c,1920 ) + % c,t where the dependent variable is the percent change in the Republican vote share. We examine changes between 1920 and 1924 when the Klan’s membership growth was in overdrive and their political strength hit its zenith. We also investigate the change in Republican vote share between 1924 and 1926 and between 1924 and 1934. The Stephenson case, which initiated the crumbling of the Klan, occurred in 1925. This provides a rare opportunity to estimate the impact of the Klan during its decline. In all cases, the unit of observation is a county, with observations weighted by the county's population. Table 5a presents results for Pennsylvania. Columns 1-2 correspond to changes between 1920 and 1924, the rise of the Klan; columns 3-4 relate to changes between 1924 and 1926, the immediate impact of the Stephenson trial; and column 5-6 cover the years between 1924 and 1934, the years subsequent to the Klan's fall. We present regression coefficients on our Klan measure with no controls and including controls for the log of population and log of Black and foreign population, respectively. The estimates in Table 5a provide little evidence that the Klan had much influence on vote shares in Pennsylvania during its peak years. Between 1920 and 1924, the effects actually go in the opposite direction that one might have expected. Similarly, from 1924 to 1926, there is 34 an increase in Republican vote shares, the opposite of what one might expect, which then declines modestly.34 Table 5b provides the estimates of most interest, partly because many have argued that the Klan had an overwhelming influence on Indiana politics (Moore 1991) and partly because Stephenson provides a quasi-experiment regarding the effect of the Klan's rapid rise and fall on politics. Yet, as the table suggests, the Klan had surprisingly little influence. During the Klan’s peak growth period, the coefficient on percent Klan is .22. Evaluated at the mean, this implies that a 10 percentage point increase in Klan was associated with a 2.2 increase in the Republican vote share between 1920 and 1924. Klan popularity in a county is associated with a small decline immediately following the Stephenson trial, a decline that grows larger with time. Figure 6 further underscores the lack of causal influence the Klan had on politics. The figure plots the change in Republican vote share in three states: Pennsylvania, Indiana, and Illinois. Illinois makes an interesting “control” state, as it was deeply nativist but experienced only limited Klan activity. The trend lines follow a remarkably similar pattern. Despite our estimates regarding the relatively small effect of the Klan on changes in vote shares, eleven of the thirteen elected candidates elected to the U.S. House in Indiana in the 1924 elections were those backed by the Klan, and the Klan controlled virtually the entire state government. A key question, then, is whether these politicians were able to pass legislation that furthered Klan goals. In 1925, a series of bills were introduced, mainly concerning education, to further the Klan’s “100 percent Americanism” campaign. Weaver (1954) provides careful documentation of Klan-initiated legislation. The bills included the Religious Garb Bill, which 34 These results are robust to excluding the two most populous counties, Philadelphia and Pittsburgh, from the estimation. 35 would have prevented any person (i.e. Catholic nuns) who wore religious garb from teaching in the public schools; the Public Schools Graduate Bill, which would have required that teachers’ licenses in Indiana be granted only to individuals who attended public schools; the Uniform Textbook Bill, which would have made it compulsory for private and public schools to use the same textbooks; and the Bible Reading Bill, which would have made Bible reading mandatory in the public schools. Also on the Klan’s agenda were bills to allow college students to receive credit for Bible study outside the school, to allow public school students to leave school early for religious education, to create an entirely new board of education, and to have all students in Indiana study the U.S. Constitution. The final bill was the only one enacted. VI. Conclusion The Ku Klux Klan symbolizes the extremes of race and religious hatred in America. Since its inception in the months after the Civil War, the Klan’s organization, mission, and power have varied tremendously, with membership and political influence peaking in the mid- 1920s when over one million Americans were members. Our statistical analysis, however, finds that, contrary to conventional wisdom, the Klan had little impact on Black or foreign born migration, lynchings, or politics during this time period. Rather, the 1920s Klan is best described as an enormously successful marketing ploy; a classic pyramiding scheme – officials at the top getting rich off of the individuals at the bottom – energized by sales agents with enormous financial incentives to sell hatred. 36 References Alexander, Charles. “Kleagles and Cash.” 1965. The Business History Review. Vol. 39, No. 3, pp. 348-367 Bachman, Frank P. (1929) “Salaries and Supplemental Earnings of College Teachers.” Science, Vol. 70, No. 1802, pp 26-29. Chalmers, David. Hooded Americanism. 1987. Durham, NC: Duke University Press. Davis, Lenwood and Janet Sims-Wood. 1984. The Ku Klux Klan: A Bibliography. Westport, CT: Greenwood Press. Glaeser, Edward L. 2005. “The Political Economy of Hatred.” Quarterly Journal of Economics. Vol. 120, No. 1, pp. 45-86. Glaeser, Edward L., David Laibson, and Bruce Sacerdote. 2002. “An Economic Approach to Social Capital,” Economic Journal Vol. 112: 437-458. Goldberg, Robert Alan. 1981. Hooded Empire: The Ku Klux Klan in Colorado. Urbana: University of Illinois Press. Jackson, Kenneth T.The Ku Klux Klan in the City, 1915-1930. 1992. Chicago, IL: Elephant Paperback. Kennedy, Stetson, 1990, The Klan Unmasked. University Press of Florida. Ku Klux Klan, 1921, Constitution and Laws of the Knights of the Ku Klux Klan, Atlanta, GA. Krueger, Alan and Maleckova, Jitka. “Education, Poverty, and Terrorism.” 2003. The Journal of Economic Perspectives. Vol. 17, No. 4, pp. 119-144. Loucks, Emerson Huntsberger. The Ku Klux Klan in Pennsylvania, a Study in Nativism. 1936. New York, Harrisburg, PA: The Telegraph Press. MacLean, Nancy. Behind the Mask of Chivalry: The Making of the Second Ku Klux Klan. 1994. New York: Cambridge University Press. McVeigh, Rory. “Power Devaluation, the Ku Klux Klan, and the Democratic National Convention of 1924.” 2001. Sociological Forum 16(1):1-30. Moore, Leonard. Citizen Klansmen. 1991. Chapel Hill, NC: University of North Carolina Press. 37 Morison, Samuel, Commager, Henry, and Leuchtenburg, William. The Growth of the American Republic. 1980. Oxford: Oxford University Press. Newton, Michael. The Ku Klux Klan. 2006. Jefferson, NC: McFarland and Company. Newton, Michael. The Invisible Empire. 2001. Gainesville, FL: University of Florida Press. Skocpol, Theda. Diminished Democracy. 2003. Norman, OK: University of Oklahoma Press. Tucker, Richard. The Dragon and the Cross. 1991. Hamden, CT: Archon Books. Wade, Wyn Craig. The Fiery Cross. 1987. Oxford: Oxford University Press. 38 Appendix A: Data Appendix We have developed an extensive database of members of the Ku Klux Klan during the 1920s. The database was developed in three stages. First, we scoured the country for archives, libraries, and historical societies that contained data on the Ku Klux Klan. Once we found those that contained potentially relevant information, we attempted to partner with these institutions by offering to pay for their collections to be microfilmed. Our price: one copy of all microfilm. Second, we analyzed these microfilms for data on individual Klan members, minutes from Klavern meetings, expenses, filed applications, and so on. Many of these forms contain information about each Klan member (name, height, weight, e.g.). From this, we developed the initial list of Klan members from all the microfilm that had such information. The final stage in creating our database was to link up Klan members to census information so that they could be compared to non-Klan members. Below, we detail each of these stages. Below, we provide the details of this three-stage process. In Search of Klan Members For obvious reasons, identifying members of secret societies or terrorist groups is a difficult task. We began by searching university libraries, archives, and historical societies around the country. Twenty-three of these organizations possessed a total of thirty collections regarding the Ku Klux Klan. Most were simply old copies of Klan propaganda, a robe here and there, or various copies of the Kloran and the Klan monthly newsletter. A number of repositories had remarkable data on members of the Invisible Empire including membership roles and dues paid, financial records, first and last names, as well as other useful information. These included Emory University (Knox County Klan Number 14, Knoxville, Tennessee), the University of Georgia (Athens Klan #5), the Pennsylvania State Archives (Census of Pennsylvania, Delaware, and New Jersey Klans), Bowling Green University (Ohio Knights of the KKK, Wood County), the Indiana Historical Society (Census of Indiana), Clark Historical Library (Newaygo and Mecosta County, Michigan and Calvin Enders Collection), and the Montana Historical Society (Wheatland Klan Number 29, Harlowton Montana). As with most data on organizations that are difficult to obtain, one worries that the information that has been discovered is not representative of the organization. While we cannot answer this directly, we have gone to great lengths to understand the origins of each of the collections in our database. The Klan collections arrived at their respective locations in various ways. In Pennsylvania, the Grand Wizard was burglarized and all the Klan’s confidential files were dropped off at the State Police headquarters in Harrisburg in the middle of the night. The collections at Emory and Bowling Green Universities and the Montana Historical Society were donated by unknown sources; typically an individual who discovered the files in their attic and believes they may have historical value but for which they want to distance themselves. The Indiana Historical Society was donated by a series of individuals in various parts of Indiana. Further information about the Indiana Klan was made publicly available through the criminal trial of D.C. Stephenson. 39 Once the files from the assorted collections were microfilmed, we printed each page to PDFs, which are available from the authors upon request. We then keypunched all information on individual members from these PDFs into Excel. This serves as the set of Klan members. Other information of interest such as dues records, expenses, checks written, bank statements, or other financial information were also type into separate spreadsheets. Matching Klan Members to the 1920 and 1930 Censuses To link our database to the census, we used three approaches. For the smallest community (Harlowton), it was feasible to manually keypunch all of the information from the 1920 and 1930 decennial censuses using PDF versions of the original hand-written census rolls available at ancestry.com. Appendix Figure 1 provides a sample of a typical census image. The census rolls are small and hand-written, which can make it difficult to be certain about some entries. In cases in which information was either blank or not legible, we created an indicator variable to denote both of these types of missing data. We then linked the data we have from Klan membership rolls directly to the Census data. Thus, for Harlowton, we not only have a complete match of Klan members and census data, but also have all non-Klan members included in the data set. For Knox and Wood counties along with Athens, we follow the same matching procedure using the original Census rolls, but found it impractical to type in the entire population of non- Klan members. Instead, for these two cities, our control group of non-Klan members is constructed from two complimentary sources. First, we include individuals from those counties who are included in the Integrated Public Use Microdata Series (IPUMS) from the University of Minnesota Population Center in 1920 and 1930. Unfortunately, the IPUMS is only a 1 percent sample in 1920 and a 0.5 percent sample in 1930, yielding a relatively small control samples.35 To supplement this, we constructed a random sample of 500 individuals in Knox County and 1,000 individuals in Wood County by randomly choosing PDF pages of the original hand-written census rolls for the 1920 decennial census available at ancestry.com, and keypunching all of the information on white males 21 or older. The randomization was accomplished in a few steps. First, we set the number of observations equal to the total number of sheets available across all enumeration districts in a county. Then we generated a variable that took the values of 1 through the total number of sheets. Next, we took a random sample of 50 of those numbers. Before we took the sample we set the seed, which allows you to re-produce the same random numbers each time you run the do file. Finally, we took those numbers and calculated the actual sheet number within each district. For example, random number 151 in Knoxville represents sheet 39 of District 95. The Colorado and Pennsylvania data sets were too large to feasibly link individual Klan members (more than 30,000 of them in each case) to original census rolls by hand. Thus, in Colorado and Pennsylvania, we exclusively matched our Klan data to IPUMS data for their 35 wwThe data from Knoxville was quite comprehensive; 6 of the census variables were already in our files. Further, there were other interesting variables in the data such as years of education. As such, we just appended census data onto these datasets and assumed that those from the census were not Klan members. 40 respective states. Unfortunately, the data on klan members from these two states consisted mainly of last names and first names, or in many cases, first initials, along with city or county of residence. For instance, in Pennslyvania this merge yielded 394 matches on last name, first name and county – roughly 1% of our original klan data. We treated the observations from the census data that did not merge with the klan data as controls. Merges were performed on last name, first name, and county. In both the klan dataset and the IPUMS data there were duplicates based on these variables. In these cases, we counted the number of times that the duplicate arose and weighted each observation by one over this number. Dropping these observations does not in any way alter the results. Appendix B: A Guide to Klan Terminology Realm: In both the Reconstruction and modern Klans, a realm is equivalent to a state, administered by a grand dragon. The term is never applied to smaller geographical units, although the 1920s saw the term occasionally applied to larger areas, when membership in one state was insufficient to justify an imperial charter. Thus, the states of Maine, Vermont, and New Hampshire were considered a single realm. Province: In Reconstruction Klan terminology, the province was equivalent to a county, with a grand giant presiding over KKK affairs. In the modern Klan, province replace the obsolete Reconstruction term “dominion,” generally referring to the congressional district but more flexible, adapted to larger and smaller areas based on the needs of the Klan. The modern ruling officer is the great titan. Lictor: In the early Reconstruction Klan, the title was applied to a guard of the den. With the passage of time, the designation was apparently replaced by, or used interchangeably with, that of the nighthawk. Kludd: the chaplain of a klan, this rank is occupied by an ordained Protestant minister whenever possible. A great kludd offers prayers at the province level, with a grand kludd serving the realm and an imperial kludd selected for the Invisible Empire at large. Klokard: designated as the lecturer of a klan, the klokard has no counterpart at the province level. A grand klokard serves for the realm, while an imperial klokard is elected for the Invisible Empire at large. Klongress: Described in official documents of Mississippi’s White Knights as the Klan’s legislative branch, the Klongress theoretically consisted of two houses, the Klonvocation and the Klanburgesses. Members of the Klonvocation were designated “senators,” empowered to meet if and when called on by the Klanburgesses. The authoritarian control exercised by Imperial Wizard Sam Bowers suggests that the Klongress existed only on paper. 41 Klexter: Designated as the outer guard of a klan, the klexter is responsible for external security during regular meetings. The office does not exist at the province level. Each realm is served by a grand klexter, while the imperial klexter operates from KKK national headquarters. Kligrapp: As secretary of a klan or other jurisdictional unit, this title is further designated by rank to include the great kligrapp (at the province level), the grand kligrapp (of a realm), and the imperial kligrapp (for the overall Klan). Klokan: An individual member of the klokann, this officer serves as an adviser, auditor, and investigator for his respective geographical unit. Klokann: The advisory board of a klan, serving as auditors and investigators, the klokann is composed of several members dubbed klokans. At the klanton and province levels, three members are selected for each board, with the latter body known as the great klokann. Five members are specified for the grand klkann (at the realm level), and the imperial klokann (at national headquarters). Klavern: Technically applied to the local headquarters of a klan, this term is often used interchangeably to describe the klan itself. Kleagle: IN Klan parlance, a kleagle is an organizer or recruiter, appointed by the imperial wizard or his imperial representative to “sell” the KKK among nonmembers. Kleagles are generally paid on commission, receiving an established percentage of each recruiter’s klectokon. Klanton: a subdivision of a province, a klanton is the domain of a klan, with an exalted Cyclops and his twelve terrors presiding. Klarogo: As the inner guard of a klanton, the klarogo is roughly equivalent to a sergeant-at-arms. The office does not exist at the province level. Other rank designations include the grand klarogo (of a realm) and the imperial klarogo (for the national Klan). Klabee: The treasurer of a klanton, klabees are further designated by rank with the title of great klabee (at the province level), grand klabee (for a realm), and imperial klabee (for the national Klan). Kladd: As the conductor of a klanton, a kladd is the custodian of ritual paraphernalia, and he also introduces candidates for “naturalization.” The office does not exist at the province level, but rank is otherwise designated by the titles of grand kladd (for a realm) and imperial kladd (at the national levels). Klaliff: The vice president of a klanton, klaliffs multiply as one ascends the Klan ladder of rank. At the province level, three great klaliffs serve as an advisory board for the great titan. The grand 42 klaliff is vice president of a realm, while the imperial klaliff serves as second in command after the imperial wizard. Klan: The smallest unit of the Invisible Empire, a klan is administered by an exalted Cyclops and his twelve terrors presiding. To further confuse the issue, local units of the Constitutional Union Guard were also referred to as “klans” during Reconstruction. Imperial Kloncilium: In the modern United Klans of America, the kloncilium consists of fifteen genii serving as the imperial wizard’s command staff. Members include the imperial klaliff, imperial klokard, imperial kludd, imperial kligrapp, imperial klabee, imperial kladd, imperial klarogo, imperial klexter, imperial klonsel, imperial nighthawk, and the five-man imperial klokann. Imperial representative: The acting grand dragon of a prospective realm, this officer is appointed by an imperial wizard to organize local Klansmen in sufficient numbers to elect state officers and obtain a realm charter. He may also be empowered to appoint kleages, if that authority is not restricted to the imperial wizard. Imperial Wizard: Since 1915, this title has generally applied to the presiding officer of the Invisible Empire. He is the Klan’s supreme chief executive, its military commander-in-chief, and (at least in the United Klans of America) chairman of the imperial kloncilium, consisting of his fifteen genii. Some Klans use other titles for their leader: a few cling to the original “grand wizard” designation used by Nathan Forrest during Reconstruction. Other Klan chiefs have proclaimed themselves “emperor,” “grand emperor,” or simply “president.” Imperial Giant: Theoretically reserved for retired imperial wizards in the United Klans of America, this honorary title has never been used, since founding wizard Robert Shelton still remains in office. Hydras: IN Klan parlance, hydras are the advisory officers chosen to serve a grand dragon. During Reconstruction, there were six, each one without individual titles. The modern Klan specifies nine hydras, including the grand klaliff, grand klokard, grand kludd, grand kligrapp, grand klabee, grand kladd, grand klarogo, grand klexter, and grand nighthawk. Grand Wizard: As supreme leader of the Reconstruction Klan, the grand wizard was advised and assisted by ten genii. Grand Magi: During Reconstruction, this officer served as second in command of a den, presiding in the absence of the grand Cyclops. Today, the klaliff holds an equivalent rank as vice president of a klavern. Grand Monk: IN the original Klan, this officer was third in command of a den, presiding when both the grand Cyclops and the grand magi were absent. There is no equivalent rank in the 20th century KKK. 43 Grand Scribe: As secretary of the Reconstruction Klan, the grand scribe served at various levels from local dens to imperial headquarters without further differentiation in title. His modern equivalent is the kligrapp. Grand Sentinel: During Reconstruction, this officer held responsibility for den security, including selection and assignment of the grand guard. The modern KKK divides this task between the klexter and klarogo, with occasional assistance from special security squads. Grand Titan: The commanding officer of a dominion during Reconstruction, this officer was advised by six furies. Grand Turk: In the Reconstruction Klan, the grand turk served as den’s executive officer, charged with informing the rank and file of any irregular of informal gatherings called by the grand cyclops. Grand Cyclops: In the Reconstruction Klan, this officer commanded a local den, assisted by two nighthawks. Grand Dragon: Throughout Klan history, this title has applied to the chief officer of a realm. During Reconstruction, the grand dragon was advised by eight hydras, a number increase to nine for the 20th century Klan. Grand Ensign: In the Reconstruction Klan, the grand ensign was a local officer in charge of the Klan banner for a particular den. Today, the kladd fulfills that function, while the term “grand ensign,” at least in the United Klans of America, is reserved for the banner itself. Grand Exchequer: During Reconstruction, the grand exchequer was a Klan treasurer, serving at various levels from individual dens to imperial headquarters without different titles to designate rank. His modern equivalent would be the klabee. Grand Giant: In the Reconstruction Klan, this title designated the chief officer of a province who was advised by four goblins. Some modern groups retain the office, while the United Klans of America has added a unique twist, treating “grand giant” as an honorary title sometime (but not always) bestowed on retired grand dragons. Grand Guard: Selected as sentries for the original KKK, members of the grand guard were chosen by a den’s grand sentinel to stand watch during meetings. Their symbolic function is fulfilled in modern times by the klexter, but Klans of significant size normally maintain a special, uniformed security force to patrol public rallies and demonstrations. Ghouls: The original prescript of the Reconstruction KKK applied this designation to rank-and- file Klansmen. Published sources differ on whether the term still applies to modern Klansmen, 44 but it does not appear in the Kloran or in any current constitution of the various competing Klans. Dominion: According to the prescript of the Reconstruction KKK, a dominion was equivalent to a state congressional district, with Klan business supervised therein by an officer dubbed the grand titan. In modern Ku Klux jargon, the term has been replaced with “province,” carrying a flexible—but often identical—definition. Den: The official designation for local Klan unit or meeting place during Reconstruction, the term has generally been supplanted by “klavern” in 20th century Klan jargon. Klectokon: A sum of money not less than $10 and not more than $25. Figure 1: Membership in the Ku Klux Klan, 1921-2000 0 500000 1000000 1500000 2000000 2500000 3000000 3500000 4000000 4500000 1921 1922 1923 1924 1927 1929 1930 1970 1980 2000 Year KKK Members Sources: 1921, Jackson (1992); 1922, Chalmers (1965), 1923-1930, Newton (2006); and 1970-2000, Anti-Defamation League. Empire: U.S. Grand Wizard: Chairman Imperial Wizard: CEO 15-member Imperial Kloncilium (cabinet) composed of Genii: Klaliff (first vp), Klazik (second vp), Klokard (lecturer), Kludd (chaplain), Kligrapp (secretary), Klabee (Treasurer), Kladd ("Conductor")*, Klarogo (Inner-guard), Klexter (Outer-guard), Klonsel (Attorney), Night Hawk (Courier), and the four Klokann (auditors). Realm Realm Realm: State Grand Dragon: Head of Realm Imperial Wizard: CEO 15 Hydras: Same positions as Kloncilium, except at state level. Congressional District Grand Titan: Head of electoral operations in each district Furies: Aides to the Grand Titan District District Klavern Klavern Klavern: Local Organizations Exalted Cyclopes: Head of Klavern Several local officers, depending on Klavern, almost always including Kligrapp (secretary), Klabee (treasurer), and Kludd (chaplain) Figure 2: Klan Basic Organization Ghoul Rank-and-file Klansman who mostly attending Klavern meetings Imperial Wizard National: 12.5 percent Great Goblin Great Goblin Nine nationally: 5 percent Figure 3: Auxiliary Recruiting Structure of the Ku Klux Klan Ghouls Members Imperial Kleagle National: 12.5 percent Grand Wizard National: 20 percent Great Goblin Kleagle Local Salesman: 40 percent KleagleKleagle King Kleagle State: 10 percent King Kleagle King Kleagle Figure 4: Black Lynching, 1880-1930 Figure 5: An Example of Klan Initiated Voter Intimidation .3 .4 .5 .6 .7 % R ep ub lic an V ot e 1900 1910 1920 1930 1940 Year Indiana Pennsylvania Illinois Figure 6: Republican Vote Share, by State City/State: Klan Non-Klan Klan Non-Klan Klan Non-Klan Klan Non-Klan Klan Non-Klan Klan Non-Klan Observations: 41 569 551 942 2208 1316 566 498 243 1573 475 40789 share native born 0.902 0.851 1.000 0.983 0.961 0.882 0.995 0.974 0.895 0.765 0.886 0.740 (0.300) (0.357) (0.000) (0.129) (0.194) (0.323) (0.073) (0.160) (0.307) (0.424) (0.318) (0.438) share unskilled 0.439 0.711 0.595 0.664 0.499 0.812 0.489 0.430 0.518 0.593 0.713 0.737 (0.502) (0.454) (0.491) (0.473) (0.500) (0.391) (0.500) (0.496) (0.501) (0.491) (0.453) (0.441) share service workers 0.390 0.171 0.297 0.265 0.373 0.127 0.389 0.345 0.395 0.336 0.265 0.219 (0.494) (0.377) (0.458) (0.442) (0.484) (0.333) (0.488) (0.476) (0.490) (0.473) (0.442) (0.413) share professionals 0.171 0.118 0.108 0.071 0.127 0.061 0.041 0.084 0.086 0.071 0.022 0.045 (0.381) (0.323) (0.310) (0.257) (0.333) (0.239) (0.198) (0.278) (0.282) (0.257) (0.147) (0.207) literacy rate 1.000 0.991 0.959 0.936 0.987 0.960 0.970 0.964 0.996 0.987 0.994 0.952 (0.000) (0.093) (0.198) (0.245) (0.112) (0.197) (0.171) (0.187) (0.065) (0.112) (0.079) (0.214) share own home 0.771 0.408 0.607 0.594 0.272 0.217 0.465 0.509 0.484 0.521 (0.426) (0.492) (0.489) (0.491) (0.445) (0.413) (0.500) (0.500) (0.500) (0.500) share never married 0.075 0.294 0.183 0.246 0.205 0.188 0.189 0.341 0.255 0.335 0.251 0.297 (0.267) (0.456) (0.387) (0.431) (0.404) (0.391) (0.392) (0.475) (0.437) (0.472) (0.434) (0.457) average age 46.900 41.820 34.849 37.069 43.336 43.948 39.634 39.739 41.619 40.405 40.162 39.642 (11.112) (14.911) (11.064) (14.204) (16.852) (15.480) (14.031) (15.755) (14.593) (15.640) (14.752) (15.116) share head hh 0.927 0.698 0.724 0.765 0.726 0.576 0.657 0.556 0.638 0.621 (0.264) (0.460) (0.447) (0.424) (0.446) (0.495) (0.476) (0.497) (0.481) (0.485) share veterans 0.171 0.167 0.124 0.078 (0.381) (0.374) (0.330) (0.269) pct of county blk 0.020 0.020 0.041 0.037 (0.009) (0.008) (0.034) (0.035) pct of county frgn 0.130 0.131 0.171 0.147 (0.022) (0.024) (0.056) (0.068) Table 1: Summary Statistics Harlowton Knoxville Notes: Data in this table correspond to the six different geographic areas noted in the column headings. The first column in each area corresponds to Klan members that we successfully matched to individuals in either the 1920 or 1930 Census; the second column contains Census data for white males over 21 who did not match anyone in our Klan records. For Harlowton, the Non-Klan data reflects the entire Census population. For the other areas, the non-Klan samples are based on all individuals included in the Public Use Microdata for 1920 and 1930 in these areas. In addition, we supplemented the PUMS data with a random sample drawn from the original Census rolls in Athens, Knoxville and Bowling Green. For Colorado and Pennsylvania, a probablistic match was made between Klan records and the PUMS data (see appendix for details). Any record with a positive match with the Klan data is treated as a Klan member in these summary statistics; some of these are false positives do to multiple people in a county sharing the same name. Standard deviations in parentheses. AthensBowling Green Colorado Pennsylvania Initiation Fees Robes Imperial Tax Realm Tax Total Recipient of Revenue National Headquarters 532 796 2,971 0 4,299 Imperial Wizard 100 796 2,971 0 3,867 Grand Wizard 100 0 0 0 100 Imperial Kleagle 332 0 0 0 332 Grand Dragon of Indiana 663 133 0 1,650 2,447 Great Goblin 133 0 0 0 133 King Kleagle 265 133 0 0 398 All other Kleagles combined 1,062 133 0 0 1,195 Total 2,655 1,195 2.971 1,650 8,472 Source of Revenue Table 2: Estimates of the Revenues Generated by the Klan in Indiana in 1924, (in thousands of $2006) Notes: Entries are the estimated revenues (in thousands of $2006) generate by the Klan in the state of Indiana in the year 1924 for Klan leaders and the sales force. Estimates are based on author calculations. Estimated number of Klan members in Indiana in 1924 is 140,000, with 22,511 new Klan members that year. The sources of revenue and the distribution across Klan leaders are based on a number of (sometimes conflicting) historical accounts, but especially on Anderson (1965). Entries are inflation adjusted using the CPI. City/State: Harlowtown Knoxville Bowling Green Athens Colorado Pennsylvania Variables: Klan Klan Klan Klan Klan Klan nativity 1.38 - 2.27 5.89 2.74 3.26 (0.45 - 4.26) (1.29 - 4.00)** (1.60 - 21.70)** (1.76 - 4.28)** (2.43 - 4.37)** service workers 3.66 1.19 5.20 0.92 1.22 1.00 (1.68 - 7.98)** (0.86 - 1.63) (3.47 - 7.80)** (0.69 - 1.22) (0.89 - 1.68) (0.81 - 1.24) professionals 2.00 1.75 3.15 0.46 1.27 0.40 (0.77 - 5.22) (1.06 - 2.89)* (1.75 - 5.65)** (0.27 - 0.80)** (0.73 - 2.19) (0.21 - 0.75)** literacy 1.36 1.68 3.62 1.58 4.66 (0.69 - 2.66) (0.66 - 4.31) (1.28 - 10.28)* (0.20 - 12.28) (1.47 - 14.74)** own home 4.38 0.96 1.08 0.90 0.87 (1.89 - 10.16)** (0.65 - 1.42) (0.77 - 1.51) (0.66 - 1.22) (0.72 - 1.06) never married 0.59 0.69 2.08 0.63 0.88 0.64 (0.14 - 2.58) (0.48 - 1.00) (1.28 - 3.36)** (0.40 - 0.98)* (0.57 - 1.36) (0.47 - 0.86)** age 1.67 0.73 0.88 0.81 1.21 1.23 (0.78 - 3.59) (0.54 - 0.99)* (0.63 - 1.24) (0.60 - 1.09) (0.89 - 1.65) (1.00 - 1.51)* head of household 5.51 1.70 0.78 1.11 0.80 (0.84 - 36.03) (0.90 - 3.18) (0.36 - 1.69) (0.72 - 1.73) (0.61 - 1.05) veteran 1.02 1.54 0.53 (0.39 - 2.68) (0.69 - 3.45) (0.30 - 0.93)* % of county pop black 0.94 0.82 (0.62 - 1.44) (0.66 - 1.01) % of county pop foreign 1.45 3.03 (0.73 - 2.87) (2.39 - 3.82)** Observations 528 1493 3522 1064 1805 41150 Klan Observations 41 551 2208 566 243 475 Table 3: Logistic Regression Estimates for Predictors of Klan Membership Notes: Coefficients in the table are estimated odds ratios. Each column in the table represents results from a single regression. The first three columns are from a logistic regression. Because the Klan/Census match is done probabilistically for Colorado and Pennsylvania, the dependent variable in the final columns is a continuous variable rather than an indicator variable. The entries in the final two columns are based on linear probability models, with the coefficients transformed into odds ratios. 95 percent confidence intervals in parentheses. For the final two columns, these confidence intervals are computed using the delta method. One asterisk denotes statistical significance at the 5% level; two asterisks denote statistical significance at the 1% level. Literacy in column 1 and nativity in column 2 are dropped because there is no variation within Klan members. Decade: (1) (2) (3) (4) (5) (6) (7) (8) Variable: % change in blacks % change in blacks % change in foreigners % change in foreigners % change in blacks % change in blacks % change in foreigners % change in foreigners % Klan in 1924 -0.3054 0.0021 -0.3263 -0.7244 -0.6269 0.0040 -0.4408 -0.6631 (0.0834)** (0.0724) (0.2560) (0.2151)** (0.1606)** (0.1223) (0.2002)* (0.1523)** log (county pop in 1910) 0.0036 -0.0062 _ _ (0.0022) (0.0067) log (black pop in 1910) 0.0007 0.0066 _ _ (0.0007) (0.0019)** log (foreign pop in 1910) 0.0001 -0.0110 _ _ (0.0011) (0.0033)** log (county pop in 1920) _ _ 0.0075 0.0147 (0.0034)* (0.0043)** log (black pop in 1920) _ _ 0.0022 0.0003 (0.0010)* (0.0013) log (foreign pop in 1920) _ _ -0.0012 -0.0161 (0.0018) (0.0022)** Observations 67 67 67 67 67 67 67 67 R-squared 0.17 0.61 0.02 0.57 0.19 0.69 0.07 0.65 Table 4A: The Relationship Between KKK Membership and Population Changes, Pennsylvania 1910 -1920 1920-1930 Notes: The dependent variable in odd columns is the percent change in the black population in a county between decennial censuses; the dependent variable in even columns is the percent change in percent foreign born in the county. The first four columns are the changes between 1910 and 1920; the last four columns represent changes between 1920 and 1930. The key dependent variable in the specifications is the percent of the eligible county residents in the Klan at its peak in 1924. Klan membership began to grow rapidly beginning in 1921. Standard errors in parentheses. One asterisk denotes significance at the 5 percent level; two asterisks reflect significance at the 1 percent level. See the data appendix for further details on the data sources and construction. The unit of observation is a county in Pennsylvania. Decade: (1) (2) (3) (4) (5) (6) (7) (8) Variable: % change in blacks % change in blacks % change in foreigners % change in foreigners % change in blacks % change in blacks % change in foreigners % change in foreigners % Klan in 1924 -0.0254 0.0083 0.0990 0.0320 -0.0709 -0.0285 0.1483 0.0423 (0.0146) (0.0120) (0.0244)** (0.0174) (0.0175)** (0.0160) (0.0337)** (0.0229) log (county pop in 1910) 0.0054 0.0123 _ _ (0.0030) (0.0043)** log (black pop in 1910) -0.0023 0.0003 _ _ (0.0007)** (0.0011) log (foreign pop in 1910) 0.0038 -0.0139 _ _ (0.0012)** (0.0017)** log (county pop in 1920) _ _ -0.0098 0.0303 (0.0038)* (0.0054)** log (black pop in 1920) _ _ 0.0002 -0.0020 (0.0010) (0.0014) log (foreign pop in 1920) _ _ 0.0083 -0.0219 (0.0015)** (0.0022)** Observations 84 83 84 83 84 82 84 82 R-squared 0.04 0.49 0.17 0.67 0.17 0.48 0.19 0.72 Table 4B: The Relationship Between KKK Membership and Population Changes, Indiana 1910 -1920 1920-1930 Notes: The dependent variable in odd columns is the percent change in the black population in a county between decennial censuses; the dependent variable in even columns is the percent change in percent foreign born in the county. The first four columns are the changes between 1910 and 1920; the last four columns represent changes between 1920 and 1930. The key dependent variable in the specifications is the percent of the eligible county residents in the Klan at its peak in 1924. Klan membership began to grow rapidly beginning in 1921. Standard errors in parentheses. One asterisk denotes significance at the 5 percent level; two asterisks reflect significance at the 1 percent level. See the data appendix for further details on the data sources and construction. The unit of observation is a county in Indiana. (1) (2) (3) (4) (5) (6) Variable: % change in republican vote 1920-1924 % change in republican vote 1920-1924 % change in republican vote 1924-1926 % change in republican vote 1924-1926 % change in republican vote 1924-1934 % change in republican vote 1924-1934 % Klan in 1924 -0.4829 -0.5313 0.6555 2.3358 2.6619 -0.3480 (0.8892) (1.0899) (0.8327) (0.9708)* (1.1883)* (1.0713) log (county pop in 1920) -0.0468 0.0261 0.0580 (0.0321) (0.0286) (0.0316) log (black pop in 1920) 0.0094 0.0093 -0.0363 (0.0093) (0.0083) (0.0092)** log (foreign pop in 1920) 0.0255 -0.0137 -0.0360 (0.0163) (0.0145) (0.0160)* Observations 67 67 67 67 67 67 R-squared 0.00 0.06 0.01 0.15 0.07 0.52 Table 5A: The Effect of th KKK on Political Vote Shares, Pennsylvania Notes: The dependent variable in all cases is the percent change in the Republican vote share for the U.S. House of Representatives in a county across elections. Virtually all Klan supported candidates in the North were Republican. The key dependent variable in the specifications is the percent of the eligible county residents in the Klan at its peak in 1924. Klan membership began to grow rapidly beginning in 1921 and then quickly declined after 1924. Standard errors in parentheses. One asterisk denotes significance at the 5 percent level; two asterisks reflect significance at the 1 percent level. See the data appendix for further details on the data sources and construction. The unit of observation is a county in Pennsylvania. (1) (2) (3) (4) (5) (6) Variable: % change in republican vote 1920-1924 % change in republican vote 1920-1924 % change in republican vote 1924-1926 % change in republican vote 1924-1926 % change in republican vote 1924-1934 % change in republican vote 1924-1934 % Klan in 1924 0.1186 0.2182 -0.0541 -0.0932 0.0450 -0.1584 (0.0694) (0.0597)** (0.0706) (0.0693) (0.1088) (0.0749)* log (county pop in 1920) 0.0169 -0.0586 0.0194 (0.0143) (0.0166)** (0.0179) log (black pop in 1920) -0.0036 0.0101 -0.0038 (0.0036) (0.0042)* (0.0045) log (foreign pop in 1920) 0.0125 0.0088 -0.0390 (0.0059)* (0.0069) (0.0074)** Observations 84 82 84 82 84 82 R-squared 0.03 0.42 0.01 0.24 0.00 0.61 Table 5B: The Effect of th KKK on Political Vote Shares, Indiana Notes: The dependent variable in all cases is the change in the Republican vote share for the U.S. House of Representatives in a county across elections. Virtually all Klan supported candidates in the North were Republican. The key dependent variable in the specifications is the percent of the eligible county residents in the Klan at its peak in 1924. Klan membership began to grow rapidly beginning in 1921 and then quickly declined after 1924. Standard errors in parentheses. One asterisk denotes significance at the 5 percent level; two asterisks reflect significance at the 1 percent level. See the data appendix for further details on the data sources and construction. The unit of observation is a county in Indiana. Collection Name Location Description Size Ku Klux Klan collection (ca. 1915- [ongoing]) Atlanta History Center, Atlanta GA No membership roles or finances 2 boxes (.5 cu.ft.) Ohio Knights of the KKK, Wood County Bowling Green State U Detailed Membership Records very large Ku Klux Klan (1915- ) CALIFORNIA STATE UNIV, NORTHRIDGE Correspondence, Newspapers, and Pamphlets 33 items (1 box) ; 28 cm. Ku Klux Klan (Newaygo County, Mich.), (Mecosta County, Mich.), Calvin Enders Collection. CLARKE HIST LIBR Membership Cards 3.5 cubic ft. Ku Klux Klan (1915-) Colorado Historical Society, Stephen H. Hart Library, Denver CO One volume of members and applications for membership 3 linear ft. (1 box, 2 oversize volumes). Collection of miscellaneous papers, DUKE UNIV LIBR No membership roles or finances 146 items. Ku Klux Klan East Carolina University, J. Y. Joyner Library, Special Collections, Greenville NC Klan Publications 0.073 cu. ft. (6 items) Ku Klux Klan. No. 30 Butte, Montana EASTERN WASHINGTON STATE HIST Membership and financial records 2 ft. Ku Klux Klan (1915- ) Knox County Klan No. 14, Knoxville, Tenn. Emory University, Robert W. Woodruff Library, Special Collections Department, Atlanta GA Membership and Dues Records 288 items (8 linear feet) Ku Klux Klan (1915- ), Crown Point Klan #72 (Crown Point, Ind.), Whitewater Klan #60 (Richmond, Ind.), Local Officers, Indiana, Odon Unit, Klan 90, Logansport, Indiana Indiana Historical Society Membersip and financial records 4 microfilm Reels Ku Klux Klan (1915- ). Wheatland Klan No. 29 (Harlowton, Mont.) Montana Historical Society, Library and Archives Dept., Helena MT Membership Records .1 linear ft. Ku Klux Klan (1915- ). Ohio Knights Ohio Historical Society, Archives-Library Division, Columbus OH Correspondence of Brown Harwood 1/6 ft. Appendix Table 1: Results From our Search for Klan Archives and Special Collections Ku Klux Klan (1915- ). La Grande, Or., Chapter No. 14 Oregon Historical Society, Library, Portland OR Minutes of meetings relating to Klans Involvement in politics 190 items Pennsylvania State Police Collection, 30.16 – 30.19 Pennsylvania State Archive Membership and financial records 30 microfilm reels Ku Klux Klan (1915- ). Pond Creek Klan, no. 117 (Ky.) UNIV OF KENTUCKY LIBR Membership Records 1 microfilm reel Ku Klux Klan Women's Organization (Okla.) UNIV OF OKLAHOMA Oklahoma Women's group from 1924-1928 .66 ft. Women of the Ku Klux Klan. Klan 14 (Chippewa Falls, Wis.). UNIV OF WISCONSIN, EAU CLAIRE Minutes of meetings 0.2 c.f. (5 folders) Ku Klux Klan (1915- ). Athens Klan #5 (Athens, Ga.) University of Georgia, Libraries, Hargrett Rare Book and Manuscript Library, Athens GA Membership and financial records 2.5 linear ft., 3068 items Ku Klux Klan. Klan No. 51, Mt. Rainier, Maryland University of Maryland, Archives and Manuscripts Department, College Park MD Minutes 0.5 ft. or 5 ft. (two listings differ) Ku Klux Klan ((1915- )) University of Memphis, Special Collections / Mississippi Valley Collection, McWherter Library, Memphis TN Miscellaneous papers 1 box Ku Klux Klan (1915- ) Tillamook, Or., Chapter No. 8 University of Oregon, Knight Library, Special Collections, Eugene OR Membership records 1.25 linear ft. (4 containers) Ku Klux Klan University of Tulsa, McFarlin Library, Tulsa OK Membership list 140 (11x17) printed pages Bayfield Ku Klux Klan Collection Fort Lewis College, Colorado Membership Records .8 linear ft. Klan Figures and Tables 2.21.11 final.pdf klanpaperfigs 9.11.07.pdf Slide Number 1 Slide Number 2 Slide Number 3 Slide Number 4 Slide Number 5 Slide Number 6 KlanPaperTables 2.21.11 Table 1 - Summary Stats Table 2 - Revenues Table 3 - Regressions Table 4a - Migration, Penn Table 4b - Migration, Indiana Table 5a - Politics, Penn Table 5b - Politics, Ind A Table 1