PROJECT ABSTRACT

Racial rioting has long been an important issue of intellectual inquiry in sociology and the sub-field of social movements and collective behavior. The proposed project seeks to make an important contribution to this line of research by providing a new and considerably more comprehensive tabulation of racial riots that occurred in the late 1960s and early 1970s. Specifically, event information will be extracted from news clipping files recently recovered from the Lemberg Center for the Study of Violence, the premier riot research center operating during the riot era. The new data set will address several shortcomings in prior riot data that cause them to be incomplete. First, all cities in the US are included in the data. Second, all racial riots, irrespective of who initiated the violence or whether the riots occurred in a school, campus, or urban environment, are included. Third, multiple newspaper sources reduce event size-related bias and distance-related bias in the data set.

A more comprehensive riot data set will be particularly useful at the present time because riot research has experienced a revival in recent years. New statistical techniques and well as new theories focusing on competition processes and diffusion promise substantial advances over past work. However, incomplete cataloging of riot events continues to hamper progress. The proposed project will fill that gap and allow not only a reappraisal of past work, but also detection of patterns in riot cycles that remain uncovered at the present time.

BACKGROUND

Although the race riots of the 1960s and 1970s have been a key focus of empirical research and theory development for years (Spilerman 1970; 1971; 1976; Jiobu 1971; Carter 1983; 1986; 1990; Myers 1996; 1997 Olzak and Shanahan 1996; Olzak, Shanahan, and McEneaney 1997), the data supporting these analyses has been considerably less than ideal. In most cases, data were tabulated from a single media source (such as the New York Times) and supplemented with a motley conglomeration of other sources including the Congressional Quarterly's Civil Disorder Chronology, the Report of the National Advisory Commission on Civil Disorders, and the U. S. Senate's compilation reported in Riots, Civil, and Criminal Disorders. These data compilations are now known to be incomplete and biased, and therefore the theory and empirical work built on them is suspect.

Three major problems exist with respect to most of these data. The first is that the sources used to compile the data exhibit various kinds of selection bias. For example, many newspaper-derived data on protests and demonstrations have been shown to under-represent smaller events and events farther from where the newspaper(s) is (are) produced (McCarthy, McPhail, and Smith 1996; Mueller 1997; Snyder and Kelly 1977; Oliver and Myers 1998b).

Second, many data sets were purposely sanitized to make the events within them more homogeneous. For example, Spilerman (1970; 1971; 1976), Carter (1983), Myers (1997), and Olzak and Shanahan (1996) all eliminated riots that were focused on institutional conflicts (such as those in schools and union halls) and those that were not spontaneous (those that arose as a result of a planned protest or demonstration). Furthermore, only incidents involving primarily Black aggression were included in most of the data thereby ignoring riots in which the action was initiated by some other racial group. These decisions seemed reasonable because the studies were limited in scope, but these data can not legitimately be used for more general purposes.

Third, most riot data has been explicitly limited to large cities with large Black populations. Rather than substantive or theoretical concerns driving these decisions, analysts usually chose this path for practical reasons--namely, the easy availability of economic and structural information for large cities. The result, however, was that many riots were either ignored or undetected. For example, one of the more inclusive sets of criteria was used by Spilerman (1970; 1976): A city was included in the study if the total population was 25,000 or greater and the Black population was 1,000 or greater. Spilerman argued that the critical mass of individuals necessary to produce a riot did not exist in cities that did not meet these criteria. We now know this claim was wrong. In Carter's (1983) data, for example, he used the same definition of a riot that Spilerman used and located 752 riots: Of these, 136 (18%) occurred in cities that did not meet Spilerman's criteria.

Table 1 details the characteristics of several important riot compilations. The final entry in the table is a preliminary tabulation constructed following a method parallel to that proposed herein. It is apparent that all of the other data compilations are seriously incomplete. Across only a three year period, 1967-1969, we located 1357 riot events. Even the most complete (Carter 1983) of the others contains only 752 events and this is over an eight year period (1964-1971). According to our tabulation, 458 cities experienced at least one riot from 1967-69 compared to the 313 riots cities in Carter's longer-term study.

 Table 1: Sample Characteristics of Key Prior City-Level Analyses of Racial Rioting in the 1960s

Author(s)	Temporal Scope	Number of Cities Examined (Number that experienced riots)	City Population Lower Limit	Black (or Non-white) Population Lower Limit	Riot Size Criterion	Number of Riots
Carter 1983; Myers 1996	1964-1971	313 (313)	no limit	no limit	30 or more participants	752
Spilerman 1970; Spilerman 1976; Myers 1997	1961-1968	410 (169)	25,000	1,000	30 or more participants	341
Jiobu 1971	1965-1969	74 (61)	100,000	Black population at least 85% of non-whites	Not stated	162
Lieske 1978b	1967-1969	119 (93)	50,000	1,000	4 or more participants	334
Olzak and Shanahan 1996	1954-1993	204 (86)	204 largest cities	no limit	"large scale" (most had at least 50 participants)	249
Olzak, Shanahan, and McEneaney 1996	1960-1993	55 (43)	Sample of large SMSAs for which all covariate information was available		50 or more participants	154
Preliminary Results using the proposed approach	1967-1969	No limit: All cities in the US included (458)	No limit	No Limit	4 or more participants	1357

Some of the differences in these riot counts stems from legitimate differences in the definition of a riot. While most riot analysts have similar notions about the violence and damage that constitutes a riot, the number of participants required differs quite widely. Again, these choices have been made primarily as a matter of the practicality rather than substance. Because smaller riots are more likely to be neglected by the press, the analyst can improve the relative completeness of the sample by limiting the focus of the study to only large events (Snyder and Kelly 1977). Because most analysts did not have effective ways of tracking small events, focusing on large events was viewed as a strategy to reduce selection bias.

Other reasons for omitting events are even less compelling. For example, the elimination of school and campus riots is particularly significant. In our preliminary study of 1967-1969 we identified 526 riots (39% of the total) that occurred in educational settings, all of which would have been eliminated in prior studies. Even if college campus riots do not reflect local conditions, riots and racial tensions in secondary schools likely do, and thus the omission is unjustified.

Unfortunately, these limited-scope studies have often been touted as the comprehensive word on the riots (e.g., McPhail 1994), but because we now know that the samples behind these studies are incomplete and biased, the empirical conclusions and theoretical advances based on them are suspect and must be re-examined. To do so, a more complete riot tabulation must be constructed.

DATA SOURCE

During the years 1966-1974, a research center called the Lemberg Center for the Study of Violence operated at Brandeis University. The singular focus of the research center was the ongoing racial rioting that engulfed the country during that period. Although the center engaged in several research lines and methods, one of the most important was their attempt to tabulate civil disorders. To create this catalog, the center employed a national news clipping service that monitored most of the local newspapers in the country. The center staff further filtered and processed the clippings, and organized them into event order. For two years, 1968 and 1969, the center published tabulations of riot events. In addition, clippings were collected and organized for 1967, and 1970-1972, but compilations were not created or published.

After the Lemberg Center closed in 1974, its archives seemed to have disappeared and were presumed lost. In early 1998, I located these archives in the library of Manchester College, and with the assistant of the Notre Dame Archives, negotiated a transfer of the materials to Notre Dame. Along with the other components of the Center's research, the complete clipping files are now on site here at Notre Dame.

Using these clipping files to produce a catalog of riots defeats or substantially reduces the problems related to prior riot tabulations. First, because the articles are drawn from local newspapers all over the country rather than a single source like the New York Times, distance-related bias is all but eliminated. Second, using local newspapers also substantially reduces size bias because while a small event in Charlotte, North Carolina, for example, is not likely to be reported in the New York Times, it is much less likely to be missed by the Charlotte Observer. Third, the Lemberg Center made no systematic attempt to sanitize the data by eliminating certain kinds of events or events that occurred in specific contexts. As a result, the data contains a huge number of previously neglected school-related riots. Finally, the Lemberg Center purposely chose a low threshold for the number of participants to provide the most comprehensive tabulation of events possible. Because the strategy of using local newspapers reduces the size-related bias that generated the large-riot strategy, there is no particular reason to focus solely on large events (Baskin et al. 1971).

METHODOLOGY

The processing of the news clippings into usable data will proceed through three phases under this project: (1) narrative coding of information from newspaper articles, (2) numerical coding of information from the narrative coding sheets, (3) entry of numerical codes into a machine-readable data file.

The news-clippings contained in the Lemberg Center Files are arranged such that all newspaper articles related to a single incident are filed in a single folder. To process the news-clippings in the first phase of data collection, coders will read all articles associated with a single event and enter information on the Narrative Coding Form. This form requires information about the time and location of the event, the newspaper of record, a series of characteristics of the event such as duration, number arrested, and property damage, and information about the outcomes of the event. If the outcomes of or follow-up to the event is extensive, a supplementary outcome form is also completed. The Narrative Coding Form follows a format established by Professor Susan Olzak of Stanford University, a leading scholar of protests and violence who has made extraordinarily good use of newspaper records of collective events (Olzak 1992; 1989; 1987; Olzak and Shanahan 1996). Professor Olzak has also made extensive use of undergraduate research assistants to perform these kinds of research tasks, and I have consulted with her on structuring the proposed project. Our pilot efforts here at Notre Dame also indicate that when properly trained, undergraduate students are highly capable, accurate, and reliable coders of newspaper data.

The second phase of data collection involves translating the narrative data recorded on the Narrative Coding Form to a second form that will actually be used in the data entry process. Using coding guidelines, coders will select the appropriate numerical code reflecting the narrative information and record it on the Data Entry Form. This numerical coding system is based on a detailed system developed by Clark McPhail of the University of Illinois and John McCarthy of Penn State University for recording information about collective events and crowd behavior (McPhail and McCarthy, 1993; 1995; Schweingruber and McPhail 1995). Variants of their coding system has been used in several NSF-sponsored projects in recent years.

The final phase involves entering data into pre-formatted data entry forms. The Laboratory for Social Research has agreed to construct these data-entry forms for the proposed project.

Based on pilot tests, we estimate that completing the entire run of events from 1970-1972 will require 4000 person hours. Those participating in the pilot procedure contributed approximately 600 hours of the total. We anticipate that another 600 hours will be completed by the end of the academic year, leaving 2800 hours worth to be completed. Through this proposal, I am asking for support of 1050 hours, which (in conjunction with a sociology department graduate assistant) will allow completion of 1970 and 1971. I am pursuing other avenues of support for cataloging the remainder of the clippings as well as processing other components of the Lemberg Archives.

DISCUSSION

Riot research has experienced a recent revival among social movement and collective behavior scholars. This return to riots has come about for several different reasons. One primary impetus was the unrest that followed the Rodney King incident (Useem 1997). Perhaps more important though, developments in research methods and theory have permitted a critical reconsideration of earlier riot work. In terms of theory, several scholars have begun to apply ethnic competition theory to racial rioting (Olzak and Shanahan 1996; Olzak, Shanahan, and McEneaney 1996; Myers 1997; Bergesen and Herman 1997; Shanahan and Olzak 1998). In these studies, scholars have argued that competition over key resources (jobs, money, and territory) contributes to rioting such that economic down-turns, migration, and desegregation interact to produce inter-group violence. Given that earlier riot studies were unable to find substantial effects consistent with virtually any theory of rioting, the support found for the competition model in initial analyses is extremely important.

The second theoretical development of note is growing attention to diffusion processes in protest and collective violence. These models of collective action "waves" posit that earlier protests change the likelihood of additional protest in the future. A series of important advances in statistical modeling of diffusion processes has allowed direct testing of diffusion notions in ways that were previously impossible (Strang and Tuma 1993; Strang and Soule 1998; Myers 1997; Myers 1996). As a result, diffusion processes have come to the forefront of protest research and some even claim that diffusion will provide a unifying framework for disparate strands of theory that have accumulated over the past 30 years (Oliver and Myers 1998a). Initial work on the diffusion of riots has shattered past views of riot diffusion. While many have claimed that no contagion existed between riots (e.g., Spilerman 1970), these advanced techniques have revealed that contagion/diffusion was not only present, but was an extremely powerful force in determining when and where riots would occur (Myers 1996; 1997).

These new analyses have their own shortcoming though. Although competition theory, diffusion notions, and new statistical techniques all show great promise for understanding riots, the new analyses still depend on the same incomplete data used in the past (or they use newly-collected data with the same shortcomings). The incompleteness of this data is troublesome enough for testing competition theory arguments, but is even more problematic for diffusion analysis. If we are to understand the flow of rioting, we must have a quite comprehensive tabulation of events. Without a comprehensive catalog, diffusion effects are likely to be underestimated or perhaps not even detected. For example, I have shown in recent work that smaller riots tend to flow from larger riots and that riots in smaller cities tend to flow from riots in larger cities. If we were to test for these relationships on a larger scale using an existing riot data set, we would probably miss this effect because small riots and riots in small cities are left out of these data. In addition, rioting in schools and on campuses might result (in part) from the contagion of urban riots (or the other way around). Obviously, these kinds of effects cannot be traced if school and campus riots are not included in the data.

In conclusion then, the riot data I propose to collect will make an extraordinarily important contribution to the developing literature on, and understanding of, collective violence. I plan several detailed analyses of the data, each of which will produce an article appropriate for a main-line sociology journal. Should I procure funding to process the other parts of the Lemberg Archive, I plan to write a book developing a diffusion-based theory of violent protest that draws heavily on all the data in the archive. I have identified several target foundations (NSF, Guggenheim, MacArthur, and NIMH) for the larger project and will be developing proposals for those grant targets in the months ahead. Further progress with the news-clipping portion of the archive will help to demonstrate the viability and value of the Lemberg Data and should bolster my efforts to secure outside funding.

SCHEDULE

Previous Phases Completed:

A. Collection of basic riot characteristics including location, timing, and severity indicators for 1,357 race-related riot events in the United States occurring from 1-1-67 through 12-31-69. Compilations created by the Lemberg Center for 1967-1969 were the sources of these data. Completed in August of 1998.

B. Negotiation of the transfer of the Lemberg Center riot news clipping archives. An agreement was struck between Brandeis University, Manchester College, and Notre Dame in September of 1998 and the transfer was completed in October of 1998. (Other parts of the data archive were also retrieved but are not the focus of this proposal.)

Phases Currently in Progress:

C. Reorganization of the Lemberg news clippings into a consistent sequence.

D. Continuing collection of basic riot characteristics including location, timing, and severity information for all events from 1-1-70 through 12-31-72. Approximately 15% has been completed.

Phases to be Funded under the Current Proposal:

Sept. 1 – Dec. 31, 1999:

1. Continuing collection of basic riot characteristics including location, timing, and severity information for all events from 1-1-70 through 12-31-72. We anticipate completing 1970 during this period.

2. Begin numerical coding of narrative data.

Jan. 1 – May 31, 2000:

3. Continued narrative and numerical coding. We anticipate completing 1971 during this period.

4. Begin entry of numerical data into machine-readable form.

June 1 – August 15, 2000 :

5. Complete data entry for 1970 and 1971.

6. Combined data for 1970 and 1971 with previously collected data for 1967-1969. Perform initial statistical analysis and begin drafting journal articles documenting initial findings.

Bibliography