Racial Rioting in the 1960S: An Event History Analysis of Local Conditions Author(s): Daniel J. Myers Reviewed work(s): Source: American Sociological Review, Vol. 62, No. 1 (Feb., 1997), pp. 94-112 Published by: American Sociological Association Stable URL: http://www.jstor.org/stable/2657454 . Accessed: 21/01/2013 18:26
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp
. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].
.
American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to American Sociological Review.
http://www.jstor.org
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions RACIAL RIOTING IN THE 1960S: AN EVENT HISTORY ANALYSIS OF LOCAL CONDITIONS*
Daniel J. Myers University of Wisconsin, Madison
Using Spilerman's (1970b) data on the timing and locations of race riots from 1961 to 1968, I use event history analysis to investigate the effects of local conditions on city-level hazard rates of rioting. First, several struc- tural strain and social-psychological arguments, which were the focus of Spilerman's original studies, are reexamined in light of event history analy- sis. Second, I reinterpret ethnic competition arguments recently used to ex- plain a wide variety of collective violence, and apply these ideas to the 1960s' riots. Third, I test two models of the diffusion of rioting. Like previous analyses, my findings fail to support structural strain theories. Contrary to previous analyses however, the size of the non-White population is not the singular predictor found for the 1960s' riot locations. Instead, my results support both competition and diffusion arguments.
D uring the 1970s, Spilerman(1970a, States, were better predictors of riots and 1971, 1972, 1976) published an influ- their severity than the 16 theoretical indica- ential set of studies examining racial rioting tors combined. in the 1960s. For many, this research became Despite the apparentclarity of Spilerman's the definitive word on the 1960s riots, and it findings, scholars did not abandon the theo- remains influential today (McPhail 1994). retical constructs he tested. Much subsequent Spilerman examined a number of theoretical research has attempted to challenge his find- explanations for the occurrence and severity ings and revive older explanations of the of riots, including structuralstrain arguments 1960s riots (for reviews, see Bryan 1979; (Smelser 1962), absolute deprivation (Dow- McPhail 1994; McPhail and Wohlstein 1983; nes 1968), relative deprivation (Gurr 1968, Schneider 1992). Despite slight deviations, 1970), unresponsive political structures(Lie- however, the results of these studies gener- berson and Silverman 1965), and geographi- ally support Spilerman's findings: The size cal contagion (National Advisory Commis- of the non-White population accounted for sion on Civil Disorders 1968). In summary, the majority of the variance of riot frequency he concluded that differences among cities explained, and other theoretical variables ex- were unrelated to the frequency and severity plained only small increments beyond it. of rioting. In fact, only two variables, the Ultimately, however, there is still a consid- size of the non-White population and a erable lack of agreement about what factors dummy variable for region in the United should remain under consideration as poten- tial causes of rioting. Some scholars con- * Direct correspondence to Daniel J. Myers, clude that structural strain explanations and Department of Sociology, 1180 Observatory other social-psychological explanations have Drive, Madison, WI 53706 ([email protected]. been thoroughly trounced and that new ap- edu). I thank Pamela Oliver, Gerald Marwell, proaches must be developed to identify the Charles Halaby, Seymour Spilerman, Robert underlying determinantsof rioting and other Mare, Alberto Palloni, Ronald Wohlstein, Ralph collective behaviors (McPhail 1994). Others Turner, Alexandra Corning, and the ASR review- ers for comments on earlier drafts. Parts of this maintain the worthiness of grievance and paper were presented at the 1995 annual meeting deprivation explanations and continue to in- of the American Sociological Association in voke slightly recast versions of these theo- Washington, D.C. ries (Heskin 1985; Koomen and Frankel
94 American Sociological Review, 1997, Vol. 62 (February:94- 12)
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions RACIAL RIOTING IN THE 1960S 95
1992; Polletta 1992). These ideas may con- sense of relative deprivation by comparing tinue to be popular in part because of the their situation to that of proximal Whites. ambiguity resulting from different samples Spilerman'sthird variablecluster tested the and methods, or simply because "it is very "expectations"argument (Berkowitz 1968) in difficult to abandon old explanations" (Mc- which expectations for improved social and Phail 1994:5). However, the continued at- economic conditions increase as the disad- tractiveness of these previous explanations vantaged group's position improves, thereby may hint that the empirical examinations that leading to heightened frustrationif the rising tested them were inadequate analyses of the expectations are not met. In this argument, processes underlying rioting. This question the opposite relationship between deprivation can be answered by critically examining pre- variables and racial rioting is hypothesized: vious empirical work using new analytical As conditions improvefor Blacks, an increase procedures and recent theoretical develop- in expectations, frustration, and ultimately ments that are better suited to the problem. the number of riot events occurs. Thus, I reexamine Spilerman's data on ra- Finally, Spilerman tested the idea that the cial rioting in the 1960s using event history lack of access to political representation in- analysis. First, I closely follow his analysis, creases rioting (Lieberson and Silverman re-examining his conclusions regarding the 1965): With no way to address grievances or structural strain and deprivation arguments. to have group interests represented in gov- Then, I investigate two additional explana- ernmental decision-making bodies, Blacks tions for rioting-one based on notions of may turn to violence as a method of express- social diffusion and the other based on ing their demands. As noted above, Spiler- Olzak's (1987, 1992) recent work on ethnic man (1970a) found little or no support for competition. any of these argumentsin his initial work, in which the frequency of rioting was the de- pendent variable. When riot severity was the BACKGROUND dependent variable and also failed to show these Structural Strain and Deprivation support (Spilerman 1976), ideas were further discredited. Spilerman (1970a, 1971, 1976) conducted his In addition to the original four arguments, analysis of the outbreak and severity of riots Spilerman (1970a) also noted several other by grouping structuralstrain and other related potential sources of the tendency to riot, but structural explanations into four basic clus- he provided only partial tests. Of these, I ex- ters and adopting indices for each cluster of amine competition and diffusion. arguments (Spilerman's variable clusters ap- pear in the stub of Table 2 on page 103). His Models first variable cluster tested the "social disor- Competition ganization thesis"-that poorly integratedin- Competition for scarce resources is often as- dividuals are outside the general control of sumed to be a key process underlying most community norms and have less access to tra- conflicts. When two or more groups compete ditional mechanisms for addressing social for a common pool of resources, they inevi- grievances (Downes 1968). Spilerman's sec- tably come into conflict, and as the demand ond and third clusters of variables examined for a good increases and the supply de- the "absolute"and "relative"versions of dep- creases, the conflict will increase. Park rivation arguments.The absolute deprivation (1950) applied such an economic model to argument posits that the most severely dis- ethnic and racial conflict. His faith, however, possessed in society will be the ones who that assimilation would in time relieve eth- engage in collective violence (Downes 1968). nic conflict and lead to peaceful accommo- The relative deprivation argument (Gurr dation has not been substantiated; thus, the 1968, 1970) states that a disadvantagedgroup continued application of economic competi- determines its own level of social and eco- tion models to ethnic conflict in general and nomic deprivation by comparing itself to ethnic collective violence in particular(Lieb- some reference group; in the case of erson and Silverman 1965; Nielsen 1985; Spilerman's studies, Blacks would develop a Olzak 1987, 1992; Olzak and Nagel 1986;
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions 96 AMERICAN SOCIOLOGICAL REVIEW
Shorter and Tilly 1974; Spilerman 1970a, Black-initiated violence is more easily under- 1971, 1976; Tilly, Tilly, and Tilly 1975; stood under such circumstances. Tolnay and Beck 1995). Using the above insight, I treat non-White Olzak's (1992) recent work offers a prom- unemploymentas an indicator of competitive ising model that connects economic pro- labor market outcomes that disadvantage cesses with collective action using ethnic Blacks. I also incorporate variables indicat- competition as its main mechanism. Her ing increased competition in the traditionally study of ethnic conflict and protest supports Black labor market stemming from non- three main competition mechanisms: (1) The White migration to a city and all immigra- breakdown of labor market segmentation tion to a city from outside the United States. leads to increased competition between eth- Both of these variables are particularly rel- nic groups, thereby leading to increased con- evant to the 1960s because of high levels of flict and collective action; (2) immigration interstate migration of Blacks and increasing increases competition directly (Bonacich immigration throughout the 1950s and 1972; Lieberson 1980) and also through its 1960s. And because competition theory pre- effects on labor marketsegmentation; and (3) dicts that economic downturns exacerbate economic contraction exacerbates competi- any competition-driven ethnic violence, I in- tion and conflict as it increases competition clude indicators of general economic con- for employment. tractions in the analysis. Finally, I examine While Olzak's analysis is compelling, the interactions between immigration and eco- collective violence she and others attempt to nomic conditions implied by competition ar- explain is predominantly that of Whites guments. Intuitively, the effect of any in-mi- against ethnic and racial minorities. Rioting gration should be greater when the economic in the 1960s, however, was an exception to situation is poor in a particular city (Olzak that pattern. Therefore, if ethnic competition 1989b). Conversely, an influx of foreign im- arguments are to be used to explain the migrants or American non-Whites should Black-initiated property riots of the 1960s, have a smaller effect on competition and sub- the arguments must be congruent with mi- sequent violence when the economy is far- nority-initiated as well as majority-initiated ing well.' collective violence. I propose that while changes in competi- tion may in fact change the levels of collec- Diffusion of Rioting tive violence, predicting who will initiate the Diffusion refers to a process in which the oc- violence and who will be the target of vio- currence of an event to one member of a lence requires knowledge of the outcomes of population changes the likelihood of such an increased competition. In the cases examined event occurring to other members of the by Olzak, White-initiated violence in re- population (Strang 1991). So when a riot oc- sponse to labor market de-segregation makes curs in one city in the United States, the rates intuitive sense because White advantage is of future rioting in other cities are hypoth- threatenedby desegregation. However,apply- esized to increase as a result. Many analysts ing the labor market desegregation argument have recognized that collective protests and to Black riots is problematic because it is not collective violence tend to cluster both tem- clear that Blacks would respond to improved porally and geographically (McAdam 1982; market positions with violence against Pitcher, Hamblin, and Miller 1978; Olzak Whites. Yet it is possible that decreased labor 1987; Koopmans 1993; Lichbach 1985; Diek- market segregation could lead to violence by mann 1979). This is not imitation or mindless Blacks if desegregation led to worse market copying, (as was the view in early contagion positions for Blacks. Consider the possibility theories (see McPhail 1991 for a review). In- that desegregation resulted from White inva- stead, it is viewed as a social learning process sion of labor niches traditionally dominated in which the favorable outcomes of others' by Blacks; the result for some Blacks thus would be unemployment. Rather than an im- I For a complementary analysis of racial riot- proved labor market position for Blacks, de- ing that also supports competition theory, see segregation would have the opposite effect. Olzak and Shanahan (1996).
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions RACIAL RIOTING IN THE 1960S 97 collective behaviors provides a model for spatial proximity is usually indicated by geo- subsequent protests (Pitcher et al. 1978). graphic distance (Hedstrom 1994; Spilerman The process of social diffusion is driven by 1970a), and temporal proximity is indicated a variety of communication processes among by the amount of time that has passed since established social networks (Hamblin, Jacob- a collective event (Olzak 1987, 1992). I use sen, and Miller 1973; McPhail and Wohlstein spatial and temporal proximity to create 1983; Paige 1971). These networks may be three diffusion indicators. defined by friendship and family connec- The first diffusion variable is a declining tions, or they may be more dispersed, as function of geographical distance between a when individuals are connected through the given city and each city that experienced a mass media (Oliver 1989). For example, riot at time t - 1. The amount of contagion, Rude (1964, 1972) documented that rebel- c,, experienced by a given city at time t is lion in England and France during the eigh- thus given by: teenth and nineteenth centuries diffused along major transportationroutes suggesting ct =Xd7mti , (1) that information about collective actions was j being communicated by travelers along trade where di1 is the distance in miles between routes. More recently, the mass media and two cities, i and j, and m is a dummy vari- the telephone have been considered to be the able indicating whether or not a riot occurred most prominent communication modes driv- in city j at time t - 1. This contagion variable ing the diffusion of collective behavior tests the hypothesis that riots in one city in- (Spilerman 1970a; Morris 1984). Clearly, the crease the likelihood of rioting in other cit- telephone is more apt to be used by networks ies and that other riots are more likely in cit- of acquaintances, while the mass media has ies closest to where the original rioting oc- a more widely dispersed influence governed curred. in scope by its audience (i.e., a national or Given that diffusion is driven by commu- local audience) (Oliver 1989). nication processes, the media in particular, I assume that these communication net- an alternative pattern of diffusion may have works are geographically concentrated. operated in the 1960's rioting. Given the sub- Friends and family (and thus their face-to- stantial national media coverage of the riots, face and telephone contacts) are likely to be particularly of large riots, rioting may have geographically concentrated. And, although diffused nationally rather than regionally. In some riots receive national media attention, other words, the proximity of cities may have media coverage of riots and civil violence is been irrelevant if all cities nationally re- generally concentrated in the regions where ceived equivalent information about the riots they occurred. Thus, while a large riot in De- as they occurred. To test for this possibility, troit may receive national newspaper and I examined two additional diffusion vari- television coverage, news broadcasters and ables. The first variable measures the num- newspaper editors in and around Michigan ber of riots in the United States at time t - 1 would devote higher proportionsof their cov- (M,_1 where Mt-1 = Ymj1t,). This variable as- erage to the riot than would the media in sumes that a riot in any city has an equal ef- Texas or Florida. Small riots that receive re- fect on other cities, regardless of relative lo- gional media attention may not be covered cations. The effect of the number of riots at nationally and therefore may be completely time t - I may not be linear however. The unknown to people in distant regions. first riot at t - 1 may contribute substantially Any adequate model of a diffusion process to the occurrence of additional riots, but the must identify the underlying factors that ac- tenth riot may contribute little more beyond count for variationin the "contagiousness"of that contributedby the ninth riot or may even a social phenomenon. The two most impor- decrease the likelihood of rioting. To reflect tant of these factors are temporal proximity this possibility, I incorporate the number of and spatial proximity (Hamblin et al. 1973; riots squared, (M,_1)2, in the analysis. Rogers 1983; Brown 1981; Mahajan and Unlike other social diffusion processes, Petersen 1985; Strang and Tuma 1993). In such as the adoption of a new technology the case of riots and other collective events, (Hamblin et al. 1973; Rogers 1983), disrup-
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions 98 AMERICAN SOCIOLOGICAL REVIEW tive protest activities do not continue to dif- not find any city-level differences, he con- fuse unabated. Rather, protest activity ini- cluded that rioting must be the product of a tially accelerates, then peaks, and then be- national phenomenon. Drawing on Tom- gins to decline. In the case of rioting, each linson (1968), Spilerman suggested that the riot subsequent to the apex slows the rate of process was actually driven by nationwide rioting, almost as if the process were becom- conditions that affected all Blacks and by the ing distasteful to its participants. This ex- acceptance of a "riot ideology" among all haustion effect has been hypothesized to re- Blacks; therefore riots would break out ran- sult from either the tiring and calming of the domly, and when and where were governed participants or the increase in repressive only by the number of available rioters. mechanisms by agents of social control (Kel- Thus, the current data cannot provide for ly and Isaac 1984; Koopmans 1993; Ober- comprehensive tests of theories about the schall 1978). Several analyses have sug- causes of racial rioting; but they can be gested that the diffusion and exhaustion pro- more appropriately applied to the question cesses for collective violence require rela- of which cities were more likely to experi- tively brief periods. Olzak (1987) found the ence rioting given that the riot wave was oc- diffusion-exhaustion cycle to be completed curring. Conclusions drawn from this data within a period of 30 days. With regard to must take into account these limitations. the 1960s rioting, the Kerner Commission Thus, my aim is to examine differences in report (National Advisory Commission on local conditions and to determine if these Civil Disorders 1968) suggested that events differences contributed in any systematic clustered in two-week periods. Given these way to the pattern of rioting observed dur- observations and the nature of the data on ing the 1960s.2 1960s riots, I estimate four sets of models using rioting in the past one week, the past DATA two weeks, and the past four weeks. Models using the one-week period were clearly su- The riot data were collected by Spilerman perior; the addition of longer time periods (1970b), who cataloged urban riots in the did not improve prediction over the one- United States for the years 1961 through week model. Therefore, here I present only 1968. He used these in three important stud- models using the one-week measures. ies that tested alternativeexplanations of riot frequency and riot severity (Spilerman 1970a, 1971, 1976). His original data were What Can and Cannot Be Explained drawn from a number of sources, including Every analysis is bounded by the scope of the the Congressional Quarterly's Civil Disorder data examined, and the current study is no Chronology, the New York Times Index, the exception. As were previous studies of the Report of the National Advisory Commission 1960s riots, this analysis is also limited by on Civil Disorders, and the Lemberg two key aspects of the data: their temporal Center's Riot Data Review (for details, see scope and their units of analysis. Because the Spilerman 1970a, 1970b). The 673 cities in- data I examine are limited to the 1960s and cluded in the data are located in the contigu- compare city-level characteristics, they can- ous United States and had populations of at not explain the rise of the unprecedented least 25,000 in 1960. For each city, the wave of Black-initiated rioting that occurred weeks in which a riot occurred are recorded in the 1960s. A different type of data set, from the beginning of 1961 through the end which provides informationon long-term na- of 1968. tional conditions, is necessary to explain the 2 riot wave (see McAdam 1982; Jenkins and Although the data I analyze do not speak to Eckert 1986; and Kelly and Isaac 1984 for long-term cycles of collective behavior, evidence of these is mounting indicating that competition processes empirical investigations long-term contribute to long-term patterns of collective vio- trends and contexts). In fact, it is exactly this lence (Olzak 1992). Olzak's work and the results limitation that led Spilerman (1970a) to con- of the present study suggest that competition ar- jecture about the causes of Black-initiated ri- guments provide strong hypotheses for explain- oting in the 1960s. Given that Spilerman did ing long-term patterns of racial rioting.
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions RACIAL RIOTING IN THE 1960S 99
150 134 141
z 100
0 5 0 ~~~~~~~~~~~~~36
.0 0
1961 1962 1963 1964 1965 1966 1967 1968 Year Figure 1. Distribution of Riots by Year in 410 U.S. Cities: 1961 to 1968
For the current study, I furtherreduced the "spontaneous"or arose from a protest (e.g., set of cities examined by eliminating those a civil rights demonstration).Because Spiler- that had 1,000 or fewer non-Whites in 1960. man was primarily concerned with random My rationale for this procedure was that a rather than coordinated rioting and because critical mass of Blacks was necessary to he wished to reduce heterogeneity among the cause rioting. Obviously, a city with no types of riots, he limited his analysis to spon- Blacks cannot experience a spontaneous riot, taneous Black aggression: This accounted for and including cities with few or no Blacks 341 separate riot events from 1961 to 1968 has a powerful effect on the statistical rela- (see Spilerman 1970b for details). In the in- tionship between non-White population size terest of comparability with Spilerman's and and racial rioting. With no correction, a large subsequent studies and because Spilerman's portion of the relationship between non- documentation suggests that his tabulations White population size and rioting would be of the other types of riots may be incomplete, due to the fact that Blacks must be present I also limit my analysis to spontaneous Black for Black riots to occur. This rather uninter- aggression. Figure 1 presents a tabulation of esting finding would confound subsequent spontaneous Black riots by year in the 410 interpretations of the effects of non-White U.S. cities in the study. population size on rioting. A more interest- The Governmental Units Analysis Data, ing question asks whether increases in the 1960 (Alford and Aiken 1970) provides the non-White population above an essential city-level economic, structural, and political minimum increase the likelihood of rioting. information used as independent variables in Spilerman (1971) provides evidence suggest- both this study and in Spilerman's studies. ing that the critical number of Blacks needed The sources of these data are documented in for rioting in the 1960s was somewhere in Alford and Aiken (1970) and include the excess of 1,000. Therefore, I chose a non- 1960 Census and The Municipal Yearbook. White population of at least 1,000 as the critical criterion for including a city in the study. I eliminated 263 cities by this proce- ANALYSIS dure, and it is not surprising that none of Past Analysis of Racial Rioting these cities experienced a riot during the study period. Since the eruption of racial rioting in the Riots included in Spilerman's (1970b) data 1960s, research methods for examining riot were those involving at least 30 people, that patterns have made important advances. Ef- resulted in violence or the destruction of forts prior to Spilerman's landmark study property, and that were not focused on insti- had employed pairings of similar cities, one tutional conflicts (such as those in schools of which had experienced rioting and another and union halls). Spilerman also classified which had not (Lieberson and Silverman riots into five types according to the race of 1965). As riots occurred in more and more the participants and whether the riot was cities throughoutthe 1960s, adequatepairs of
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions 100 AMERICAN SOCIOLOGICAL REVIEW riot and non-riot cities matched on region have diffusion effects. In short, to draw a le- and population size became increasingly dif- gitimate conclusion, a comprehensive analy- ficult to find. Furthermore,these paired-city sis of the data must incorporate explanatory studies allowed only a dichotomous classifi- variables that change as functions of both cation of cities (riot or non-riot), thereby ig- geographic proximity and time. noring differences among cities in which more than one riot had occurred. In response Event History to these problems, Spilerman (1970a) took a Analysis large step forward when he used multivariate In response to some of the shortcomings of analysis in which the number of riots in a earlier approaches to rioting, I apply survival city was the dependent variable. Since that analysis or event history analysis to the data study and his subsequent examination of riot described above (Tuma and Hannan 1984; severity (Spilerman 1976), most studies have Allison 1984; Yamaguchi 1991). Rather than used some version of multivariate linear re- differentiating cities on the basis of the pres- gression to examine both riot occurrence and ence or the number of riots, event history riot severity (Carter 1986, 1990, 1992; Dot- analysis focuses on the duration of time be- son 1974; Jiobu 1971; McElroy and Singell tween riot events for each individual city. By 1973). approaching the analysis of events in this Despite the advances of the multivariate way, researchers can examine processes that approach, the techniques used were not ide- cluster differently in time and use time-vary- ally suited for examining riot data. The first ing covariates to examine dynamic processes problem arises from using frequencies as the (see Olzak 1989a and 1992 for a complete dependent variable. While this approach in- discussion of using event history analysis to cludes more informationthan does a dichoto- analyze collective action events).3 Recently, mous indicator, it nonetheless represents the Olzak (1989b,1990, 1992; Olzak and West set of riots as more homogenous than it actu- 1991) has successfully applied event history ally is. In particular, any temporal relation- analysis to a number of different types of ship within the riots is ignored. That is, a dif- collective events, including labor strikes, eth- ferent underlying process may operate for ri- nic violence, lynching and the establishment ots that are temporally clustered than for of ethnic newspapers. those that are spread evenly across the time Formally, an event history analysis that period studied. models an event that can occur at any point A second important shortcoming of previ- in continuous time specifies the instanta- ous analyses is their inability to introduce neous rate of transition from one state to an- covariates that change over time, a particu- other; it is defined as larly difficult problem when attempting to ? T
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions RACIAL RIOTING IN THE 1960S 101
I model the hazardof rioting using the pro- assumptions either about the distribution of portional hazards model and estimate param- unobserved characteristics or the functional eters using the method of partial-likelihood form of the baseline hazard; none of which (Cox 1972).4 Parametersare estimated using are tenable for the present analysis. There- an ordinary maximization routine such as the fore, I correct for unobserved heterogeneity Newton-Raphson algorithm by maximizing by incorporatinga control variable: the num- - S. ber of prior riots that have occurred in a city. Including variables that represent the prior I exp[ J&Xik (ti )] j k history of the individual city is a practical PL Y --- procedurewhich minimizes the effects due to IL Xexp[XfPkXk (0)] correlation within the same city and does not demand the restrictive assumptions of more where t1is the time of occurrence of either an general methods (Allison 1984).5 event or censoring, the Plsare coefficients es- timating the effects of the k hypothesized ex- planatory variables, and 3i is a dummy vari- RESULTS able that indicates whether the event associ- Spilerman 's Analysis Revisited: with i was censored = or was ated case (8i 0) Disorganization, Deprivation, and a = 1). The coefficients generated failure (3i Political Structure are most readily understood by interpreting eP (the hazard ratio) as the amount by which Spilerman's (1970a) analysis of riot fre- the hazard is multiplied for each unit increase quency led him to draw a number of conclu- in the related independent variable. sions about propensities to riot. In summary, In event history analysis, the effects of un- he found that non-White population size and observed heterogeneity must be carefully region (south versus non-south) were by far considered (Allison 1984; Yamaguchi 1991). the most powerful predictors of riot fre- Of particular concern here is the lack of in- quency. These two variables subsumed near- dependence among repeated observations ly all the effects of the variables represent- made on the same city that occurs when all ing social disorganization, deprivation, and sources of inter-city dependence are not ac- political structure arguments. He concluded counted for in the estimated models. Because that ". . . the racial disturbances of the 1960s it is impossible to know if all relevant vari- were not responses to conditions in the local ables have been included, unobservedhetero- community," and that non-White population geneity usually introduces a downward bias size was really the only relevant variable: in standarderror estimates. In the data stud- "[T]he larger the Negro population, the ied here, 76 cities experienced more than one greater the likelihood of a disorder. Little spontaneous Black-initiated riot (up to a else appears to matter"(p. 645). maximum of 11 in Chicago), therefore the Tables 1 and 2 present analyses that paral- present analysis is highly susceptible to this lel those conducted by Spilerman (1970a). problem. Table 1 reiterates a portion of Spilerman's Many methods have been suggested to cor- Table 6, reporting the increments in the vari- rect for bias introduced by unobserved het- ance of riot frequency explained. It also pre- erogeneity in survival analysis (Allison 1984; sents an event history analysis analogous to Flinn and Heckman 1982a, 1982b; Heckman Spilerman's. I tested the same variables ac- and Singer 1982, 1984; Yamaguchi 1986; Tuma 1985; Trussell and Richards 1985). Of 5 False duration dependence due to different these, a number of general methods require risk levels among unidentified strata within the sample also can arise in survival analysis. How- ever, this problem is not relevant here because the 4 For the basics of survival analysis and par- approach to repeated events does not allow the tial-likelihood estimation see Tuma and Hannan membership of the population at risk to vary over (1984), Cox (1972), Allison (1984), and Yama- time. Details regarding this issue as well as alter- guchi (1991). Details on the application of event native methods considered for controlling unob- history and partial likelihood to the riot data used served heterogeneity in the riot data are available herein are available from the author. from the author.
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions 102 AMERICAN SOCIOLOGICAL REVIEW
Table 1. A Comparison of the Results of Spilerman's Analysis with Results from Event History Analysis
Regression Analysis of Riot Counts Spilerman (1970a) Event History Analysis (1) (2) (3) (4) Likelihood Likelihood Ratio Test Percent of Percent of Ratio Test Comparing Variance Variance Comparing Model with Explained by Explained by Model with Non-White Non-White Variable Cluster Variable Cluster Population Population when Entered and South to and South to when Entered after Non- Same Model Same Model after Variable White Popu- Adding Non-White Adding Cluster and lation and Population Variable Cluster Variable Cluster a South South x2 (d.f.) x2 (d.f.)
(A) Social disorganization 43.1 2.3 179.32*** (1) 8.33* (3) (B) Absolute deprivation 40.6 .1 187.28*** (1) 2.42 (4) (C) Relative deprivation 27.7 1.3 126.04*** (1) 2.68 (5) (D) Political structure 19.2 .9 137.58*** (1) 12.95* (4)
All variables except 9.3 4.5 70.04*** (1) 30.86* (16) non-White population
a See the variables listed in Table 2 for indices that Spilerman included in each variable cluster. Numbers in parentheses are degrees of freedom. * .05 **< .01 ***< .001 cording to their contribution to the hazard 46.8-percent of the variance was explained rate of rioting; the values reported are incre- by only non-White population and the south ments in chi-squares resulting from a likeli- dummy variable, these findings mean that hood ratio test comparing two models. Al- there is a large overlap among non-White though minor differences exist, an obvious population size, the south dummy, and the parallel occurs between these results and variables in the four clusters. This Spilerman's regression analysis. The incre- multicollinearity is also apparentin the event ment gained by adding non-White population history analysis presented in Table 2. Mod- to each variable cluster is large (columns 1 els run with the individual variable clusters and 3) compared with the increment gained (Models A throughD) show many significant by adding the variable cluster after non- relationships between the theorized vari- White population size (columns 2 and 4). ables. As the models are combined and Furthermore,while two variableclusters pro- Spilerman's key variables are added, nearly duce significant gains when added after the all of the theoretical variables become non- south dummy variable and non-White popu- significant. lation, the improvement is marginal. Due to these patterns of collinearities, re- Despite their seeming clarity, interpreta- sults relying on variance explained or com- tion of these results is problematic because parisons of models using likelihood ratio of collinearities among the variables. In tests fail to provide convincing evidence Spilerman's analysis, 51.3 percent of the against the theories tested or against the no- variance in riot frequency was explained by tion that community conditions contribute to the four variable clusters in Table 1, and an rioting. Rather than interpreting the differ- additional 9.3 percent of the variance was ences in community conditions as negligible, explained by non-White population size. a more reasonable interpretation is that dif- Thus 42 percent of the variance in riot fre- ferences are related to rioting, but that the quency was explained by the four clusters of conditions are related to the size of the non- variables. Taken together with the result that White population. In fact, the relationships
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions RACIAL RIOTING IN THE 1960S 103
Table 2. Partial-Likelihood Estimates Showing the Effects of Social Disorganization, Deprivation, and Political Structure on Hazard Rates for Racial Riots: 410 U.S. Cities, 1961 to 1968
Model Model Model Model Model Model Independent Variable A B C D 1 2
South (vs. non-south) a - 1.17 *** (.304) Log non-White population .610*** (.074) (A) Social Disorganization Percent change in total -.008*** - -.005** -.004* population (.002) (.002) (.002) Percent change non-White .138 .024 .466 population (.699) (.637) (.689) Percent dilapidated housing -.026* -.047** -.021 (.012) (.018) (.019) (B) Absolute Deprivation Percent of non-White males in -.014** -.002 -.002 traditionally Black occupations b (.005) (.006) (.007) Non-White male unemployment .033** - .019 .003 rate (.011) (.017) (.019) Non-White median family income .019* .032* .004 (in 100s of dollars) (.008) (.014) (.017) Non-White median education -.051 -.132 -.119 (.056) (.079) (.085) (C) Relative Deprivation and Socioeconomic Expectations Percent of males in traditionally -.072* -.024 .050 Black occupations C (.033) (.049) (.056) Income - 1.74** -1.42 -.030 (.639) (1.21) (1.30) Unemployment C - .067 -.018 -.090 (.048) (.075) (.085) Education - 1.21* - 1.21 -1.02 (.544) (.684) (.770) Non-White population percentage - .281*** - .268*** --.057 (.038) (.048) (.070) (D) Political Structure Population (in 1,000s) per council -- .007*** .006*** -.002 member (.001) (.001) (.001) Percent of council elected at-large - .001 .001 .002 (.002) (.002) (.002) Presence of non-partisan elections - .507*** .658*** .375* (.129) (.138) (.148) Presence of mayor-council .693*** .632*** .155 government (.137) (.140) (.157) Previous riots (control) ,414*** ,413*** .374*** .279*** .244*** .180*** (.027) (.027) (.028) (.033) (.033) (.033) Model chi-square (d.f.) 224.27 214.57 256.58 257.37 354.92 434.82 (d.f. = 4) (d.f. = 5) (d.f. = 6) (d.f. = 5) (d.f. = 17) (d.f. = 19) Note: Numbers in parentheses under coefficients are standard errors. The four clusters of indices are taken from Spilerman (1970a). a "South" indicates Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Caro- lina, Oklahoma, South Carolina, Tennessee, Texas, Virginia, and West Virginia. b Traditionally Black occupations are service, household, and laborers. C Relative deprivation variables are calculated by dividing figure for non-Whites by figure for Whites. "p n<.05 "n <.01 < .001 (two-tailed tests)
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions 104 AMERICAN SOCIOLOGICAL REVIEW between the non-White population size and to 1960. Both variables increase the level of the various city characteristics are them- competition for jobs in labor market seg- selves of considerable theoretical interest. ments occupied by Blacks and thus are pre- In the end, the comparison reported in dicted to have positive effects on rioting. Table 1 does little to either refute or validate Table 3 presents partial likelihood esti- the theories tested. It is the patternof coeffi- mates of the effects of competition model cients associated with the theoretical vari- variables on the hazard rates of racial riot- ables in Table 2 that provides evidence ing. Models A, B, and C estimate the effects against the structural and deprivation theo- for each respective variable cluster. The chi- ries examined by Spilerman. In each theo- square associated with each of these models retical cluster, some significant coefficients is highly significant (p < .001), although the are in the predicted direction, but others are economic effects on Blacks (Model A) is not. Across the four clusters approximately clearly a more powerful model than the other one-half of the theoretical variables produce two. Although nearly all significant coeffi- coefficients with signs opposite those which cients are in the direction predicted by the the theories would have predicted. Because intergroup competition approach, there is none of the four theories finds reasonable one notable exception: In Model B, the support in the data, alternative explanations manufacturing wage was predicted to de- must be pursued. crease riot risk by alleviating competition, but the data indicate that higher wages actu- ally contributedto an increased risk of racial Competition Models riots. All of the other significant indicators To test intergroup competition perspectives, testing competition theory, including both I adopted variables to indicate three central economic and migration indicators, produced notions of competition models: economic af- coefficients that support the hypotheses. fects of Blacks, general economic factors. Model 1 in Table 3 combines selected and minority group migration. First, I in- variables from Models A, B, and C and of- clude non-White unemployment rates and fers a significant improvement over each raw numbers of non-Whites unemployed as single-cluster, as shown by the likelihood- indicators of labor market competition out- ratio test (p < .001 in each case). The coeffi- comes that affect primarily Blacks. Both cients indicate that each variable cluster variables are predicted to have positive rela- adds explanatory power above that achieved tionships with riot risk. Although non-White by Model A, B, or C alone. Two nonsignifi- unemployment rate is often used in such cant variables, the non-White unemploy- analyses, I introduce the absolute number of ment rate and the change in non-White unemployed non-Whites because it may have population, are dropped from the analyses at different effects in cities where the unem- this point in the interest of parsimonious ployment rate is the same. Larger numbers presentation. Model 2 indicates that drop- of unemployed non-Whites may make the ping these two variables does not result in a unemployed segment of the group more vis- poorer prediction of rioting rates compared ible and thus exacerbate feelings of competi- to Model I (Model I vs. Model 2 in Table 3 tion with other groups. is nonsignificant).6 Therefore I use Model 2 The general economic factors included are as the baseline against which to examine in- the median manufacturing worker's wage teraction effects between economic factors and the overall unemployment rate. Higher and migration factors. manufacturingwages signal a growing econ- While most coefficients in Model 2 remain omy, which according to competition theory consistent with Models A, B, and C, the co- would reduce competition and thereby re- duce intergroup collective conflict. Unem- 6 I conducted alternative analyses carrying for- ployment indicators are predicted to have ward the variables measuring non-White unem- positive effects on rioting. ployment rate and the 1950-1960 change in non- Minority group migration is representedby White population through the remainder of the two variables, percent foreign-born and the models presented in this paper. Neither variable change in non-White population from 1950 achieved significance any model.
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions RACIAL RIOTING IN THE 1960S 105
Table 3. Partial-Likelihood Estimates Showing the Effects of Competition Variables on Hazard Rates for Racial Riots: 410 U.S. Cities, 1961 to 1968
Model Model Model Model Model Model Independent Variable A B C 1 2 3
(A) Economic Effects on Blacks Non-White unemployment rate -.005 .002 (.012) (.014) Ln of number of non-White .497*** .504*** .500*** .629*** unemployed (in lOGOs) (.037) (.039) (.038) (.059)
(B) General Economic Factors Median manufacturing wage .029* .028*** 029*** .0268** (in lOs of dollars) (.007) (.008) (.008) (.008) Unemployment rate - .0838** -.088* -.085* -.090* (.030) (.042) (.035) (.035)
(C) Minority Group Migration Change in non-White population, -.574 .313 1950-1960 (in 1,OOOs) (.628) (.560)
Percent foreign-born - 048*** .029 * .030** .026* (.010) (.01 1) (.01 1) (.01 1) Ln non-White unemployed x - - - -.016** percent foreign-born (.0055) Prior rioting (control) .214*** .430*** .399*** .188*** .188*** .205*** (.031) (.026) (.027) (.032) (.032) (.033)
Model chi-square (d.f.) 359.86 208.25 202.04 387.26 386.96 395.32 (d.f. = 3) (d.f. = 3) (d.f. = 3) (d.f. = 7) (d.f. =5) (d.f. = 6)
Likelihood Ratio Tests Chi-Square d.f. Model A vs. Model 1 27.40** 4 Model B vs. Model 1 179.01*** 4 Model C vs. Model 1 185.22*** 4 Model 1 vs. Model 2 .30 2 Model 2 vs. Model 3 8.36*** 1
Note: Numbers in parentheses under coefficients are standard errors. *p < .05 ** < .01 ***p< .001 (two-tailed tests) efficient for the general unemployment rate singling out White unemployment reveals a becomes negative in Model 2, indicating that completely different effect. lower unemployment leads to increased riot- I also examined a series of interaction ef- ing. It is apparentthat the change in sign re- fects suggested by competition theory. I ex- sults from adding the (In) number of non- amine interactions between economic con- White unemployed to Model B. Thus, the traction and migration because the effect of coefficient for the general unemployment migration on rioting should differ at varying rate in Model 2 represents its independent levels of economic contraction. Specifically, effect when holding non-White unemploy- when the economy is faring poorly, migra- ment constant, an effect that must be due tion into a city should have greater effects on largely to White unemployment. While gen- competition and violence than when the eral unemployment may produce the rela- economy is healthy. Thus, I predict a posi- tionship predicted by competition theory, tive coefficient for interactions between the
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions 106 AMERICAN SOCIOLOGICAL REVIEW unemployment variables and the migration Table 4 presents results supporting diffu- variables, and a negative coefficient for the sion predictions: All coefficients are in the interaction between manufacturing wages predicted directions. In Model 1, the posi- and the migration variables. tive and significant spatial diffusion coeffi- In general, the data do not support the cient indicates that riots in the prior week theory in this respect. Of the three possible increase the likelihood of riots in proximal interactions, two (unemployment rate x per- cities and that this effect diminishes as the cent foreign-born and manufacturingwage x distance increases. In Model 2, the national- percent foreign-born) do not achieve statisti- level diffusion variable also has positive and cal significance (results not shown). Further, significant effects. The number of riots in the one significant interaction term (In num- the previous week increases the likelihood ber of non-Whites unemployed x percent for- of rioting in any given week. Also, the num- eign-born) is in the opposite direction of that ber of riots in the previous week has a di- predicted. The negative coefficient indicates minishing effect as the number of riots in that increases in the percentage of immi- the prior week increases, as evidenced by grants has a lower effect on rioting when the significant and negative coefficient asso- non-White unemployment is high and a ciated with the squared national-level diffu- greater effect when Black unemployment is sion term, (Mt-1)2. low. The coefficients for the model contain- The nature of this particular data set sug- ing this significant interaction term is pre- gests an alternative to the diffusion models I sented in Model 3 in Table 3. Instead of these offer. As is apparentin Figure 1, the vast ma- economic and migration variables compoun- jority of riot events during the study period ding each other, the effect of one appears to occurred during 1967 and 1968. Because of attenuate the effect of the other, thereby sup- the unique character of this period and the pressing the main effects of both. In sum, specific events that occurred then (such as however, the interaction effects I have exam- the Vietnam War and the assassinations of ined contribute little to the prediction or ex- Martin Luther King, Jr. and Robert planation of racial rioting. Kennedy), it is possible that the observed dif- fusion effect is merely a result of the tight clustering of riot events during a relatively Diffusion of Rioting short time period. To test for this possibility, Analyses of three diffusion variables are pre- I introduce a dummy variable indicating sented in Table 4. First, c,, the spatial diffu- whether the time period for each observation sion variable defined in equation 1, is incor- fell in the 1967-1968 time period. Given the porated to model the effects of spatial het- distributionof the riot events across the study erogeneity among the riots occurring in the period, it is clear that the dummy variable previous week. If the diffusion hypothesis is will be highly significant. Interestingly, its correct, this diffusion variable should pro- effects on the diffusion variables are much duce a positive coefficient. That is, recent ri- less dramatic.As Model 3 in Table 4 shows, oting in cities proximal to a given city when the 1967-1968 dummy is entered, the should increase that city's hazard of experi- national diffusion effect decreases slightly encing a riot. Second, the measure for na- but remains highly significant. And there is tional-level diffusion effects (Mti) simply virtually no effect on the spatial diffusion calculates the total number of riots nation- variable. Despite the unique character of the ally in the previous week. Again, if national- 1967-1968 period, clearly the diffusion pro- level diffusion is occurring, a positive coef- cess was operating. ficient will be observed. In addition, the What remains ambiguous to this point is square of the national-level diffusion vari- the relative importance of the competition able represents the tapering effect of addi- and diffusion variables when compared to tional riots over time. A negative coefficient the influence of non-White population size. is expected when this term is added after the Thus, to complete the analysis, the size of the unsquared national-level variable, meaning non-White population must be introduced to that a high number of recent riots would pro- evaluate its effects in conjunction with those duce a diminishing hazard for rioting. derived from competition and diffusion argu-
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions RACIAL RIOTING IN THE 1960S 107
Table 4. Partial-Likelihood Estimates Showing the Effects of Competition and Diffusion Variables on Hazard Rates for Racial Riots: 410 U.S. Cities, 1961 to 1968
Independent Variable Model 1 Model 2 Model 3 Model 4
Ln of number of non-White .617*** .625*** .644*** .482* unemployed (in 1000s) (.059) (.059) (.061) (.203) Median manufacturing wage .026** .024** .021** .024** (in 100s of dollars) (.008) (.008) (.008) (.008) Unemployment rate -.081 * -.101** -. 139*** -.113* (.035) (.036) (.036) (.048) Percent foreign-born .020 .023* .035** .034** (.01 1) (.01 1) (.01 1) (.01 1) Ln non-White unemployed x -.013* -.012* -.004 -.005 percent foreign-born (.006) (.006) (.006) (.006) Spatial diffusion (c,) .034*** .024** .025** .025** (.007) (.008) (.009) (.009) National-level diffusion (M,-1) .295*** .170*** .170*** (.025) (.029) (.030) National-level diffusion squared -.005*** -.003*** -.003*** (M_ l)2 (.000) (.001) (.001) Dummy indicating years 2.48*** 2.48*** 1967-1968 (.260) (.260) Ln of non-White population .166 (.199) Prior rioting (control) .202*** .132*** -.0470 -.047 (.033) (.036) (.042) (.042)
Model chi-square (d.f.) 412.49 525.98 629.86 630.5 (d.f. = 7) (d.f. = 9) (d.f. = 10) (d.f. = 11)
Likelihood Ratio Tests Chi-Square d.f.
Model 1 vs. Model 2 113.49 2 Model 2 vs. Model 3 103.88 1 Model 3 vs. Model 4 .69 1
Note: Numbers in parentheses under coefficients are standard errors. *p < .05 p < .01 *** < .001 (two-tailed tests) ments. If demographic aggregation of na- 3. The results for variables, other than non- tional discontent is in fact operating, effects White population variable itself, are consis- of the non-White population size should ex- tent with the previous three models. And the ist above and beyond any community char- addition of the logged size of the non-White acteristics that are not directly associated population, however, has no effect on the with non-White population size. In other hazard of rioting.7 Therefore, although the words, if two cities have equal Black unem- variable may still have effects collinear to the ployment, equal manufacturingwages, equi- competition and diffusion variables (as indi- valent regional histories, and so forth, but cated by the attenuation of the coefficients one city has a larger Black population than the other, this city should have a higher risk I This finding is not an artifact of the particular of rioting than the city with fewer Blacks. model. When added to any model beginning with Model 4 in Table 4 adds the natural loga- Model 1 of Table 3, the logged size of the non- rithm of the non-White population to Model White population is not significant.
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions 108 AMERICAN SOCIOLOGICAL REVIEW for non-White unemployment and general and regional diffusion processes. Although unemployment), the demographic aggrega- the definitions of diffusion I used here are tion argument is not supported here because robust, many questions remain regarding the no additional effect is shown. exact nature of the diffusion process. First, given that the process appears to trail off quickly over a two-week period, a more de- SUMMARY AND DISCUSSION tailed approach that examines riot data on a Contraryto the conclusions drawn by Spiler- daily ratherthan on a weekly level may pro- man (1970a, 1971, 1976) in his classic stud- vide a fuller picture of this dynamic process. ies, I demonstrate that local. conditions did Models incorporatingvarious decaying func- indeed contribute to the occurrence of racial tions of time may also be useful. Second, al- rioting in the 1960s. The propensity to riot though the measure for proximity I used here was a function of far more than simply the has proven effective in other studies (Hed- number of Blacks available for rioting in a strom 1994), future work should test mea- particularcity. I find strong supportfor argu- sures that recognize boundaries that may fil- ments drawn from ethnic competition and ter or limit diffusion processes. Third, het- conflict literatureand for the existence of riot erogeneity within the diffusion phenomenon diffusion processes. Economic and migration suggests interesting possibilities for model- pressures in U.S. cities created different lev- ing diffusion. What differences among riots els of intergroup competition, which in turn can cause them to contribute more or less to contributed to the chances of experiencing the diffusion process? Size, severity, amount racial riots. Once a riot broke out, that event and type of media attention, and repressive increased the likelihood that other riots would responses are all possible riot characteristics break out elsewhere in the near future-par- that could affect the diffusion process. Fi- ticularly in cities geographically close to the nally, advances in understandingriot and col- original riot. These results are robust across a lective action diffusion processes will also variety of models, including models contain- depend on learning more about the different ing very conservative control variables. communication processes and networks that The structural strain arguments originally transmit the information driving the process examined by Spilerman (1970a) again failed itself. It may be that information communi- to find support. The arguments do not fail, cated throughacquaintance networks has dif- however, because of the lack of significant ferent effects at different rates than does in- relationships between individual variables formation gained through the mass media. and rioting. The variables associated with Furthermore,the dynamic interaction of all structural strain arguments actually do have these possible factors suggests fascinating importantrelationships with collective racial complexities and hypotheses for future mod- violence. Although these relationships are els of collective violence diffusion. often obscured in multivariate analysis due to collinearity among the variables, this is not the important finding-instead, it is the Competition lack of a coherent pattern among these vari- The hypotheses derived from competition ables relative to any of the tested theories theory found support here, although the in- that stands out. Thus I do not conclude that teractions suggested by the theory were not community characteristics have no relation- important. A city9s median manufacturing ships with rioting, but rather that the rela- wage does not appearat first to support com- tionships which do exist do not form a co- petition notions. Because manufacturing herent patternto support the specific theories wage is taken as an indicator of a healthy city Spilerman tested. economy, it was predicted to reduce inter- group competition and thus to decrease col- lective violence. The apparently contradic- Diffusion tory finding can be explained easily if it is My conclusions regarding the diffusion of understood how economic conditions may racial rioting are quite straightforward, as effect Blacks and Whites differently. While strong evidence emerged for both national high manufacturingwages may indicate that
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions RACIAL RIOTING IN THE 1960S 109 the economy is generally strong, labor mar- opportunity (via non-White population size) ket segregation causes a strong manufactur- as the sole source of rioting variability is not ing sector to benefit Whites more than supported because increases in population Blacks. As Olzak (1992) and Thernstrom size do not increase riot propensity beyond (1973) point out, Whites (both foreign-born that of other explanatory variables. Clearly, and native-born) dominate skilled occupa- the argument that 1960s rioting was an es- tions and expanding industries. Thus, when sentially random process once the demo- manufacturing wages are high, Whites dis- graphic aggregation of potential rioters was proportionately reap the benefits. Further- controlled is not adequate. more, Olzak's analysis supports the idea that A detailed stepwise examination of Model during periods of expansion, foreign-born 4 in Table 4 reveals that one of the main vari- Whites are able to move up and out of segre- ables superseding the size of the non-White gated occupations, which has the effect of population variable is the number of non- increasing the concentration of Blacks in Whites unemployed in a given city. In this low-paid and low-prestige jobs. Thus, one analysis, regardless of the percent unem- can see how general economic expansion, ployed, it is the number unemployed that has which might be expected to lower competi- the most powerful effect on riot rates. Three tion, could actually increase the likelihood of tempting interpretations come to mind. The collective violence on the part of Blacks. first is that the non-White unemployment Second, while greater overall unemploy- rate is simply a proxy for the non-White ment would seem to indicate greater compe- population size. While this interpretation is tition for jobs, my analysis suggests that partially true and clearly plausible in the case Black unemployment and White unemploy- of cities with similar unemployment rates but ment have very different effects on rioting, different numbers of Blacks (those with particularly in segregated labor markets. higher populations will experience higher While White unemployment may cause more rates of rioting), the interpretation breaks White aggression against Blacks, and Black down when Black unemployment rates vary. unemployment can be expected to cause Under such circumstances, cities with larger more Black aggression against Whites, it is Black populations and lower unemployment much less reasonable to expect that White rates could have lower raw numbers of un- unemployment would cause Black aggres- employed Blacks than some cities with smal- sion. Thus, overall unemployment must be ler Black populations and higher Black un- decomposed by race to make sense of its ef- employment rates. Given the results pre- fects on the 1960s rioting. If unemployment sented here, the cities with larger Black is viewed as an outcome of losing to corn- populations would experience less rioting petitors in the market place, then Black un- than cities with smaller Black populations, a employment, not White unemployment, result which is exactly the opposite of what would be expected to increase Black collec- would be predicted from the proxy argument. tive violence. The position I take, however, is not that the size of the Black population is unimportant to riot rates. In fact, the models presented in Non-White Population Size and Table 4 demonstrate that the number of non- Unemployment Whites unemployed and the size of the non- In this study, the combination of competition White population are collinear with regard to and diffusion provides a model that most co- their ability to predict rioting. Nevertheless, herently accounts for differences in riot risks the parameterestimate for the number of un- among U.S. cities in the 1960s. This model employed non-Whites is significant, indicat- is superior to models using structural-strain ing that there are important differences in variables and to a model dependent only on riot rates that cannot be explained simply by non-White population size. The competition/ the overlap between these two variables. diffusion model is superior to the non-White A second interpretationof the relationship population size argument because it illumi- between the numbers of non-White unem- nates theoretical explanations for rioting. ployed and riot rates is that rioting requires The notion of demographic aggregation or biographical availability, and the unem-
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions 110 AMERICAN SOCIOLOGICAL REVIEW ployed, given the lack of demands on their ticipation by the unemployed specifically. At time, are the most available segment of the present, available empirical data are not ad- population. While biographical availability equate to assess these two possibilities. This has proven to be an extremely useful variable study, however, strongly suggests that the in social movement and collective behavior role of unemployment is one key to under- research (particularly with regard to student standing riot outbreaks, and that further re- protest movements), it is less useful in this search should focus on determining exactly context. Many studies conducted on riot par- how unemployment contributes to civil un- ticipants have found no significant differ- rest. ences in employment status between rioters and non-rioters. For example, Opp (1989) Daniel Myers is a Ph.D. candidate in Sociology found that unemployment had no relationship at the University of Wisconsin-Madison. His re- to illegal protest and a negative relationship search interests include collective violence, for- inal models collective to legal protest. Similar relationships have of action, game theory, and media coverage of protest activity. His disserta- been found in a numberof surveys of partici- tion research focuses on developing and testing pants and nonparticipantsof the 1960s riots models of the diffusion of collective violence. (Rodgers and Bullock 1974; Caplan 1970; Mason and Murtagh 1985). Furthermore, REFERENCES many studies have established that urban ri- ots occur "at times when (e.g., after working Alford, Robert and Michael Aiken. 1970. Gov- hours or on weekends) people are available" ernmental Units Analysis, 1960 [MRDF]. Uni- versity of Wisconsin, Data and Program Li- (McPhail 1994:9; also see McPhail and Mil- brary Service, Madison, WI. ler 1973; Burbeck, Raine, and Stark 1978). Allison, Paul D. 1984. Event History Anahlysis: Because the unemployed are not constrained Regression for Longitudinal Event Data. New- by working hours, their availability for riot- bury Park, CA: Sage Publications. ing is continual. Therefore, if the unemployed Berkowitz, Leonard. 1968. "The Study of Urban were largely responsible for rioting, this tem- Violence: Some Implications of Laboratory poral pattern of rioting would not exist. Studies of Frustrationand Aggression." Ameri- A third explanation of the importance of can Behavioral Scientist 2:14-17. non-White unemployment to rioting reasons Bonacich, Edna. 1972. "A Theory of Ethnic An- tagonism: The Split Labor are the most Market."American that the unemployed discontent Sociological Review 37:547-59. and therefore would be the more likely than Brown, Lawrence A. 1981. Innovation Diffusion: the employed to lash out in violence. Again A New Perspective. London, England: Meth- evidence that the unemployed no more con- uen. stitute the body of rioters than they do the Bryan, Marguerite. 1979. "The Social Psychology general population contradicts this explana- of Riot Participation." Research in Race and tion. Furthermore, few studies have estab- Ethnic Relations 1:169-89. lished any attitudinal differences between Burbeck, Stephen L., Walter J. Raine, and M. J. those who participate in riots and those who Abudu Stark. 1978. "The Dynamics of Riot Growth: An Jour- Bullock Epidemiological Approach." do not (Rodgers and 1974). nal of Mathematical Sociology 6:1-22. In short, the reason that the numberof un- Caplan, Nathan. 1970. "The New Ghetto Man: A employed is so important to riot rates is not Review of Recent Empirical Studies." Journal presently clear. It may be that even though of Social Issues 26:59-73. the unemployed are not overrepresented Carter, Gregg Lee. 1986. "The 1960s Black Riots among riot participants, they may be over- Revisited: City Level Explanations of Their Se- represented among those who initiate rioting. verity." Sociological Inquiry 210-28. If this were true, biographical availability, 1990. "Collective Violence and the Prob- severe discontent, or both could be operat- lem of Group Size in Aggregate-Level Stud- ies." Sociological Focus 23:287-300. ing. the other it be that On hand, may higher .1992. "Hispanic Rioting During the Civil numbers of unemployed Blacks make Rights Era." Sociological Forum 7:301-22. Blacks' economic problems highly visible Cox, D. R. 1972. "Regression Models and Life and thus highly salient. Such a condition Tables." Journal of the Royal Statistical Soci- would increase the likelihood of riots inde- ety Series B 34:187-202. pendent of biographical availability or par- Diekmann, Andreas. 1979. "A Dynamic Stochas-
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions RACIAL RIOTING IN THE 1960S 111
tic Version of the Pitcher-Hamblin-Miller Koopmans, Ruud. 1993. "The Dynamics of Pro- Model of 'Collective Violence."' Journal of test Waves: West Germany, 1965 to 1989." Mathematical Sociology 6:277-82. American Sociological Review 58:637-58. Dotson, A. Bruce. 1974. "Social Planning and Lichbach, Mark. 1985. "Protest: Random or Con- Urban Violence: An Extension of McElroy and tagious: The Postwar United Kingdom."Armed Singell." Urban Affairs Quarterly 9:283-301. Forces and Society 11:581-608. Downes, Bryan T. 1968. "Social and Political Lieberson, Stanley. 1980. A Piece of the Pie: Characteristics of Riot Cities: A Comparative Blacks and White Immigrants since 1880. Ber- Study." Social Science Quarterly 49:504-20. keley, CA: University of California Press. Flinn, Christopher J. and James J. Heckman. Lieberson, Stanley and Arnold R. Silverman. 1982a. "Models for the Analysis of Labor 1965. "The Precipitants and Underlying Con- Force Dynamics." Advances in Econometrics ditions of Race Riots." American Sociological 1:35-95. Review 30:887-98. . 1982b. "New Methods for Analyzing In- Mahajan, Vijay and Robert A. Peterson. 1985. dividual Event Histories." Sociological Meth- Models for Innovation Diffusion. Beverly Hills, odology 11:99-140. CA: Sage Publications. Gurr,Ted R. 1968. "UrbanDisorder: Perspectives Mason, David T. and Jerry A. Murtagh. 1985. from the Comparative Study of Civil Strife." "Who Riots? An Empirical Examination of the American Behavioral Scientist 2:50-55. 'New Urban Black' versus the Social Margin- . 1970. Why Men Rebel. Princeton, NJ: ality Hypotheses." Political Behavior 7:352- Princeton University Press. 73. Hamblin, Robert L., R. Brooke Jacobsen, and McAdam, Doug. 1982. Political Process and the Jerry L. L. Miller. 1973. A Mathematical Development of Black Insurgency, 1930-1970. Theory of Social Change. New York: John Chicago, IL: University of Chicago Press. Wiley and Sons. McElroy, Jerome L. and Larry D. Singell. 1973. Heckman, James J. and Burton Singer. 1982. "Riot and Nonriot Cities: An Examination of "The Identification Problem in Economic Mod- StructuralContours." Urban Affairs Quarterly els for Duration Data." Pp. 35-95 in Advances 8:28 1-302. in Econometrics, edited by J. Hildebrand.Cam- McPhail, Clark. 1991. The Myth of the Madding bridge, England: Cambridge University Press. Crowd. New York: DeGruyter. . 1984. "A Method for Minimizing the Im- . 1994. "Presidential Address-The Dark pact of Distributional Assumptions in Econo- Side of Purpose: Individual and Collective Vio- metric Models for Duration Data." Econo- lence in Riots." The Sociological Quarterly metrica 52:271-320. 35:1-32. Hedstr6m, Peter. 1994. "Contagious Collectivi- McPhail, Clark and David L. Miller. 1973. "The ties: On the Spatial Diffusion of Swedish Trade Assembling Process: A Theoretical and Em- Unions, 1890-1940." American Journal of So- pirical Examination." American Sociological ciology 99:1157-79. Review 38:721-35. Heskin, Ken. 1985. "Political Violence in North- McPhail, Clark and Ronald T. Wohistein. 1983. ern Ireland." Journal of Psychology, 119:481- "Individual and Collective Behavior within 94. Gatherings, Demonstrations and Riots." An- Jenkins, J. Craig and Craig M. Eckert. 1986. nual Review of Sociology 9:579-600. "ChannelingBlack Insurgency:Elite Patronage Morris, Aldon D. 1984. The Origins of Civil and Professional Social Movement Organiza- Rights Movement: Black Communities Orga- tions in the Development of the Black Move- nizing for Change. New York: Free press. ment." American Sociological Review 51:812- National Advisory Commission on Civil Disor- 29. ders. 1968. Report of the National Advisory Jiobu, Robert M. 1971. "City Characteristics and Commission on Civil Disorders. New York: Racial Violence." Social Science Quarterly Bantam Books. 55:52-64. Nielsen, Francois. 1985. "Toward a Theory of Kelly, William R. and Larry Isaac. 1984. "The Ethnic Solidarity in Modern Societies." Ameri- Rise and Fall of Urban Racial Violence in the can Sociological Review 50:133-49. U.S.: 1948-1979." Research in Social Move- Oberschall, Anthony. 1978. "The Decline of the ments Conflict and Change 7:203-33. 1960s Social Movements." Research In Social Koomen, Willem and Erik G. Frankel. 1992. "Ef- Movements, Conflict and Change 1:257-89. fects of Experienced Discrimination and Dif- Oliver, Pamela E. 1989. "Bringing the Crowd ferent Forms of Relative Deprivation among Back In: The Nonorganizational Elements of Surinamese, a Dutch Ethnic Minority Group." Social Movements." Research in Social Move- Journal of Communityand Applied Social Psy- ments, Conflict and Change 11:1-30. chology 2:63-7 1. Olzak, Susan. 1987. "Causes of Ethnic Protest
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions 112 AMERICAN SOCIOLOGICAL REVIEW
and Conflict in Urban America, 1877-1889." Spilerman, Seymour. 1970a. "The Causes of Ra- Social Science Research 16:185-210. cial Disturbances: A Comparison of Alterna- . 1989a. "Analysis of Events in the Study tive Explanations."American Sociological Re- of Collective Action." Annual Review of Soci- view 35:627-49 ology 15:119-41. . 1970b. Governmental Units Analysis . 1989b. "Labor Unrest, Immigration, and Data, 1960: Urban Racial Disorders, 1961- Ethnic Conflict in UrbanAmerica, 1880-1914." 1968 [MRDF]. University of Wisconsin, Data American Journal of Sociology 94: 1303-33. and Program Library Service, Madison, WI. . 1990. "The Political Context of Competi- . 1971. "The Causes of Racial Distur- tion: Lynching and Urban Racial Violence, bances: Test of an Explanation."American So- 1882-1914." Social Forces 69:395-421. ciological Review 36:427-42. . 1992. The Dynamics of Ethnic Competi- . 1972. "Strategic Considerations in Ana- tion and Conflict. Stanford, CA: Stanford Uni- lyzing the Distribution of Racial Disturbances." versity Press. American Sociological Review 37:493-99. Olzak, Susan and Joane Nagel. 1986. Competitive . 1976. "StructuralCharacteristics of Cit- Ethnic Relations. Orlando, FL: Academic. ies and the Severity of Racial Disorders." Olzak, Susan and Suzanne Shanahan. 1996. "Dep- American Sociological Review 41:771-93. rivation Race Riots: An Extension of Spiler- Strang, David. 1991. "Adding Social Structureto man's Analysis." Social Forces 74:931-61. Diffusion Models: An Event History Frame- Olzak, Susan and Elizabeth West. 1991. "Ethnic work." Sociological Methods and Research Conflict and the Rise and Fall of Ethnic News- 19:324-53. papers." American Sociological Review 56: Strang, David and Nancy Brandon Tuma. 1993. 458-74. "Spatial and Temporal Heterogeneity in Diffu- Opp, Karl-Dieter. 1989. The Rationality of Politi- sion." American Journal of Sociology 99:614- cal Protest: A Comparative Analysis of Ratio- 39. nal Choice Theory. Boulder, CO: Westview. Thernstrom,S. 1973. The Other Bostonians: Pov- Paige, Jeffrey M. 1971. "Political Orientation and erty and Progress in the American Metropolis, Riot Participation."American Sociological Re- 1880-1970. Cambridge, MA: Harvard Univer- view 36:810-20. sity Press. Park, Robert. 1950. Race and Culture. Glencoe, Tilly, Charles, Louise Tilly, and Richard Tilly. IL: Free Press. 1975. The Rebellious Century: 1830-1975. Pitcher, Brian L., Robert L. Hamblin, and Jerry Cambridge, MA: HarvardUniversity Press. L. L. Miller. 1978. "Diffusion of Collective Tolnay, Stewart E. and E. M. Beck. 1995. A Fes- Violence." American Sociological Review 43: tival of Violence: An Analysis of Southern 23-35. Lynchings, 1882-1930. Urbana, IL: University Polletta, Francesca. 1992. "Politicizing Child- of Illinois Press. hood: The 1980 Zurich Burns Movement." So- Tomlinson, T. M. 1968. "The Development of a cial Text 33:82-102. Riot Ideology among Urban Negros." Ameri- Rodgers, Harrell R. and Charles S. Bullock III. can Behavioral Scientist 11:27-3 1. 1974. "Political and Racial Attitudes: Black Trussell, James and Toni Richards. 1985. "Cor- versus White." Journal of Black Studies 4:463- recting for Unmeasured Heterogeneity in Haz- 85. ard models Using the Heckman-Singer Proce- Rogers, Everett M. 1983. Diffusion of Innova- dure." Sociological Methodology 14:242-76. tions. New York: Free Press. Tuma, Nancy Brandon. 1985. "Effects of Labor Rud6, George. 1964. The Crowd in History: A Market Structure on Job-Shift Patterns." Pp. Study of Popular Disturbances in France and 327-363 in Longitudinal Analysis of Labor England, 1730-1848. New York: Wiley. Market Data, edited by J. J. Heckman and B. .1972. Europe in the Eighteenth Century: Singer. Cambridge, England: Cambridge Uni- Aristocracy and the Bourgeois Challenge. New versity Press. York: Praeger. Tuma,. Nancy Brandon and Michael T. Hannan. Schneider, Hans Joachim. 1992. "Criminology of 1984. Social Dynamics: Models and Methods. Riots." International Journal of Offender Ther- San Diego, CA: Academic. apy and Comparative Criminology 36:173-86. Yamaguchi, Kazuo. 1986. "Alternative App- Shorter, Edward and Charles Tilly. 1974. Strikes roaches to Unobserved Heterogeneity in the in France: 1830-1968. New York: Cambridge Analysis of Repeatable Events." Sociological University Press. Methodology 16:213-49. Smelser, Neil. 1962. Theory of Collective Behav- . 1991. Event History Analysis. Newbury ior. New York: Free Press. Park, CA: Sage.
This content downloaded on Mon, 21 Jan 2013 18:26:05 PM All use subject to JSTOR Terms and Conditions