General Strain Theory and Racial Insurgency: Assessing the Role of Legitimate Coping

Abstract

While General Strain Theory (GST) recognizes the broad range of legitimate and illegitimate coping behaviors people adopt in reaction to strain, tests of the theory focus almost exclusively on criminal coping. We advance the theory by articulating the role of legitimate coping in the GST process. We test the theory’s assumptions that strain increases both legitimate and illegitimate coping and that negative emotions more strongly increase illegitimate coping. We also draw on recent work by Agnew and expect that these coping strategies co-occur rather than being mutually exclusive. We investigate these hypotheses in relation to black insurgency, specifically nonviolent civil-rights protest (legitimate coping) and rioting (illegitimate coping). Using data from a large 1968 survey of blacks, multivariate findings are consistent with the theory’s expectations regarding the role of legitimate coping. More broadly, our results echo calls to extend the boundaries of the discipline beyond traditional conceptualizations of “crime.”

Keywords:

• GST
• legitimate coping
• rioting
• nonviolent protest

Notes

¹ Though scholars disagree about the utility of the Hausman test, we present these tests as an additional check for model bias given the simultaneity concerns associated with cross-sectional data. To test whether nonviolent protest and rioting simultaneously affect the endogenous predictor variable (frustration), we identified an instrumental variable (dissatisfaction with one’s current financial situation; 1 = pretty well satisfied, 2 = more or less satisfied, 3 = not at all satisfied) that is significantly correlated with the endogenous variable (r = .13). This indicates that blacks who are unhappy with their financial situation are frustrated with the non-existent pace of racial progress. The instrument variable is not correlated with the residuals from regression models predicting nonviolent protest (Table 2, model 3; r = .03) or riot participation measures (Table 2, model 5; r = .01; Table 2, model 7; r = .05). We then ran a separate model regressing frustration with the instrument variable plus controls. As expected, the instrument variable is significantly correlated with frustration. Following the recommendations of Terza, Basu, and Rathouz (2008) for non-linear models like ours, we then include the residualized variable (rather than the predicted values) from that model along with the measure of frustration as an additional independent variable in the estimation of violent and nonviolent protest in models 3, 5, and 7 (Table 2). Including residuals in this model purges endogeneity from the coefficient of frustration and provides a Hausman-style test of the significance of the endogeneity. These residualized variables are not significant in any of the three models, suggesting that the association of frustration with nonviolent and violent protest works primarily in one direction. Results available on request.

² A Hausman test for simultaneity would allow us to test and control for simultaneity between nonviolent protest and rioting. However, the data-set includes only a limited range of survey items and we could not identify a suitable and theoretically relevant instrument variable that was both correlated with nonviolent protest (our endogenous variable) and not correlated with residuals form the regression models predicting riot outcomes. As we note throughout the paper, we suspect nonviolent protest generally precedes rioting, but this may not always be the case.  These data allow us to speak convincingly to the relationship between the two but only to speculate on their time order.

³ 5.8% of the sample had missing income data. Our rare events models cannot accommodate the chained equation imputation approach, so for these models we used conditional mean imputation by replacing missing values with predicted values from a regression analysis of the complete data. Notably the chained regression and standard regression imputation strategies produce income estimates that are highly correlated and neither is significant in our models predicting coping behavior. To check if respondents with missing data differed from those with non-missing data, we created a variable coded 1 if the respondent had missing data on income, 0 otherwise, but in our regression analyses this term was never statistically significant nor did its inclusion meaningfully alter other results.

⁴ In Stata 13.0, the rare-events logistic procedure does not support the city-level clustering procedure. However, those models still include dummy variables to control for city effects and also use robust standard errors.

⁵ We use the formula recommended by Paternoster et al. (1998) to compare regression coefficients across models: We compare models using the broader riot measure with the nonviolent protest outcome models since these rely on the same set of imputed data and are thus fully comparable. The coefficients for the strain/violence link are even closer (though we cannot formally compare them since the models predicting riot participation use a rare events regression model that cannot accommodate imputed data) when we compare the nonviolent protest model (that includes imputed data) with that for riot participants only (model 6).

⁶ We see our contributions mainly to the field of criminology but our study also has implications for other fields. Political scientists have long been interested in the link between deprivation and civil wars, revolutions, and the like. These studies almost always investigate processes occurring at the macro-level and find mixed results. Our work shows strong links between strain and rioting, suggesting that micro-level processes offer fertile ground for political science theorizing. In the field of social movements, strains at one time were seen as the prime motivator of collective behavior but since the 1970s such “breakdown” approaches have fallen out of favor. Strain theories fell out of favor in criminology around this time as well. GST addresses questions that beleaguered strain theory in both criminology and political science in the 1970s—why does strain only sometimes lead to non-normative outcomes (crime and rioting)? From the perspective of GST, the answer lies in the critical role of emotional processing and related coping resources—mechanisms that original strain and breakdown theories did not articulate but that our study suggests are of central import. Attention to these individual-level processes are likely relevant to political science as well and can augment the focus on opportunity structures by detailing how opportunity structures influence the range of coping strategies individuals might be able to access and outlining the ways in which structural contexts might influence the potential utility and success of these strategies.

Additional information

Notes on contributors

Lisa Broidy

Lisa Broidy is a professor of Sociology at the University of New Mexico and an Adjunct professor at the Griffith Criminology Institute in Brisbane, Australia. Her scholarship centers on the etiology of crime with a particular focus on the influence of gender, life course transitions, institutional contact, and strains (including victimization and trauma).

Wayne A. Santoro

Wayne A. Santoro is an associate professor of Sociology at the University of New Mexico. His work lies at the intersection of race, politics, and social movements. He examines how African Americans, Mexican Americans, and Arab Americans have mobilized to compel governments to become responsive to community concerns. Other recent work investigates how city political contexts affect crime in immigrant and black neighborhoods, the political dynamics that took place during the decline of the civil rights movement, and the links between generation and conventional political participation among Puerto Ricans.