How Did Things Get So Bad So Quickly? An Assessment of the Initial Conditions of the War Against Organized Crime in Mexico

Abstract

Objectives: This study explores the initial conditions of the current war against organized crime in Mexico. The theoretical framework is institutional anomie theory (IAT). Composite measures were used to summarize local initial conditions for the occurrence of organized crime deaths by gang execution, confrontation, and aggressions to authority. Spatial and temporal elements were included to assess the validity of the initial conditions approach. Evidence presented here suggests that Mexican states significantly differed in their initial conditions for organized crime deaths to have occurred. Also, although trends in gang executions and confrontations have been slowing down, aggressions to authority are speeding up considerably. The evidence presented corroborates IAT. However, the significance and direction of the relationships among institutional anomie correlates and initial conditions of the war against organized crime depended upon the type of death.

Appendix

The Global Moran I autocorrelation coefficient is given by the following formula (Holt, 2007):

$$ I=\left(\frac{1}{s^2}\right)\frac{{{\displaystyle \sum}}_{i=1}^n{{\displaystyle \sum}}_{j=1}^n{w}_{ij}\left({x}_i-\overline{x}\right)\left({x}_j-\overline{x}\right)}{{{\displaystyle \sum}}_{i=1}^n{{\displaystyle \sum}}_{j=1}^n{w}_{ij}}, $$

where N is the number of states (N = 32), x_i and x_j are the values of the dependent variables, deaths by execution, confrontation and aggression towards authority (Ln) in the states i and j. s² is the sample variance. And W_ij is the neighbouring matrix. The element x_[i,j] of the resulting weight matrix X is 1 if polygon j is adjacent to polygon i, and is 0 otherwise.

The local Moran I autocorrelation coefficient is given by the following formula (Holt, 2007):

$$ {I}_i=\left({x}_i-\overline{x}\right){\displaystyle \sum}_{j=1}^n{w}_{ij}\left({x}_j-\overline{x}\right). $$

Same notation as above.

The baseline growth-curve model applied was the following (Kubrin and Herting, 2003):

$$ (ln){Y}_{it}={a}_i+{b}_{1i}(Time)+{b}_{2i}{(Time)}^2+{\mathrm{e}}_{it}, $$

where (Ln)Y _it is the logged count of the number of organized crime deaths in state i at time t, ai is the estimated mean constant, b₁ is the mean effect of time on deaths, b₂ is the mean effect of time squared, and e_it is each observation error. The constant reflects the initial levels of organized crime deaths at the beginning of the time series, b₁ reflects the main linear trend in deaths levels, and b₂ reflects the extent to which the trend accelerates or decelerates over time.

I applied the following test (Paternoster et al. 1998; Kubrin and Weitzer; 2003):²⁰

$$ Z=\frac{a_i-{a}_j}{\sqrt{ SE{a}_i^2+ SE{a}_j^2}}, $$

where a _i and a _j the intercepts and SE the standard errors.

The weighted mean centres (WMC) are calculated in the following form (Vilalta, 2011):

$$ WMC=\left({\overline{x}}_{cmp},{\overline{y}}_{cmp}\right)=\left(\frac{{{\displaystyle \sum}}_{i=1}^n{w}_i{x}_i}{{{\displaystyle \sum}}_{i=1}^n{w}_i},\frac{{{\displaystyle \sum}}_{i=1}^n{w}_i{y}_i}{{{\displaystyle \sum}}_{i=1}^n{w}_i}\right), $$

where x_cm y y_cm are the coordinates of the centre mean, x_i y y_i are each centroid coordinates for polygon i,n is the sample size, and W_i is each polygon´s weighted variable.

The spatial spatial autoregressive (SAR) model takes the form (Lesage, 1998):

$$ Y=\uprho \mathrm{Wy}+\mathrm{x}\upbeta +\mathrm{e}, $$

where y contains an nx1 vector of dependent variables, X represents the usual nxk data matrix containing explanatory variables and W is a known spatial weight matrix, in this case a first-order or strict contiguity matrix. Rho (p) is the coefficient on the spatially lagged dependent variable (Wy-). Maximum likelihood estimation of this model is based on a concentrated likelihood function.