The study of the causes of violence and crime has been a part of western sociology since the beginning. A variety of causes have been suggested: the incidence of absolute poverty, the extent of inequalities, the phenomenon of relative deprivation, the drug trade, the breakdown of traditional community and family values, and the effects of racism, to name a few.
Notice that we can put the causal question in several ways:
- What causes variation across communities with respect to crime rates?
- What factors increase the likelihood of a particular individual becoming a violent criminal?
- What social factors cause an increase or reduction in the crime rate?
That is, we can ask about explaining variation across cases; we can ask about explaining particular individuals' behavior; we can ask about "inducing" and "inhibiting" causes of changes in the crime rate; and there are other causal questions as well. The final pair of questions have to do with explaining variation within a case across time.
Consider this small set of possible causal factors that might influence behavior:
- rational incentives (risk and gain calculation)
- material circumstances (unemployment, education)
- community cohesiveness
- a broad sense of injustice (exploitation, unfair exclusion)
- moral and religious values
- alienation and disaffection
- racial or ethnic polarization
- imitation of the behavior of others
- organizations (gangs, youth groups, social networks)
- laws and policing
Now we might try to construct an agent-centered theory of violent crime based on these sorts of factors conjoined with a description of the social environment at the time -- and then predict variation across time and place as a behavioral result of the incorporation of these factors influencing action.
A very different approach -- and one that is probably closer to the quantitative-methodology mainstream in sociology today -- is to assemble a set of cases (a list of US cities, for example); measure a number of variables for each city (unemployment rate, index of social capital, level of education, degree of neighborhood segregation, presence/absence of mass transportation, ...); and then test the degree of correlation that one or more of these variables has with the observed variation in the crime rate. Do variations in the incidence of violent crime correlate with levels of unemployment? Then unemployment is a plausible causal factor in determining the level of the crime rate. Does some measure of social capital correlate negatively with variations in the crime rate? Then the social cohesion that is hypothetically linked to higher scores for social capital in a community is a negative causal factor in variations in the crime rate. And so forth.
Both these approaches are compatible with the research methodology associated with trying to identify causal mechanisms. If it is in fact true that a young person's disposition to engaging in violent crime is decreased if he/she is a member of a church -- then we ought to find at the macro-level that a higher index of church membership will be associated with a lower crime rate. Reciprocally, if we discover a positive correlation between "degree of neighborhood segregation" and the crime rate -- then we need to be able to disaggregate the story and discover the individual-level mechanisms through which segregation increases individuals' propensity for violent crime.