Gary F. Jensen and Ronald L. Akers
A major development in criminology over the last ten to fifteen years has been the revival of research testing theories relevant to variation in crime among ecological units, ranging from neighborhoods to societies (e.g., Sampson and Groves 1989; Gartner 1990; Patterson 1991; Bursik and Grasmick 1993; Bellair 1997; Markowitz et al. 2001). This resurgence of macro-level theory and ecological analysis contrasts with the dominant method for testing criminological theories during much of the 1970s and 1980s, the analysis of correlates of self-reported offense behavior based on questionnaire or interview data collected from individuals. That type of research allowed an assessment of the comparative importance of variables central to several distinct theories about motivational and constraining forces affecting the odds of delinquent behavior and, overall, provided substantial support for theories emphasizing the causal primacy of peer and family relationships, school experiences, personal commitments and conventional values. Theories emphasizing economic disadvantage and frustrations in the pursuit of conventional goals (so-called “strain” theories) did not fare as well as theories focusing on social constraints or social learning processes (see Akers 2000; Jensen and Rojek 1998). Such conclusions have generated considerable controversy and one of the lines of argument behind a resurgence of macro-level research flows from a critique of the self-report survey and analyses of survey data collected from individuals.
Conflicting Themes in the Ecological Revival
In 1987, Thomas Bernard challenged the adequacy of any analyses based on survey data from individuals for testing strain theory, arguing that a proper
assessment of Merton’s (1938, 1957) version of the theory required analysis of variations at a macro level (referred to as anomie theory at the macro level). Steven Messner and Richard Rosenfeld have endorsed Bernard’s position and reiterate the argument that tests at the individual level have no necessary bearing on the adequacy of the theory at the macro level. Indeed, they present examples where variations at the individual level within nations do not apply across nations (2001: 38-46) and propose an “institutional anomie theory” that is supposed to explain variations in “serious” crime among nations.1
The fact that tests of competing theories using data collected from individuals have provided strong support for social learning theory as opposed to strain theory is considered irrelevant to a global analysis.
Messner and Rosenfeld are very critical of the self-report survey data as measures of “serious” crime and theories supported by such data are depicted as relevant only to minor offenses.2
In their opinion, the dominance of the self-report method and analyses of data at the individual level constitute an “individualistic bias in contemporary criminology” (Rosenfeld and Messner 1995: 161-164; Messner and Rosenfeld 2001: 38-42). In contrast, their “institutional anomie” theory is supposed to overcome these limitations and explain macro-level variations in serious crimes.
They qualify their critique of individual explanations by stating that “Any macro-level explanation of crime will inevitably be predicated on underlying premises about individual behavior” (2001: 42), and “it is possible ... to link these individual-level theories with macro-level explanations that share a causal logic.” However, these statements are countered by (1) the dismissal of the very type of research that has challenged the validity of strain theory at the individual level while supporting social learning theory, (2) the use of examples of discontinuities at different levels of analysis with no discussion of continuities, and (3) the rejection of the analysis of survey data as an appropriate technique for testing macro-level theories or studying truly “serious” crime.
This critique can be contrasted with the approach taken by other theorists and researchers involved in the resurgence of ecological analysis. Much of the central research in a revival of Shaw and McKay’s “social disorganization” theory in recent years measures crime using survey data, including data on self-reported offending collected from individuals as well as victimization data (e.g., Sampson and Groves 1989; Sampson, Raudenbusch and Earls 1997; Markowitz, Bellair, Liska and Liu 2001). The survey data is aggregated to create measures appropriate for ecological analysis and used to measure rates of serious crime. Cautions about generalizing from one level of analysis to another are expressed, but survey data are accepted as an appropriate foundation for macro-level analysis.
Moreover, theoretical extensions and elaborations of social disorganization theory convey an impression of continuity between levels of analysis. It
is common to acknowledge links between the logic of Hirschi’s social control theory (1969) and social disorganization theory. Weak bonds between youth and conventional adults, institutions and values increases the probability of delinquency among individuals in Hirschi’s control theory and weak communal institutions or networks increase crime rates at the ecological level. While there has not been a thorough elaboration of the possible transitions and problems in making those transitions, the discussion of links suggests continuity as opposed to discontinuity. In fact, while stressing discontinuities between macro-micro levels in research, Messner and Rosenfeld categorize social disorganization with social control theory, cultural deviance with social learning theory, and anomie with strain theories. As will be discussed below, while there are links between cultural deviance and social learning theory, this coupling is problematic for some depictions of cultural deviance theory.
The Neglect of Social Learning Theory
The one theory that has received very little attention in this ecological revival is social learning theory. It is included together with cultural deviance theory in Messner and Rosenfeld’s classification of theories, but their subsequent discussion focuses entirely on cultural deviance theory. None of the characteristics of contemporary social learning theory are discussed, and the major social learning theorists are relegated to a footnote. Most of their discussion is devoted to Ruth Kornhauser’s (1978) controversial interpretation of cultural deviance theory. Moreover, while the links between social disorganization theory and social control theory are mentioned by most of the parties to the ecological revival, virtually nothing is said about links between social learning theory and the ecological perspectives proposed. This neglect is quite surprising when the fact that social learning theory has fared better than any other theory in the explanation of variation in criminal and delinquent behavior using self-report data collected from individuals.
Several possible reasons for this neglect can be proposed. For one, its development was overshadowed by Hirschi’s Causes of Delinquency
(1969). While Hirschi’s theory is classified together with social learning theory as an “individualistic” theory of crime, one of the reasons Causes of Delinquency
became the most quoted work in criminology in the twentieth century was the fact that it proposed differences among social control, cultural deviance and strain theories in macro-level assumptions about American society and conducted analysis relevant to those assumptions and correlates distinctively important to each of those theories. For example, Hirschi reasoned that were America characterized by the class-linked cultural or normative variations emphasized by Walter Miller (1957), then attitudes towards the law and moral beliefs should vary by social class. Variation in such attitudes were found to
be correlated with delinquency, but, in contrast to Miller’s version of cultural deviance theory, they were not found to be significant correlates of social class. Moreover, if such sizeable subcultures linked to class or race existed, measures of interaction and attachment to adults in such categories should be positive correlates of delinquency for the young. In contrast, Hirschi found that such variables inhibited delinquency in all categories. In short, Hirschi addressed arguments about the subcultural organization of American society by exploring variations across groups and examining correlates within socially differentiated groups. He took the same approach when testing strain theory. If acceptance of widely shared, culturally approved goals facilitated delinquency among disadvantaged youth, then such aspirations should be positive correlates of delinquency among youth in categories where chances of realizing such goals were slight. He did not find that pattern when examining such categories of youth. In short, Hirschi’s work was attended to carefully by sociologists because it dealt with criminogenic properties of American social structure and culture.
When Hirschi’s work was published, a nascent “social learning theory” was under construction by Ronald Akers, elaborating on Burgess and Akers’ (1966) operant reformulation of Sutherland’s “differential association” theory. Initially presented as an operant restatement of Sutherland’s basic propositions, Akers’ specification of a more general social learning theory (1973) elaborated new ideas that distinguished the theory from its predecessors, including cultural deviance theory, and some interpretations of differential association theory. However, since social learning theory did not emerge as a repudiation of alternatives, issues that distinguished it from prior theories were not initially addressed and no specific macro-level counterpart was identified.
Another source of neglect stems from the tendency to link social learning theory to the most extreme versions of cultural deviance theory. For example, Messner and Rosenfeld state that “the idea that criminal behavior is learned (at the individual level) and that crime is entirely a product of culture (at the macro level) quickly runs into several inter-related problems.” However, the inter-related problems they address are all problems with a “pure” cultural explanation of crime. They criticize cultural theories for assuming that “all behavior is consistent with underlying values (2001: 47).” But, social learning theory makes no such assumption! Indeed, one of the empirically substantiated features of social learning theory is its delineation of several distinct learning mechanisms, only one of which involves cultural or normative phenomenon. Social learning theory proposes that delinquent behavior is learned and sustained through a variety of non-normative learning processes (direct and vicarious differential reinforcement, imitation and differential association) as well as through normative mechanisms. Normative socialization, or definitional learning processes, involve values and norms, but the separable
effects of non-normative processes means that criminal and delinquent activities can be engaged in despite “underlying values.”3
In short, social learning theory posits the very mechanisms that explain why behavior is not always consistent “with underlying values.”
Social learning theory is explicitly contrary to both the version of cultural deviance theory that Messner and Rosenfeld cite as its macro-level partner and to Matsueda’s initial interpretation of cultural deviance-differential association theory as well. In Matsueda’s first test of differential association theory, the effects of all other variables are viewed as channeled through “definitions” (see Matsueda 1982). Indeed, quite recently, Costello and Vowell (1999) reanalyzed Hirschi’s data and found Matsueda’s interpretation to be wrong. Contrary to Matsueda’s initial cultural interpretation of differential association theory, definitions did not fully mediate the effect of variables such as delinquent peer associations. Costello and Vowell interpret such findings as “more supportive of control theory than differential association theory” (1999: 815). Yet, despite the fact that the best fitting model included peer variables (an effect specifically excluded in Hirschi’s presentation of control theory), they totally ignore social learning theory. The best fitting model is a social learning model.
In this chapter, we advocate several positions relevant to micro-macro transitions in the development of social learning theory. However, before we can attempt to specify micro-macro links for social learning theory, we need to address some misconceptions about “individual level” versus “macrolevel” explanations. There is a tendency in critiques of certain theories as “individualistic” to confuse sources of data and levels of aggregation with levels of explanation. The delineation of “individual” versus “macro-level” explanations and the inclusion of some theories as “individualistic” by Messner and Rosenfeld misrepresent those theories. Not only do we need to address such misconceptions, but the links between the “causal logics” of theories at different levels need to be addressed. Development of any integrated macro-micro theory will require a more precise delineation of the nature of macro-micro issues than has been presented in the criminological literature thus far. The causal logics of theories depicted as applicable at the individual level have implications for their transformation into macro-level theories and/or their compatibility with existing macro-level theories.
When explicit attention is paid to such transitions, the popular classification of social learning theory with cultural deviance theory can be questioned, and the theory can be shown to be quite compatible with a modified version of Sampson’s and Wilson’s (1995) social-cultural disorganization theory of crime. A criminological theory that successfully addresses micro
macro transitions is to be preferred over theories that can be defended at only one level, or where the research at “micro” or “meso” levels is dismissed as irrelevant to the merits of a theory. James Coleman (1990: 21) expressed such an epistemological rule in his Foundations of Social Theory
, although he presents it as a standard for evaluating “good” social history: “...[G]ood social history makes transitions between micro and macro levels successfully.” Rather than attempt to integrate research at different levels of analysis or explanation, Bernard and Messner and Rosenfeld dismiss research based on data collected from individuals as irrelevant to macro-level versions of criminological theory. In contrast, we argue that the most empirically accurate macro-level theories are likely to be compatible with the causal logic of the best substantiated “micro-level” theories.
Level of Explanation, Data Sources...