The Impact of Microassumptions

Thus far, we have noted that by and large, microlevel explanations of crime dominate the discipline. We have also suggested that the results of empirical tests of microlevel hypotheses tend to produce weak to moderate empirical results. Finally, we have noted that this may have something to do with the inconsistencies between microlevel explanations and the tendency to test those theories using the traditional definition and measurement of crime that, as we shall argue later, contains significant structural influences that alters the nature of that measure of crime, and compromise its assumed microlevel nature.

Our argument is meant to propose a challenge to the validity of microlevel explanations of crime. This challenge requires additional explanation. Since the microlevel approach dominates criminology, any critical assessment of its assumptions and practices are not likely to be well received. Given that the majority of criminologists adhere to the assumptions of microlevel research and preference the legal definition of crime, the notion that those assumptions conflict with the definition and measurement of crime as a violation of the criminal law poses a critique that questions the logical validity of the methods of analysis typically employed by microlevel researchers.

We believe that our suggestion that testing microlevel theory with legal definitions is illogical and inconsistent would not be well received and would be rejected outright by those engaged in microlevel research that employs the legal definition of crime. The very nature of our argument is directly opposed to the core assumptions of orthodox criminology. Thus we suggest that these criticisms are not likely to be acknowledged, since they suggest that the way criminology has been practiced over time by the majority of academics within the discipline of criminology contain a serious validity problem and a severe logical error. This particular problem has not been recognized or addressed, and to be sure, there should be apprehension concerning the claims that we present. Criminologists might wonder if these claims are true; and if they are true, why no one else has exposed them. Others may not have thought about this problem at all, and as scientists, they may have some apprehension concerning the validity of our position particularly with respect to the traditional practices widely accepted within criminology.

Others may react to our contention on different grounds. For example, the strong ideological bias favoring microlevel explanations of crime has also been challenged by the few comments made thus far. By “ideological bias” we mean to imply that criminologists tend to favor microlevel explanations for crime even though the empirical results related to tests of those explanations are often weak to moderate as described earlier (e.g., low explained variance in equations used to explain crime and for the individual variables that compose those models; for example, see Greenberg 1999). Part of this ideological belief is that the causes of crime are to be found within individuals. The belief that the causes of crime are found at the individual level is not an epistemologically situated preference, but comes a priori to efforts to test causal explanations of crime at various levels of explanation. In other words, the assumption that individual-level explanations of crime are required is a metaphysical assumption (see Mills 1959), one that precedes empirical results. That metaphysical preference for individual-level explanations is defended despite the frequency of weak/moderate empirical results that would appear to challenge its validity. Moreover, one could argue that the utility of individual-level explanations for the causes of crime is ideological to the extent that there is significant evidence that macrolevel explanations of crime perform better than microlevel theories, at least with respect to the strength of those relationships in empirical tests where the explained variance is greater—that is, based on these general observations, one could argue that social structural differences matter more than individual differences within any particular social structure. Nevertheless, we have heard many criminologists comment, and have had numerous critiques of our own macrolevel research state, that moderate to strong goodness-of-fit indicators are simply a statistical artifact of macrolevel research. This is, however, a fallacy. For instance, Ostroff (1993, 569) has demonstrated that “aggregate-level correlation can be stronger than, weaker than, or equal to the individual correlation,” and that those outcome depend on the trends between predictors and the dependent variable. Nevertheless, we suggest the following: if there is empirical evidence that macrolevel explanations perform as well as or better than microlevel explanations (this generalization is based on our reading of the criminological literature), then why is so much of the criminological literature focused on individuals?

Microlevel explanations of crime, whether or not they state such an assumption, derive their focus on the individual from the assumption (1) that explaining the variation in crime across individuals is the correct place to begin, (2) that explaining variability in crime is necessary, and (3) that doing so requires analysis of individual differences. The assumption is typically that the variation in crime across individuals must logically be explained with other microlevel variables. For instance, the studies we previously referenced that focus on individuals all seek to explain individual-level behavior with individual-level characteristics. This conclusion may not be logical. For instance, what if there is variability across individuals with respect to crime and that variation is a product of the social structure? Since crime is widespread among the population according to self-report studies, it may be that the appearance of crime as a consequence of some individual set of conditions misrepresents what is actually driving most crime. Addressing this issue would require that microlevel researchers pay greater attention to this possibility. For example, consider Figure 2.1 that proposes that individual-level results may produce weak findings.

We are not using Figure 2.1 to suggest that criminologists engage in “the ecological fallacy”—that is, the use of aggregate-level results to draw microlevel conclusions. Instead, because the legal definition of crime is the product of social and political forces, it should be examined in relation to those forces to help explain variations in crime. Figure 2.1 demonstrates this variation in the differences in criminal definitions between and within Structures A, B, and C. As that figure hypothesizes, considerable variation exists between these structures rather than within them. This is not too hard to imagine in many cases. For instance, in studying the self-reported criminal behavior of college students for 55 different crimes, Porterfield (1943) found that every person in the sample of 337 had committed an offense. In other words, researchers could create a sample in which all participants admitted to committing an offense that violated the law. At the individual level, this means that participation in crime would not appear to be the result of individual differences, though the volume of crime still might be. Gold (1966) also found that 88 percent of his sample committed one or more delinquent acts. If, as these studies suggest, criminal and delinquent behavior is so widespread within a particular structure then how can crime and delinquency be explained by individual-l evel differences? The answer, of course, is that microlevel researchers create variation by examining specific types of crime—that is, crime can be subdivided and defined and studied to generate differences across individuals that are not visible in the aggregate. Whether or not specific crimes are the result of specific individual-level differences is a fair empirical question to address. However, such an approach still overlooks that the variability in the measure of crime produces some of the variation in crime across individuals.

If our observations are appropriate, then as a measure of crime, the orthodox definition of crime fails to include structural factors that may lead to weak/moderate results when testing individual- level theories. An alternative approach to studying crime at the microlevel might proceed by taking higher-level structural factors into account (Raudenbush et al. 2004). This is precisely what multilevel statistical models propose to do—that is, individual

Figure 2.1 Weak microlevel results and strong macrolevel results

Adapted from Ostroff (1993).

characteristics may serve as the level-one variables and structural variables can serve as the level-two variables. One could, of course, posit the opposite set of relations—that we begin with structural relationships because crime is a political construction, assess those relationships, and then add individual-level variables. Currently, the use of multilevel statistical models to examine the influence of both micro- and macrolevel determinants of crime is relatively infrequent. For instance, between 1997 and 2013, only 14.5 percent of all articles in the journal Criminology have attempted to control for structural-level variables using multilevel models with individuals. Moreover, when structural factors are used to explain crime, they generally do not control for those particular factors that influence different definitions of crime. In short, our argument is that microlevel research assumes that the measurement of the legal definition of crime is an individual-level variable when in fact it is produced at the structural level.