Introduction
What is the goal of the discipline of criminology? To those outside the field, it often appears bent on promoting a philosophy of social determinism. The leading textbooks and journal articles seem to promote the broad thesis that man is good until corrupted by “society.” There is little variation in the narrative from one study to the next. Regardless of whether the issue is crime causation or racial disparity in sentencing, somehow “good” individuals manage to form a “bad” society. The “bad” society corrupts the “good” individual by making him into a criminal or by punishing him unjustly. Critics of this narrative argue that it absolves ordinary citizens of personal responsibility while providing weak rationales for social engineering. These critics see criminology research as little more than progressive ideology disguised as social science.
To grasp the relationship between ideology and research within criminology, one must understand that it is not a free-standing discipline. For more than a century, criminology has been little more than an offshoot of sociology, which has long been dominated by Marxist or “critical” theories of social processes and institutions.
Much of this so-called critical influence can be attributed to the work of C. Wright Mills. Mills is known for several works that have helped shape this discipline, and The Sociological Imagination (1959) in particular has profoundly influenced how criminology research is conducted and interpreted.1
Although Mills condemned elitism, most notably in The Power Elite (1956), he was inclined to look down on those who did not share his perspective on the relationship between man and society—a view enlightened by what he called “the sociological imagination.” As for “ordinary men,” Mills wrote in The Sociological Imagination, “they do not possess the quality of mind essential to grasp the interplay of man and society” (4).
Mills’s use of “interplay” to describe the relationship between man and society is somewhat deceptive. He championed the view that this relationship is, for all intents and purposes, unidirectional. In making the case for employing the sociological imagination as a framework for social science research, Mills states, “[I]n our time we have come to know that the limits of ‘human nature’ are frighteningly broad” (6).
Mills consistently used human nature in scare quotes to highlight his assertion that man has no nature apart from that which is assigned to him by historical circumstances or by a variety of socio-structural determinants such as race and social class. Critiquing the work of British neurologist and psychoanalyst Ernest Jones, Mills wrote that “it is not true that man’s chief enemy and danger is his own unruly nature and the dark forces pent up within him.” To the contrary, man’s chief enemy and danger today lies in the unruly forces of contemporary society itself, with its alienating methods of production, its enveloping techniques of political domination, its international anarchy—in a word, its pervasive transformations of the very “nature” of man and the conditions and aims of his life. (p. 13)
That Mills preferred the term “nature” over the phrase “human nature” is no small distinction, for the latter suggests an innate predisposition toward choosing good or evil, while the former suggests that these inclinations are determined by social or other external conditions.
Mills’s dismissal of free will is most evident when he discusses large-scale, or macro-level, group conflict. This is particularly true when he writes of historical determinism. For example, Mills claims that “[a]s images of ‘human nature’ become more problematic, an increasing need is felt to pay closer yet more imaginative attention to the social routines and catastrophes which reveal (and which shape) man’s nature in this time of civil unrest and ideological conflict” (15). When Mills asks sociologists to employ the sociological imagination, he is asking them to commit to social determinism while ignoring human nature and the prospect of individual free will. The sociological imagination is thus limited by a priori assumptions.
In fact, Mills hoped that his legacy to the discipline of sociology would be a commitment to imagining a variety of social factors and historical circumstances sufficient to explain away both free will and the idea of an ingrained human nature. Expressing the view that his philosophical perspective should be sociology’s enduring perspective (because it is the right perspective), Mills stated that “the sociological imagination is not merely a fashion” but a quality of mind that seems most dramatically to promise an understanding of the intimate realities of ourselves in connection with larger social realities. It is not merely one quality of mind among the contemporary range of cultural sensibilities—it is the quality whose wider and more adroit use offers the promise that all such sensibilities—and in fact, human reason itself—will come to play a greater role in human affairs. (p. 15)
Mills’s perspective has become deeply rooted in sociology as well as its subdiscipline of criminology. Within criminology, Mills’s influence is evident in research on racial discrimination within the criminal justice system and in research on the causes of crime. And yet, synthesizing this perspective with reality has not always been easy. Criminologists have often deemed it necessary to bend the rules of scientific inquiry to sustain their faith in the sociological perspective and its philosophical presuppositions concerning social determinism.
The Sociological Imagination and Discrimination Research
Criminologist William Wilbanks’s Myth of a Racist Criminal Justice System is anomalous within criminology.2 His thesis is straightforward and states simply that the idea of systematic racial discrimination in the criminal justice system is a “myth.” Since its publication in 1987, the book has drawn widespread criticism within criminology. That is understandable, because Wilbanks’s thesis is rooted in stinging criticism of widely used research methods among criminologists. One of his many criticisms concerns the application of the sociological imagination to interpret unexplained variance in discrimination research.
When criminologists test statistical models using inferential statistics, they find two types of variance: explained and unexplained. Explained variance is the amount of variation in a dependent variable that is accounted for by one or more independent variables. Unexplained variance—what the independent variable or variables do not account for—is simply the leftover variance. To the extent that a statistical model explains variance it is considered to be robust. To the extent that it does not, it is considered to be weak.
Models testing racial discrimination in the criminal justice system can take many forms. For example, a researcher may decide to study racial discrimination in sentencing. In such studies, it is not uncommon to observe vast racial disparities between whites and nonwhites, at least initially. When researchers begin to control for sentencing criteria like offense severity or prior record, those differences are attenuated. Sometimes the attenuation is drastic. The disparity between black and white sentences, for instance, may be measured in years absent the proper control variables. After controlling for relevant sentencing criteria, however, the sentencing disparity may be reduced to just a few days.
The question is: How does one interpret the unexplained variance before and after introducing proper controls? The commonly accepted practice among criminologists conducting discrimination research is to interpret all unexplained variance as direct evidence of racism. Wilbanks criticized this practice on two grounds:
Some discrimination studies fail to control for legally relevant variables. A good example of this tendency is a study by Obie Clayton Jr., currently distinguished professor and department chair of sociology and criminal justice at Clark Atlanta University, which analyzed the sentences of twenty-one thousand felons convicted in Georgia between 1973 and 1980.3 After controlling for IQ, years of education, and prior record, he observed a remaining two-and-a-half-year average difference between black and white sentences. He interpreted this difference as race discrimination. Yet, only one of Clayton’s three control variables (prior record) is actually a legally relevant sentencing variable, while all other sentencing criteria (such as actual crime committed) were absent from his statistical model. Because Clayton’s study did not measure all, or even most, legally relevant sentencing variables, it is impossible to use his model to determine how much disparity was really due to extralegal variables in general—much less to race in particular.
Discrimination studies that do control for all relevant legal sentencing criteria do not measure racism directly. Studies that control for all relevant legal variables are a rarity. Even among these studies, they uniformly interpret as a “race effect” the variance that remains unexplained after all legal sentencing criteria are entered into their statistical models. In other words, they simply accuse the criminal justice system, or individual actors within the system, of racism. That accusation precludes other reasonable possibilities.
Imagine, for example, that there is a difference in courtroom demeanor between the average black and the average white convict. More specifically, imagine that the average white convict tends to have a remorseful demeanor during sentencing. Imagine, on the other hand, that the average black convict is much less remorseful during sentencing. Since courtroom demeanor is not a control variable, the results are potentially confounded.
To complicate things further, imagine that the courtroom demeanor difference is due to racism, but not racism in the criminal justice system per se. Blacks might be less remorseful or even more antagonistic toward judges due to the racism they have experienced outside the criminal justice system. Any anger or rebelliousness they choose to express in the sentencing hearing could be a residual effect of broader sources of social injustice. Therefore, imputing racist motivation to judges and other criminal justice actors would be misplaced unless the study directly measured such potentially confounding variables.
Essentially, Wilbanks’s statistical quarrel with race discrimination research is a quarrel with the legacy of C. Wright Mills. Indeed, Mills specifically asked sociologists to use the sociological imagination to see social determinism in places where it might have been overlooked. He did not believe in free will and sought to discount the role that individual choices play in the lives of ordinary citizens. (Under this ideology, the alleged racism of judges would also need to be explained away by social forces rather than individual responsibility.)
Wilbanks has shown how sociologists doing criminological research have often imagined such social determinism (in the form of racism) where it was not directly measured. He has also shown how studies appear to be rigged by omitting variables that reflect individual choices, such as type of crime committed or prior record of offending. Such studies thus inflate unexplained variance and blame it on “society.”
These highly imaginative interpretations of unexplained variance are controversial. They also raise questions of intellectual honesty and consistency. For example, when sociologists conduct criminological research that directly measures social determinism, how then do they interpret unexplained variance? That is, when the criminologist measures social forces that are said to cause crime, what does the researcher do with the variance that is left over?
The Sociological Imagination and Delinquency Research
Mills was not the only theorist over the last century who sought to explain crime by looking outside the individual to focus instead on broader social forces. Some of the most popular deterministic theories were fashioned in the 1930s and 1940s. For example, labeling theory became popular during the late 1930s due to the work of Frank Tannenbaum. Around the same time, strain theory became popular due to the work of Robert K. Merton. By the end of the 1940s, differential association theory had been popularized by Edwin Sutherland. These theories were all distinctly sociological in the sense that they assumed the inherent goodness of man and sought to explain crime as a function of external social forces.
These sociological theories were also at odds with the idea of free will and individual determinism. Prior to the formulation of labeling theory, for instance, people assumed that the choice to commit a crime produced a negative social label. Labeling theory reversed that argument, positing the label as the cause of the crime. Similarly, prior to the formulation of strain theory, people assumed that the choice to commit a crime would block social opportunities such as employment, which in turn produced strain. But strain theory reversed that argument, positing blocked opportunities and social strain as causes of crime. Finally, prior to the formulation of differential association theory, people assumed that the choice to commit a crime resulted in ostracism from conventional peers, which in turn produced delinquent peer associations. Differential association theory reversed that argument, positing those associations as root causes of crime.
For years, these deterministic theories were untested and speculative theories of crime and delinquency. Techniques designed to assess the theories using questionnaires simply did not exist when they were formulated. There were some crude efforts to measure crime and delinquency via self-reports as early as the 1940s. More sophisticated self-reported delinquency studies, which linked delinquency to progressive theories, did not become commonplace until the 1960s.
To date, thousands of studies have aimed to assess sociological theories of crime and deviance. In the process, a number of disciplinary norms have been established that govern the way such studies are conducted and interpreted. A close examination of recent criminological studies of self-reported delinquency reveals several notable characteristics of how criminologists deal with unexplained variance. These characteristics raise concerns similar to those raised by Wilbanks in his extended critique of sentencing research:
Researchers generally are reticent to discuss the explanatory value of their theoretical models. Over the course of the last ten years (2004–2013, at the time of writing this piece), the leading journal in the discipline, aptly named Criminology, has published twenty-six studies of self-reported delinquency. While each study attempts to explain individual delinquency using at least one clearly identifiable criminological theory, most of these studies do not report their actual success in explaining delinquency. In fact, only nine (or 35 percent) of those twenty-six studies actually report the amount of variance explained by the theoretical model being tested. In other words, it appears to be routine practice among criminologists simply to suppress statistics that summarize the overall predictive efficacy of the theory being tested.
It is difficult to imagine any other discipline adopting a similar practice. Yet, when one examines the studies that do report actual levels of explained variance, it is easy to understand why so many criminologists choose to ignore the issue of overall explanatory value.
Crime and delinquency researchers are consistently unable to explain the majority of the variance in the phenomenon they are examining. As stated above, among the twenty-six delinquency studies published in the last decade in Criminology, only nine report the amount of variance explained by the theoretical model. Among those nine, none managed to explain the majority of the variance in the outcome variable. In other words, for each of these studies, the amount of unexplained variance exceeds the amount of explained variance. As a result, these studies provide rather weak evidence in support of the theories they are assessing.
Empirical assessments of criminology theories are devoid of references to free will and individual determinism. Among the nine delinquency studies mentioned above, none includes any mention of free will in the discussion section. Each study begins by asserting that crime is explained by some deterministic social factor or factors. However, after the models have failed to explain the majority of the variance they sought to explain, the researchers remain mute on the issue of individual determinism. If a majority of the variance in a model is unexplained by social factors, it would seem reasonable to conclude that something else provides a better explanation. If social determinants are weak predictors, then it seems likely that free will and individual determinism are key explanatory elements. In criminology research, that possibility is consistently ignored.
In criminology research there appears to be more than just a tendency to overlook explanations that extend beyond social determinants. Researchers of good will could easily entertain a more comprehensive sociological imagination. In some cases, however, researchers appear to have outright hostility toward the concepts of free will and individual determinism. For example, sociologist Mark Konty tested strain theory with an empirical model that only explained 9 percent of the variance in self-reported delinquency.4 Despite the poor performance of a model positing social strain as a key cause of crime, he openly ridiculed the opposite view that crime is the result of human nature. Characterizing the natural propensity to commit crime as a “one-dimensional caricature,” Konty steadfastly defended the position that people are fundamentally good until forced into crime by society or, more specifically, by social strain.5
In fairness, it is worth noting that strain theory does not always produce such dismal results when subjected to empirical analysis. For example, Queen’s University professor of sociology Stephen W. Baron tested a reformulation of strain theory using homeless youths as subjects.6 In the original formulation, social strain is the difference between people’s aspirations and their opportunities to fulfill those aspirations, and social strain is supposed to cause crime. Baron’s reformulation of strain theory instead measured “anger” as a causal variable. This was predicated on the idea that anger is a natural part of social strain. In other words, when people are constrained by society and cannot get what they want, they tend to become angry. Baron reports that this “anger” variable explains 43 percent of the variance in self-reported violent crime.7
It is truly unremarkable that expressions of “anger” correlate with expressions of violent crime. After all, violent crime is often defined as an expression of anger. Therefore, Baron isn’t discovering much about crime when he finds that the same criminal outcome variable, worded in two different ways, tends to correlate with itself.
Strain theorists are not the only ones guilty of crafting “causal” variables that are little more than rewordings of the criminal outcome they are trying to explain. For example, self-control theorists have come close to explaining half the variance (as much as 48 percent) in “anti-social conduct” by using “consideration of others” as a causal variable.8 Of course, calling someone “inconsiderate” is simply another way of calling him “antisocial.” Once again, the same criminal outcome variable, worded in two different ways, tends to correlate with itself. No new intellectual ground is broken here.
It is promising for the discipline of criminology that theories once focused on macro-level causes are now being “reformulated” in order to focus on internal mechanisms. That these reformulations often produce more robust models flies in the face of the Mills tradition, which is decidedly slanted toward broader social forces. Yet, even when discussing these internal mechanisms, researchers characteristically twist them to preserve the narrative that society is ultimately responsible for every individual shortcoming.
For example, University of Georgia Distinguished Research Professor of Sociology Ronald L. Simons and his co-authors found that self-control theory explained 25 percent of the variation in self-reported delinquency. They interpreted the explained variance as suggesting that people who lack self-control (and also commit crime) had “caretakers” who fail to “teach the child self-control.”9 In other words, lack of self-control is not an individual shortcoming. It instead results from another social commodity, like wealth or power, which is not distributed in a fair and even manner.
Furthermore, what can be said of the 75 percent of the variance left unexplained by the model? Simons et al. assert that “presumably factors relating to neighborhood context account for much of the remaining variance.”10 The key word is “presumably.” They uncritically assume society is at fault even though they haven’t measured the factors upon which they have based their presumptions. This is a familiar claim in criminology research.
Other researchers react to low levels of explained variance by suggesting that the remaining variance doesn’t really matter. For example, Florida State University criminology professor Carter Hay and University of Queensland sociology senior lecturer Walter Forrest state that their “principal substantive interest is not in increasing explained variance but instead in gaining a more precise understanding of the nature of the relationship between self-control (the theory they tested) and crimes.”11 Such comments suggest that criminologists may fail to understand what high levels of unexplained variance mean. In their test of self-control theory, Hay and Forrest explained 25 percent of the variance in crime. That means the unexplained variance (75 percent) was roughly three times the explained variance. How could their model possibly become “more precise” without increasing the level of explained variance? It defies one’s sociological and logical imagination.
Conclusion
As such studies show, the goal of the discipline of criminology often appears to be the unscientific advocacy of social determinism. The basic narrative that man is good until corrupted by society appears in countless studies published in the leading journals within the discipline. Maintaining this narrative has often been accomplished by sacrificing academic standards. This sacrifice occurs in the analysis of research on crime causation and on criminal justice administration, most notably in studies of criminal sentencing outcomes.
Research on crime causation has been particularly unfruitful. In the process of trying to demonstrate that crime is caused by social forces outside the individual, criminology has consistently failed. In fact, criminology studies almost universally fail to explain a majority of the variance in crime and delinquency. Criminologists usually obscure that failure by refusing to disclose the amount of variance explained by their models. Even in the leading journal in the discipline, nondisclosure is the pattern in 65 percent of published studies within the last ten years. Among these publications, none seems willing to discuss the alternative to the thesis that crime is socially determined. Indeed, the idea that crime is individually chosen is absent in literally 100 percent of these studies. Moreover, some of the published studies, though they explain only a small fraction of crime, openly ridicule the idea that crime is a function of individual free will.
Meanwhile, one can hardly claim that research on criminal justice outcomes has been less fruitful than studies of crime causation. It is much safer to argue that outcome research, particularly sentencing outcome research, has been even less rigorous. Much of this lack of rigor appears intentional. By failing to include legal control variables, the criminologist boosts unexplained variance. This extra variance is then attributed to a variety of social determinants such as racism.
One thus finds an unacceptable double standard in the discipline of criminology. In crime causation studies, where the predictors are derived from a philosophy of social determinism as opposed to individual determinism, the goal is a robust model that explains everything. Accordingly, the evidence suggesting that the model is not robust tends to be hidden. In contrast, in criminal sentencing studies, where the predictors include individual characteristics (such as the individual crime and the individual’s prior record), the goal is a weak model that does not explain everything and thus leaves room to associate the unexplained variance with attributions of racism or other sources of social determinism. Under both standards, the sociological imagination appears to intentionally distort the evidence.
Such double standards are an affront to objective scientific inquiry. The double standard in research in criminology and criminal justice leads me to propose several specific reforms:
Editors should refuse to publish articles that fail to report explained variance. Even the top criminology journals sometimes publish studies that explain less than 10 percent of the variance in a given outcome variable, but revealing the explained variance at least gives readers the advantage of taking the research findings with a grain of salt. When this information is omitted, lawmakers and others charged with influencing public policy may be misled and give the findings more weight than they deserve.
Reviewers should demand that researchers discuss deficiencies in explanatory value. If a researcher attempts to explain a given social phenomenon using a particular perspective or theory and fails to do so, objectivity requires the researcher to consider alternate perspectives. This is particularly the case when a study leaves more variance unexplained than explained.
Editors should not publish studies of criminal decision-making that omit or fail to account for relevant control variables. Of course, not all studies omitting relevant control variables are rigged. Some sentencing studies may omit relevant control variables simply because the data are not available. For example, the researcher might not have access to inmates’ prior records. But it does not follow that such a study is worth publishing just because it isn’t rigged. Studies that compare whites and minorities and find unexplained sentencing variance without having introduced all relevant controls tell us little to nothing about racism. Criminology does not need additional incomplete sentencing studies. Criminology needs fewer but more complete sentencing studies.
C. Wright Mills probably would object to each of the above recommendations, largely because he was essentially opposed to objectivity in social science research. In fact, he did not even favor the use of quantitative research in criminology. Instead, he urged sociologists to rely on historical narratives that left room for subjectivity in all realms of analysis—hence the reliance on one’s “imagination.” Mills also envisioned social science as a political endeavor. Ultimately, he wanted the work product of sociologists to influence the public square. It is safe to say that he envisioned the role of sociology in public policy to be far greater than it is today.
Sociologists, particularly those in the subdiscipline of criminology, have listened to only half of what Mills said. They have ignored his pleas to eschew empirical research and have focused on the analysis of surveys and other quantitative sources of data. But the problem for criminologists who share Mills’s vision of policy relevance is that their studies have not provided clear enough results. They have infused their analysis with subjectivity, often to the point of haphazardly discarding control variables and relevant statistics concerning overall predictive efficacy.
Because of the influence of Mills, criminologists have dipped one foot in the realm of scientific objectivity while keeping the other in the realm of subjective imagination. This problem has made it difficult for other academics and public policy makers to take them seriously. It also, ironically, has damaged their capacity to influence the public square.
