Criminological theory in the digital age: The
case of social learning theory and digital
piracy
Introduction
The societal impact from digital technology development is undeniable; this has been
especially apparent in recent years. Such development has brought with it dramatic
improvements in the quality of life for many individuals, however harms from such
developments occur regularly. One such harm is that of digital piracy. Digital piracy
refers to the act of illegally acquiring some form of digital media through digital file
sharing (person-to-person, directly) or through illegal downloading (e.g., using peer-
to-peer file sharing networks–see Gopal, Sanders, Bhattacharjee, Agrawal, & Wagner,
2004). Today, most if not all forms of media (audio, visual, computer software, etc.)
are developed and stored using digital technology. Unfortunately, for those who rely
on the production of digital media for employment, this has become problematic.
Peer-to-peer file sharing networks, for example, have brought illegal file sharing to
ubiquity. The problem is that for many, there is now very little incentive to purchase a
product when it can be downloaded immediately and for free, though often times
illegally, from the Internet. Simultaneously, the probability of facing repercussions for
digital piracy is low, which may have played a role in creating an environment of
tolerance toward the issue.
An estimate in 2006 suggested that the costs associated with losses from music piracy
alone, worldwide, were in excess of one billion dollars (International Federation of
Phonographic Industries, 2006). More recent estimates have suggested losses from
digital music piracy at over twelve billion dollars annually (Siwek, 2007). Regarding
software piracy, research has suggested that global losses for 2005 were about thirty-
three billion dollars (Business Software Alliance, 2005). Economists have consistently
found a statistically and substantively significant link between increased file sharing
and dramatically declining record sales, beginning in about 1999, at the same time in
which file sharing became a relatively common practice—for a review see Liebowitz
(2005).1=The film industry in the United States was estimated to have lost over six
billion dollars in 2005 due to on-line movie piracy; worldwide, the costs were triple
that of the United States (Motion Picture Association of America, 2007).
Clearly, digital piracy is a significant global issue. Secondary harms stemming from
digital piracy may include lost jobs, lost tax revenue, inflated prices, and lost artistic
and creative motivations by producers. Each year, over 70,000 jobs are lost in the
United States due to digital piracy and over $400=million in tax revenue is lost
annually as a result of the crime (Siwek, 2007). Even so, attention to this problem by
scholars is limited. Criminological theory may provide a means to a better
understanding of digital piracy; hopefully this will lead to more informed social
responses to the problem. Existing theories of crime were not developed with the
digital environment in mind. With limited exception, most were developed to account
for participation in street-level type crimes and/or delinquency. Scholars have only
recently begun exploring the viability of existing crime theories in explaining
participation in technology driven crimes, such as digital piracy or malicious
computer hacking. This is critical as society becomes more dependent on technology
and people spend more time participating in the digital environment. The specific
purpose of this study was to extend what is known about digital piracy from a social
learning perspective and to help guide and extend criminological theories to
technology driven crimes; a by-product of the information age.
Akers, 1985, Akers, 1998 social learning theory (SLT) suggests that individual
differences in criminal behaviors depend on several key principles of learning.
According to SLT, the probability of engaging in crime and deviance is higher for
people who differentially associate with others who engage in deviance or elicit
favorable attitudes or definitions toward such, who are more exposed to models of
criminal behavior (directly or symbolically) from salient others, who define the act as
desirable or justified, and who anticipate a reward rather than a punishment as a result
of the act.
Differential association is typically considered the most pronounced component of the
learning process (Akers & Sellers, 2005) and can be represented by direct and indirect
(distant) interaction with people who engage in deviance or who express normative
ideals to otherwise deviant behavior. Family and friends typically represent the most
salient sources of differential association, however, interaction and exposure to
secondary peer groups may also contribute to the learning process (e.g., virtual peer
groups—see Warr, 2002). Such associations will vary in priority, duration, frequency,
and intensity. In other words, individuals whose associations start sooner, take up
more time, occur more often, and involve those who are more salient will have a
stronger impact on one's behavior, whether deviant or not (Akers, 1998). Definitions
refer to the relative composition or influence of factors associated with how one
defines the appropriateness of an act. Such involve one's rationalizations,
neutralizations, justifications, and excuses toward engaging in any given behavior.
The more someone has learned and approves of definitions favorable to deviance, the
greater the odds of engaging in such behavior. Differential reinforcement equates to
the balance of one's perception, or anticipation, of potential rewards and potential
punishments that result from a given behavior. Akers, 1985, Akers, 1998 extension to
learning theories of crime has suggested that the onset, continuity, and desistence
from crime “depends on the balance of past, present, and anticipated future rewards
and punishments for their actions” (Akers & Jensen, 2006, pp. 39-40). Restated, when
the perception of a reward is strong for a given behavior, the greater the likelihood of
onset and continuity in such behavior. Imitation refers to the modeling of a behavior
upon the direct observation of such behavior by important others. Whether imitation
occurs largely depends on the source of the behavioral model, the act itself, and
observable outcome/consequences of the behavior (also known as vicarious
reinforcement—see Bandura, 1977) in addition to other factors; however, imitation
may be more important for the onset of behavior rather than continuity in behavior
(Akers, 1998).
In developing a social structure and social learning (SSSL) model of criminal
behavior, Akers (1998) capitalized on the structural variation that stems from macro-
level predictors of offending (e.g., age, gender, race, social class, etc.) while
simultaneously accounting for individual, or group, level variation in crime causation
as explained by the principles of social learning theory (Akers, 1985, Akers, 1998).
The foundation of SSSL, according to Akers (1998), is that between individual
differences in social organization (i.e., aggregate correlates), one's location in the
social structure (i.e., individual characteristics), the impact from structural theories of
crime, and one's position in a social group/network will explain variation in rates of
crime and deviance and these differences are mediated through the principle
components of social learning theory (differential association, differential
reinforcement, imitation, and the ratio of definitions favorable or unfavorable to
offending). Thus, the factors representing one's position in the social structure
(including both micro and micro level attributes) should have an indirect effect on the
processes associated with learning deviant behavior as defined by social learning
theory (Akers, 1985, Akers, 1998).
According to Akers (1998), the basic assumption of SSSL is that “social learning is
the primary process linking social structure to individual behavior. Its main
proposition is that variation in the social structure, culture, and locations of
individuals and groups in the social system explain variations in crime rates,
principally through their influences on differences among individuals on the social
learning variables…” (Akers, 1998, p. 322). The structural variables represent macro
and micro-level correlates of offending which, according to SSSL, are mediated by
social learning variables. In developing SSSL, Akers identifies four primary
dimensions of social structure, each of which are briefly outlined below.
Differential social organization=reflects the aggregate structural correlates of crime for
a particular social system that are expected to influence rates of crime and deviance.
In SSSL, such may include ecological, community, or geographical differences across
communities. For example, social systems may vary by demographic composition
(e.g., proportion of females to males in a society) as well as other varying community
characteristics, such as being urban versus rural. These characteristics define how
social groups are organized regardless of whether reference is to society in general, to
specific culture, a specific community, or even to a subculture (Akers, 1998, Lee et
al., 2004).
According to SSSL,=differential location in the social structure=refers to the “direct
indicators of the differential location of groups or categories of individuals in the
social structure” (Akers, 1998, p. 333). Such variables represent individual
characteristics as well as one's position or role in the social structure. This category
involves characteristics such as, but not limited to, age, gender, race, social class,
religion, and marital status.
The third component of SSSL reflects=theoretically defined structural causes. For this
dimension of SSSL, Akers accounts for influences from varying structural theories of
crime such as class based theories, anomie, and social disorganization, for example.
Such factors are included to account for social conditions that, theoretically, have a
positive impact on crime and deviance. This is important for the fact that structural
theories contribute to the SSSL framework by broadening the scope of the theory.
This is done by offering a partial explanation for known correlates of crime and
deviance such as age, gender, race, class, population/community characteristics, etc.
through structural theories such as anomie, social disorganization, and conflict (Akers,
1998, p. 333).
The final component of SSSL,=differential social location in primary, secondary, and
reference groups,=involves the “…the small-group and personal networks that
impinge directly on the individual. These are the agents of informal and semiformal
social control and socialization and are referred to specifically in social learning
theory's concept of differential association” (Akers, 1998, p. 334). Specifically, this
dimension provides the social context for which each other dimension of SSSL
interacts with the individual as well to the operation of the variables involved in social
learning theory. This is important because such agents reflect the social conflict and
differential social organization of the larger society. This sets the stage for varying
social roles that are, in turn, determined by individual characteristics (Akers, 1998).
Social learning theory has received an impressive amount of empirical scrutiny in
recent decades and has typically been supported as a key predictor of many forms of
crime and deviance (e.g., Akers, 1998, Akers et al., 1979, Akers and Lee, 1996, Elliott
et al., 1989, Esbensen and Deschenes, 1998, Esbensen and Huizinga, 1993, Krohn et
al., 1985, Menard and Elliott, 1994, Verrill, 2008, Warr, 2002). More recently,
researchers have used varying components of SLT in explaining a few computer
related crimes such as computer hacking (Morris and Blackburn, 2009, Skinner and
Fream, 1997) and digital piracy (Higgins et al., 2006, Higgins and Makin, 2004a,
Higgins and Makin, 2004b, Ingram and Hinduja, 2008, Morris and Higgins, 2009,
Skinner and Fream, 1997), however, no one study to date has offered a thorough test
of SLT specific to a comprehensive measure of digital piracy using a time ordered
latent variable modeling framework. Other research has considered computer related
crime from a subcultural perspective where group communication is argued to play a
key role in participation in the behavior (e.g., Holt, 2007).
As with more traditional forms of offending, the literature suggests that exposure to
deviant peers (i.e., differential association) does at least as good of a job in predicting
these types of crimes (i.e., digital crimes) as it does in predicting more conventional
styles of offending, at least for college student samples. For example, Skinner and
Fream (1997) found modest support for differential association and for definitions in
predicting software piracy. Other findings have been mixed, suggesting that
definitions and peer association are related to physical software piracy (i.e., making
copies of software) but not when sharing occurs on-line (Higgins & Makin, 2004a).
Capitalizing on the advantages of structural equation modeling, Higgins et al. (2006)
found a significant direct effect from differential association on digital piracy, yet
SLT did not moderate the effect from an attitudinal measure of low self-control, as
defined by Gottfredson and Hirschi (1990). More recently, Morris and Higgins (2009)
focused on exploring the role of neutralizing tendencies in predicting three forms of
anticipated and retrospective digital piracy (software, music, and video piracy) while
controlling for differential association, self-control, microanomie, and several
measures of strain. Morris and Higgins’ results provided modest support for SLT, at
least among college students, in that social learning variables were consistently found
to be the strongest predictors of digital piracy. Neutralizing attitudes among serious
digital pirates has also been observed in a recent qualitative study of active persistent
digital pirates based on a differential association framework (Holt & Copes, in press).
These studies were consistent with N. L. Piquero (2005) who suggested that social
learning theory is important in understanding intellectual property theft that included
digital piracy. Each of these studies provided good evidence that SLT is likely to have
a direct link to participation in digital piracy, however, none have explored whether
SLT mediates the role of structural level correlates (macro or micro).
Regarding more traditional forms of crime and deviance, several published works
have found support in the notion that the effects of structural correlates, both macro
and micro, on offending are partially mediated by social learning theory, particularly
with regard to differential location in the social structure (i.e., micro-level correlates).
For example, Lee et al. (2004) found some support for the notion that individual level
correlates of offending are mediated in large part by social learning theory. While not
directly testing SSSL, Warr (1993) explored the impact of differential association over
varying ages of respondents participating in the National Youth Survey (NYS) and
offered some evidence that the effects of age on delinquency are mediated, at least in
part, by aspects of social learning theory. Akers and Lee (1999) explored the age-
mediating effects of both social learning and social bonding on marijuana utilizing
data from the Boys Town study. Relying on structural equation modeling, Akers and
Lee's results suggested that the age-marijuana use relationship is mediated by the
cognitive-behavioral process of learning rather than by social bonding (measured as
parental attachment, commitment, and beliefs). Though limited by cross-sectional data
that were not time ordered, Akers and Lee's findings provided at least some evidence
that at least one aspect of SSSL, differential location in the social structure (age), was
mediated by social learning theory in its effect of offending. Lonza-Kaduce, Capece,
and Alden (2006) explored the effect of gender on college drinking using the social
learning framework from a feminist theory perspective. Ultimately, Lonza-Kaduce et
al. tested whether gender, as a micro-level structural indicator, would hold in
predicting drinking behaviors as suggested by feminist theory or whether the effect of
gender would be mediated by social learning. Their results suggested that social
learning did mediate the direct effect of gender, however, a significant statistical
interaction between gender and grades was not mediated by social learning suggesting
that interaction between gender and other structural components may resist mediation
brought forth by social learning.2
Though mixed at times, extant research demonstrated that several individual-level
factors are predictive of, or related to, varying forms of digital piracy. Regarding age,
some studies of digital piracy relied on college student samples with limited age
variation and did not suggest that age significantly predicts piracy.3=It was important to
note, however, the high prevalence rates of digital piracy participation among such
samples. Clearly, digital piracy is very common among college students in the United
States (see Cheng et al., 1997, Eining and Christensen, 1991, Higgins, 2005, Hinduja,
2001, Im and Van Epps, 1992, Ingram and Hinduja, 2008, Morris and Higgins, 2009).
This reality makes age an important factor in explaining cyber related criminal
behaviors. Findings on gender and digital piracy are mixed, however much of the
extant literature on the topic has suggested that the prevalence and frequency of
engaging in digital piracy is significantly higher for males (Higgins, 2007, Hohn et al.,
2006, Ingram and Hinduja, 2008, Morris and Higgins, 2009, Skinner and Fream,
1997). Regarding race as a predictor of digital piracy, several studies reported that
non-White respondents report certain types of digital piracy more than White
respondents. For example, Morris and Higgins (2009) found that non-White
respondents were more likely to report video piracy; however, this was not the case
for music or software piracy. Hollinger (1993) found that for software piracy, Asian
and Hispanic students were more likely to endorse the behavior while Black
respondents were the least likely group of respondents to engage in software piracy.
Taken together, the literature on digital piracy suggested that many structural
correlates of digital piracy, as a behavior, warrant further exploration through a more
complete theoretical model, such as Akers, 1985, Akers, 1998 social learning model.
Akers, 1985, Akers, 1998 social learning theory has suggested that the effects of most
individual- and macro-level predictors of crime should, at a minimum, be partially
mediated by social learning theory. Further, the theory holds that the process of social
learning should explain a substantial amount of within individual variation in
offending behaviors. The present study explored the viability of the SLT model, short
of a full exploration of Akers’ SSSL framework, in explaining social structural
variation, mainly at the micro-level, on a latent measure of digital piracy.
References
M.Peitzet al.
Piracy of digital products: A review of the economics literature
Information Economics and Policy
(2006)
J.H.Imet al.
Software piracy and software security measures in business
schools
Information and Management
(1992)
R.L.Akers
Deviant behavior: A social learning approach
(1985)
R.L.Akers
Social learning and social structure: A general theory of crime
and deviance
(1998)
R.L.Akerset al.
The empirical status of social learning theory of crime and
deviance: The past, present, and future
R.L.Akerset al.
Social learning and deviant behavior: A specific test of a
general theory
American Sociological Review
(1979)
R.L.Akerset al.
A longitudinal test of social learning theory: Adolescent
smoking
Journal of Drug Issues
(1996)
R.L.Akerset al.
A longitudinal test of social learning theory: Adolescent
smoking
Journal of Drug Issues
(1999)
R.L.Akerset al.
Criminological theories: Introduction, evaluation, and
application
(2005)
A.Bandura
Social learning theory
(1977)
