1
CHAPTER
1
INTRODUCTION
Developing trust between the police and the communities they protect has become
a primary concern for policing scholars, law enforcement executives, and policymakers
alike. In particular, allegations of racial bias in policing that stem from the shootings of
unarmed African Americans in various cities across the United States have caused a crisis
in public trust in the police (Nix & Wolfe, 2015, 2016; Pyrooz et al., 2016; Rosenfeld,
2016; Wolfe & Nix, 2016). Tyler’s (1990) theory of police legitimacy argues that the
primary contributor to legitimacy evaluations is procedural justice (or the perceived
fairness of the process used to reach a decision), but that perceptions of distributive
justice (or the perceived fairness of outcomes) are also an important predictor of
legitimacy attitudes. The current crisis in public confidence in policing lends itself to a
distributive, rather than procedural, claim of injustice due to its focus on racial bias in
policing outcomes such as stops, searches, arrests, and use of force. Yet, criminological
research into claims of racial bias tend to focus on the objective question of whether
injustice exists (e.g., comparing aggregate stop data to racially sensitive benchmark data),
rather than why individuals’ subjective perceptions of injustice are present.
To date, distributive justice remains a theoretically underdeveloped concept in the
field of criminal justice. By contrast, the field of social psychology has extensively
developed definitions, theoretical propositions, and potential outcomes of distributive
justice over the past several decades (e.g. Adams, 1965; Folger, 1986; Jasso, 1980;
2
Markovsky, 1985). Using the existing literature on justice in the field of social
psychology, this dissertation intends to further refine Tyler’s (1990) theory of procedural
justice and legitimacy. In particular, this dissertation proposes that perceptions of
procedural justice influence an individual’s perceptions of distributive justice.
Additionally, a new outcome for the field of criminal justice is borrowed from social
psychology – justice-restoring responses. Justice-restoring responses are actions, such as
filing a complaint, taken directly in response to a sense of injustice, in an attempt to
rectify the perceived injustice.
The dissertation thus proceeds as follows: Chapter 2 reviews the history and
development of justice frameworks and the theories of distributive and procedural justice.
Chapter 3 reviews the state of the existing literature on justice frameworks in criminology
and criminal justice. Chapter 4 then develops an adapted theoretical model that re-orients
distributive justice within the existing Tylerian model of procedural justice while also
including a new outcome of justice evaluations, justice-restoring responses. A method for
testing these propositions is presented in Chapter 5. Chapter 6 presents results testing the
propositions. Finally, Chapter 7 reviews the refined theoretical model and discusses its
implications.
3
CHAPTER 2
HISTORY AND DEVELOPMENT OF JUSTICE FRAMEWORKS
For centuries, research on “justice” has focused on the question of what is fair or
what is perceived to be fair. Philosophers have tackled the issue since the times of Plato
and Aristotle (Fleischacker, 2004), but scientific research on the subject exploded with
the emergence of the field of social psychology and the contributions of scholars such as
Homans (1961), Adams (1965), Thibaut and Walker (1975), and Lind and Tyler (1988).
Criminology and criminal justice entered this field relatively late – especially for a field
with justice in its name – with Tyler’s (1990) seminal work Why People Obey the Law.
Throughout the development of justice research, scholars have focused on three
types of justice perceptions: distributive justice, procedural justice, and interactional
justice (Brockner et al., 1997; Hegtvedt, 2006; and Tyler, 1990). Distributive justice
evaluates the perceived fairness of the distribution of outcomes among a group (Adams,
1965; Folger, 1986; Jasso, 1980; Markovsky, 1985). Procedural justice examines the
perceived fairness of the process used to reach the decision regarding outcomes (Lind &
Tyler, 1988; Thibaut & Walker, 1975, Tyler, 1990). Interactional justice focuses on the
perceptions of the treatment an individual receives by the decision-maker (Bies, 2001;
Bies & Shapiro, 1987). Interactional justice and procedural justice have considerable
overlap, with both highlighting concepts such as respect, neutrality, and honesty (Bies,
2001; Hegtvedt, 2006; Tyler, 1990). As a result, some scholars have merged the concepts
4
of interactional and procedural justice (see discussion in Tyler & Blader, 2003; c.f.
Colquitt, 2001).
While the concept of justice is important as a goal in and of itself, social
psychologists have also focused on the consequences of feelings of injustice and
unfairness. For criminologists in particular, perceived fairness in interactions with
criminal justice personnel has been shown to be related to many positive outcomes
including the improved legitimacy of criminal justice institutions (Hough et al., 2010;
Reisig, Bratton, & Gertz, 2007; Reisig, Tankebe, & Meško, 2012; Sunshine & Tyler,
2003; Tankebe, 2013; Tyler, 1990), cooperation with legal authorities (Bradford, 2014;
Jackson et al., 2012a; Reisig & Lloyd, 2009; Reisig et al., 2007; Reisig et al., 2012),
compliance with legal authorities and the law (McLean & Wolfe, 2016; Paternoster et al.,
1997; Reisig et al., 2007; Tyler, 1990), and public confidence and support for legal
authorities (Bradford, Jackson, & Stanko, 2009; Jackson & Sunshine, 2007; Sunshine &
Tyler, 2003).
While the positive outcomes from the perceived fairness of criminal justice actors
have been relatively well documented, there is still indeterminacy in the theoretical
underpinnings of these relationships. Tyler’s (1990) original work on perceptions of
justice and the criminal justice system highlighted the importance of legitimacy as a
mediator between perceptions of justice and the positive outcomes of compliance with
legal authorities and the law. Other theories, such as the group-value model (Lind &
Tyler, 1988) and the group engagement model (Tyler & Blader, 2003), argue that social
identity serves as an important link between perceptions of fairness and compliance with
group norms. Finally, criminologists have also argued that perceived injustice acts as a
5
strain that can cause negative emotions leading individuals to resort to maladaptive
coping mechanisms, such as criminal behavior (Agnew, 1992). Though the intervening
mechanism is still debated, it is generally agreed upon that perceptions of fairness are
linked to a variety of positive outcomes.
Scholars have also debated the relative importance of procedural justice and
distributive justice. Early research on justice focused almost exclusively on distributive
justice (Adams, 1965; Homans 1961), while later research began to highlight the
importance of procedural justice (Thibaut & Walker, 1975; Lind & Tyler, 1988). Within
the field of criminology, Tyler (1990) originally presented both distributive justice and
procedural justice as antecedents of legitimacy evaluations with a greater emphasis on the
role of procedural justice. Later research, however, began to de-emphasize distributive
justice with scholars describing Tyler’s (1990) legitimacy model as the “process-based
model” (Hough et al., 2010; Reisig & Bain, 2016; Reisig & Lloyd, 2009; Tyler & Huo,
2002; Wolfe, 2011; Wolfe & Piquero, 2011). While some of this research has continued
to examine the importance of distributive justice in research on perceptions of justice in
criminology (Engel, 2005; Reisig et al., 2007; Wolfe et al., 2016; Wolfe & Piquero,
2011), much of it has dropped distributive justice altogether (Bradford, 2014; Bradford et
al., 2009; Jackson et al., 2012). Recently, Tankebe (2013) argued for a re-
conceptualization of criminology’s research on legitimacy and justice frameworks with
renewed focus on distributive justice (see also, Sahin et al., 2017). Due to the limited
research on distributive justice, its role in the larger justice framework of criminal justice,
especially with regards to legitimacy, remains uncertain.
6
The Philosophy of Justice
At its core, justice research is concerned with the question of what is considered
fair or just. Early academic writings on justice were concerned with the objective
standard of distributive justice (Fleishacker, 2004). That is to say, scholars were not
concerned so much with what individuals considered to be fair, but what distribution of
outcomes achieved some philosophical standard of justice. This research dates back to
the writings of Aristotle, but can be traced through a number of other philosophers such
as Adam Smith, Jean-Jacques Rousseau, Immanuel Kant, John Rawls, and Karl Marx
(Fleishacker, 2004).
Theories and philosophies of distributive justice tend to follow one of three
principles – equity, equality, or needs (Hegtvedt, 2006). The principle of equity argues
that the distribution of resources should be commensurate to the contributions an
individual makes to society. An equality principle argues for the objectively equal
distribution of resources regardless of the individual’s contributions. Finally, the needs
principle argues for the distribution of outcomes based on the needs of the individual
(Hegtvedt, 2006). The most prominent philosophers of justice, such as Aristotle, Adam
Smith, and John Rawls, tended to focus on the idea of equity in the distribution of
resources among citizens in a society (Fleishacker, 2004).
While these philosophers’ arguments were typically focused on the distribution of
wealth and resources in a society, criminology also experienced a similar argument in the
18
th
century. Cesare Beccaria (1764/1986) and Jeremy Bentham (1781/2000), two early
criminological philosophers, argued for the importance of punishing a criminal
proportional to the violation of the law that the criminal committed. These arguments are
7
often referenced in criminology classes as being relevant to the principles of deterrence,
but they also represent the first attempts at establishing a distributive justice principle
akin to the equity principle within the field of criminology.
Temporarily jumping ahead in the history of justice research, the needs principle
has also been used in the field of criminal justice by scholars advocating for rehabilitative
corrections. Andrews, Bonta, and Hoge (1990) put forth the risk-need-responsivity model
for offender treatment that suggests the treatment received by the offender in the
correctional system should be based in part on the needs principle. This places needs at a
much lower status in the criminal justice system than the equity principle, which serves a
foundational role in deciding the punishment an individual receives. Instead, needs
becomes a guiding principle for what type of treatment an offender receives after
sentencing. Still, there is an obvious connection to the distributive justice needs principle
as the allocation of resources within the correctional system is based not on the crime an
individual committed but on the needs of the individual.
These considerations of distributive justice, whether based on equity, equality, or
needs, were exclusively concerned with identifying the “correct” distribution of
outcomes. Beccaria (1764/1986) and Bentham (1781/2000) were not concerned with
what punishment individuals perceived as fair, but with what punishment actually was
fair. Andrews and colleagues’ (1990) use of the needs principle is similarly concerned
with identifying the correct allocation of resources within the criminal justice system.
Later justice theorizing would shift its focus to examine individuals’ perceptions of
justice.
8
The Emergence of Science
The arguments of Adam Smith, Karl Marx, Cesare Beccaria, and Jeremy Bentham
occurred early in the history of the field of sociology. At this point in the development of
sociology, theories were typically written discursively with elaborate explanations for
phenomena observed in society and little to no empirical support or prospective
hypotheses (Lepenies, 1988; Markovsky, 2016). Lepenies (1988) suggests that, in its
early days, the field of sociology existed somewhere between literature and science. In
describing the state of classical sociological theorizing, Markovsky (2016; p. 3) notes that
“only selective empirical validation is forgivable.” However, as the field of sociology
began to mature, it evolved into a social science.
Homans’ (1961) study of human behavior represents this shift in thinking from
the philosophical to the scientific for justice research. While Homans (1961) cites
previous studies rather extensively throughout his work, including some research activity
he and his colleagues conducted on distributive justice, his Social Behavior: Its
Elementary Forms is the first work to begin stating theoretical propositions regarding the
perception of justice in social groups. As Adams (1965, p. 292) would later comment,
“the concept is not new [but]…In the hands of Homans … the concept of distributive
justice has taken on the articulated character of what may be more properly called a
theory.” Homans’ theoretical propositions led to hypotheses concerning when individuals
would feel injustice and what their emotional responses to injustice would be based on
the equity principle. In so doing, social psychology’s focus shifted from philosophical to
scientific thinking. Identifying the just distribution of outcomes is a philosophical
exercise that attempts to identify some universal truth in establishing justice through
9
discussion and academic arguments. Identifying what people perceive as a just
distribution of outcomes is a scientific endeavor that can be assessed through prospective
theorizing and stringent empirical tests.
Following Homans’ (1961) lead, social psychologists continued to develop and
reformulate theories of distributive justice through the 1960s, 1970s, and 1980s (see
Adams, 1965; Folger, 1986; Jasso, 1980; Markovsky, 1985). This research agenda
continued to highlight and refine the importance of basing outcomes off the relative
inputs of the individual as originally hypothesized by Homans (1961). Though Homans
had focused on emotional reactions to injustice, Adams’ (1965) work began to use
perceptions of distributive justice as a predictor of other attitudes and behaviors. This
contribution would prove fruitful for future researchers in describing the outcomes of
perceptions of distributive justice (Folger & Konovsky, 1989; Markovsky, 1985;
Skarlicki & Folger, 1997).
Following the popularity of distributive justice theory, Thibaut and colleagues
(1974) proposed the idea that evaluations of process were important predictors of
fairness. Thibaut and Walker (1975) then formalized this concept into their theory of
procedural justice. Although the modern conceptualizations of distributive justice and
procedural justice clearly distinguish procedural justice’s role in evaluating process and
not outcome (e.g. Tyler, 1990; Tankebe, 2013), Thibaut and Walker’s original
formulation did not have such a clear distinction. Rather, their theory focused on the idea
that procedural fairness created distributive fairness. Thus, the primary goal of the study
of process was still on creating fair outcomes. Furthermore, this early theory of
procedural justice attempted to identify the most just procedure for different situations
10
rather than examining the consequences of evaluations of procedural justice (Thibaut &
Walker, 1975). This stands in contrast to later procedural justice theories that attempt to
use perceptions of fairness as a predictor of other important normative concepts (Folger
& Konovsky, 1989; Lind & Tyler, 1988; Tyler, 1990; Tyler & Blader, 2003).
Building on this early theory, Lind and Tyler (1988) attempted to shift the focus
of procedural justice from outcomes to the process and the consequences of process
evaluations. In so doing, they incorporated a concept put forth in earlier research “that a
fair process also has value as an end in itself” (Tyler & Folger, 1980, p. 283).
Additionally, Lind and Tyler (1988) address the distinctions between objective and
subjective procedural justice. Objective procedural justice, similar to the philosophies of
distributive justice, is concerned with the preferred or best process for a given situation.
This was the subject of Thibaut and Walker’s (1975) investigation into procedural justice
in the courtroom. Subjective procedural justice, on the other hand, is concerned with the
process’ effect on individual assessments of fairness. This is the concern of the Lind and
Tyler (1988) theory. Rather than merely determining which process is “best” Lind and
Tyler concerned themselves with how people evaluate procedural fairness and how these
evaluations impact other attitudes and behaviors. In this way, the Lind and Tyler theory is
similar to the work of Adams (1965) who was interested in different evaluations of
distributive justice and how those evaluations impact other attitudes and behaviors.
Thibaut and Walker’s (1975) new concept of procedural justice generated a large
amount of interest and increased research into the 1980s, especially among researchers
exploring organizational behavior (e.g. Alexander & Ruderman, 1987; Folger, 1977;
Folger, 1987; Greenberg & Folger, 1983). As research into this new concept was being
11
undertaken, Bies and Shapiro (1987) noted the possibility that another concept was being
included in assessments of procedural justice; interactional justice. Specifically, Bies and
colleagues noticed in a series of studies that individuals assessed the manner in which a
procedure was applied, in addition to the procedure itself (Bies, 1987; Bies, 2001; Bies &
Moag, 1986; Bies & Shapiro, 1987). In their conception of the issue, procedural justice
applied to the actual process itself, while interactional justice applied to the manner in
which the authority figure applied that process. Thus, interactional justice was defined as
“the quality of interpersonal treatment they receive during the enactment of
organizational procedures” (Bies & Moag, 1986, p. 44). Other scholars did not believe
this distinction to be important and argued that interactional justice was merely a term for
the interpersonal component of procedural justice (Folger & Bies, 1989; Lind & Tyler,
1988; Tyler & Bies, 1990; Tyler & Lind, 1992). As such, later theories of procedural
justice began to incorporate interpersonal treatment as a component of procedural justice
(e.g. Tyler & Blader, 2000, 2003; Tyler & Huo, 2002).
Moving to Criminal Justice
While procedural justice and distributive justice were becoming increasingly
important topics within the social psychology literature, research in criminology and
criminal justice largely ignored these theories. Much of the criminological literature
(including the beginning of this dissertation) recognizes Tyler’s (1990) Why People Obey
the Law as being the first attempt to bring these concepts into the realm of criminological
research. However, it is worth noting that the original research on procedural justice
evaluated perceptions of fair processes in courtroom proceedings (Lind & Tyler, 1988;
Thibaut & Walker, 1975). Additionally, Tyler and Folger (1980) conducted research on
12
the impact of distributive and procedural justice on satisfaction with the police a decade
before Tyler’s seminal work was published. Still, the concepts of distributive and
procedural justice were not widely used by criminal justice scholars until Why People
Obey the Law in 1990. In fact, this dissertation’s author could find only a single mention
of the term procedural justice before 1990 in the flagship journal Criminology (Cavender,
1984). Furthermore, this mention only used procedural justice in reference to the
philosophies of justice by Kant and Rawls (Cavender, 1984), rather than the scientific
work by Thibaut and Walker or Tyler and Folger.
Throughout the 1990s, procedural justice received more attention from the field of
criminology. Notably, the second earliest mention of procedural justice that this author
could find in Criminology occurred when Agnew (1992) referenced procedural justice in
his initial formulation of general strain theory. While Tyler argued extensively for the
inclusion of procedural justice in the study of criminal justice authorities throughout the
1980s and 1990s (Casper, Tyler, & Fisher, 1988; Tyler, 1987; Tyler, 1988; Tyler, 1989;
Tyler & Folger, 1980), a major test of the theory in criminal justice from researchers
other than Tyler did not occur until 1997 (Paternoster et al., 1997). Paternoster and
colleagues re-analyzed data from the Milwaukee Domestic Violence Experiment to
demonstrate that the perceived fairness of the police had an impact on future assault rates.
As the field of criminology moved into the 21
st
century, research on procedural
justice exploded. Evaluations of Tyler’s “process-based model” of policing proved
fruitful as scholars would connect perceptions of fairness in process to a number of
favorable outcomes (e.g. Bradford, et al., 2009; Engel, 2005; Hough et al., 2010; Reisig
et al., 2007). As the evidence continued to support the theoretical model formulated in
13
Why People Obey the Law (Tyler, 1990) and refined by Tyler and Huo (2002), procedural
justice would become an integral part of police reformers’ agenda. In fact, procedural
justice became such an important topic that improving the community’s trust in police
through procedural justice would form a pillar of the recommendations put forth in the
final report of the President’s Task Force on 21
st
Century Policing (2015).
Throughout this wave of criminological research on procedural justice, the issue
of distributive justice remained largely ignored. Within the fields of social psychology
and organizational management, distributive justice remained an important consideration
in studies of perceptions of fairness and reactions to those perceptions (Clay-Warner,
Hegtvedt, & Roman, 2005; Griffith, Harvey, & Lusch, 2006; Hegtvedt & Killian, 1999;
Hsu, Anen, & Quartz, 2008; Lim, 2002; Martinez-Tur et al., 2006; Skarlicki & Folger,
1997), however, few scholars would examine perceptions of distributive fairness within
their research on the police, courts, or corrections (for notable exceptions see Engel,
2005; Tankebe, 2013; Wolfe et al., 2016). Recent criminological research has raised the
issue of distributive justice and urged scholars to reconsider the concept in their
examinations of perceptions of fairness (Sahin et al., 2017; Tankebe, 2013). However, the
limited research into distributive justice within criminology makes placement of the
concept within the existing theoretical model undetermined.
Distributive Justice – Social Psychological Theories
Distributive justice research in the field of social psychology has almost
exclusively focused on the equity principle. The early philosophical writings on the
subject by Aristotle suggested that the distribution of goods should be based on the
amount of work done by the individual. In writing the first scientific theory of
14
distributive justice, Homans (1961) adhered closely to this same principle. Specifically,
Homans (1961, p. 75) proposed that, “a man in an exchange relation with another will
expect that the rewards of each man be proportional to his costs… and that the net
rewards, or profits, of each man be proportional to his investments.” Thus, as an
individual provides greater input (cost or investment), the individual should receive
greater output (rewards or profits). Furthermore, Homans argued for a “rule of
distributive justice” that established a comparison relationship for determining
perceptions of fairness. Specifically, “if the investments of two men, or two groups, are
equal, their profits should be equal” (Homans, 1961, p. 244). Adams (1965) would go on
to mathematically formalize this statement as:
=
While this relationship seems simplistic, Homans (1961) noted that establishing
justice within a group or exchange relationship would actually be quite complicated.
While Homans believed the rule of distributive justice to be universal – that every
individual would desire a relationship where the ratio of outcome to inputs was
equivalent across individuals – he recognized that the means for assessing this ratio
would not be universal. That is, individuals differ in their assessments of what should be
considered valid inputs and valid outputs. If these concepts are not agreed upon it would
be difficult, if not impossible to establish a situation where the ratios are equivalent for all
parties involved.
Another major contribution of Homans’ (1961) initial theory of distributive
justice regarded the emotional reaction to conditions of injustice. Though the
mathematical formula suggested by Homans (1961) and formalized by Adams (1965)
15
treated violations of the conditions of justice equally in each direction, Homans believed
that the content of the emotional reaction to injustice would differ by direction. “The
more to a man’s disadvantage the rule of distributive justice fails of realization, the more
likely he is to display the emotional behavior we call anger” (Homans, 1961, p. 75).
Injustice that advantaged an individual, on the other hand, was more likely to cause guilt
than anger. This proposition is especially important when discussing the concept of
injustice within the framework of Agnew’s (1992) general strain theory, which will be
discussed in more detail later in this chapter.
Adams’ (1965) work on distributive justice remained largely consistent with
Homans’ (1961) original propositions. In addition to providing the mathematical formula
for comparing rewards and outputs, Adams would also suggest that conditions of
injustice could lead to not only emotional reactions like anger, but to behavioral reactions
as well. In particular, Adams argued that perceptions of injustice in the workplace could
lead individuals to change their input behavior or even leave their job altogether.
Markovsky (1985) would later term these and other behavioral reactions to perceived
injustice (such as formal complaints) justice-restoration attempts. Markovsky (1985)
viewed justice restoration attempts as behaviors that attempted to return a situation to the
conditions of equity.
Jasso (1978) noted several flaws in Adams’ (1965) mathematical formulation of
the justice condition. In particular, Jasso noted that Adams’ formula provided no easy
measure of how much justice was being experienced. That is, solving the formula would
tell whether or not the condition of justice was met and which individual had a larger
16
ratio than the other, but not how severely it was being violated. Jasso thus transformed
Adams’ formula to:
=
−
In this equation, when the justice evaluation is equal to 0, justice is achieved.
Jasso (1978) then argued that the experience of unfavorable injustice was more
impactful than the experience of favorable injustice. An accurate justice evaluation
function should highlight this fact and place greater injustice evaluations on unfavorable
experiences. Thus, Jasso settled on a justice evaluation function where the individual’s
justice evaluation was the natural log of the ratio of actual outcome to the just outcome:
= ln (
)
Under the conditions of this new formula, justice is achieved when the justice evaluation
is equal to zero. However, when the actual outcome is less than the just outcome the
justice evaluation should move away from zero more rapidly than when the actual
outcome is greater than the just outcome. This point is illustrated in Figure 2.1. When
actual outcomes are 50% greater than is just (actual outcome/just outcome=1.5), Jasso’s
justice evaluation finds a value of .41. When actual outcomes are 50% less than is just
(actual outcome/just outcome=0.5), Jasso’s justice evaluation finds a value of -.69. Thus,
unfavorable outcomes create more severe deviations from the just condition than
favorable outcomes.
17
Figure 2.1. Justice Evaluation Function
Following Jasso’s (1978) justice evaluation formula, Markovsky (1985) noted two
problems in using the justice evaluation function in research. First, there is no clear
method for determining the value of the just outcome. Markovsky (1985) solved this
problem by going back to Adams’ (1965) original justice function:
=
Then, Markovsky simplified to create a referential rule:
=
∗
Here, “Person A’s outcome” becomes the just outcome the individual should receive.
Markovsky (1985) then noted that the base of Jasso’s log function was arbitrarily
set to the exponential constant, e. Markovsky argued that the base of the log function
should be set to the amount of justice indifference that an individual had towards the
situation. If an individual had a higher degree of justice indifference, the line in Figure
2.1 would flatten and deviations from the just condition would not create a great sense of
injustice. If an individual had a lower degree of justice indifference, the line in Figure 2.1
would steepen and deviations from the just condition would be met with greater
18
evaluations of injustice. Thus, Markovsky’s (1985) variable base allowed individuals to
evaluate justice differently depending on the level of investment he or she had in the
present situation.
While mathematical discussions of evaluations of justice may seem unnecessarily
complex, it is essential to understand the core contributions of these principles. First,
evaluations of distributive justice are based on comparisons to other individuals or other
groups. This is seen in Adams’ (1965) formula, as well as Markovsky’s (1985) formula
for determining a just outcome. Second, unfavorable deviations from a just distribution
create greater feelings of injustice, as seen in the use of the log function in both Jasso’s
(1978) and Markovsky’s (1985) formulas. If an individual’s justice evaluation is used to
predict some future behavior, the difference between favorable and unfavorable situations
of injustice need to be seen as a key measurement issue. Finally, the more indifferent an
individual is to injustice, the less deviations from the just outcome will matter. This is
demonstrated in Markovsky’s (1985) variable base for the log function.
In applying the concepts of distributive justice to the field of criminal justice, all
three of these contributions may prove crucial to understanding distributive justice’s role
in the larger justice framework. Criminal justice scholars have yet to consider how
individuals form perceptions of distributive justice. Is a reference formula similar to those
presented here used or is some other conceptualization for evaluating distributive justice
needed? Researchers in criminal justice should probably expect Jasso (1978) and
Markovsky’s (1985) assertions that unfavorable conditions are met with a greater sense
of injustice than favorable conditions to also apply to distributive justice in criminology.
Indeed, criminologists have already considered this possibility in explorations of
19
procedural justice with mixed results (Bradford et al., 2009; Maguire, Lowrey, &
Johnson, 2017; Skogan, 2006, 2012). Finally, justice indifference may vary depending on
the role of the individual in the criminal justice context. Accounting for justice
indifference can enable a more complete and nuanced understanding of evaluations of
justice in criminal justice.
Multilevel Distributive Justice
In reviewing the state of distributive justice theories in social psychology,
Markovsky (1985) notes that these theories attempt to explain justice-restoring responses,
both emotional and behavioral, at the individual level (Adams, 1965; Folger, 1986; Jasso,
1978). Markovsky's (1985) own individual-level theory, tested workers’ complaints
regarding pay as the justice-restoring response of interest. While these individual-level
assessments necessarily form the basis of our understanding of distributive justice, the
potential for distributive justice theories to explain aggregate-level factors like social
change has not gone unnoticed (Cook & Hegtvedt, 1983; Jasso, 1980, 1983; Markovsky,
1985).
Jasso (1980; 1983) first argued that the aggregate rate of justice-restoring
responses could be calculated from individual-level justice-restoring responses.
Markovsky (1985) then extrapolated this argument to make the assertion that individual-
level justice-restoring responses could be used to explain collective behavior that he
termed aggregate justice-restoring responses. Both Markovsky (1985) and Jasso’s (1980;
1983) multilevel theories of distributive justice can easily be illustrated by the traditional
Coleman’s (1990) boat of multilevel sociological theories (Figure 2.2). In multilevel
distributive justice theory, macro-level factors of distribution (e.g. social inequality, racial
20
discrimination, etc.) affect the assessment of fairness for the individual. This assessment
then affects the likelihood of the individual engaging in a justice-restoring response. The
individual justice-restoring responses create an aggregate rate of justice-restoring
responses or organize into collective behavior forming an aggregate justice-restoring
response.
Figure 2.2. Coleman Boat
This multilevel approach to distributive justice may also prove fruitful in
exploring phenomena in the criminal justice system. For example, the perceived unfair
distribution of use-of-force by policing agencies since August of 2014 has led to the
development of the Black Lives Matter movement (Edwards & Harris, 2016). This
movement is consistent with Markovsky’s (1985) description of an aggregate justice-
restoring response. It is organized, collective behavior that is attempting to restore equity
to the distribution of the use-of-force. As such, the framework of multilevel distributive
justice theory may be helpful in exploring the causes and consequences of the Black
Lives Matter movement or other collective action brought against the criminal justice
system for perceived injustices.
A Note on Distributive Justice Measurement: Organizational Management Research
The social psychological theories of distributive justice discussed to this point
have primarily focused on individuals’ perceptions of justice based on a comparison of a
21
particular outcome to other outcomes. Thus, research on these theories typically focuses
on distributing outcomes to subjects or fictitious third parties in a particular manner and
assessing subjects’ sense of injustice and their tendency to engage in justice-restoring
responses (Folger, 1977; Jasso, 1978; Jasso & Rossi, 1977; Markovsky, 1985).
Manipulation of the outcomes received by subjects or third-parties has the added
advantage of allowing researchers to implement experimental designs which provide for
stronger internal validity (Berk, 2005; Campbell & Stanley, 1963; Shadish, Cook, &
Campbell, 2002). Testing the propositions of distributive justice theories using these
experimental methods allows the researcher to state, with a considerable degree of
certainty, that manipulating the objective conditions can alter overall perceptions of
fairness or justice-restoring responses. These social psychological experiments are,
therefore, well-suited for determining the validity of distributive justice theories.
However, theories of justice and fairness have frequently linked assessments of justice to
aspects of identity and relationships within organizations (Clay-Warner et al., 2005; Lind
& Tyler, 1988; Skitka, 2003; Tyler & Blader, 2003; Wenzel, 2001). Laboratory
experiments with hypothetical circumstances are unlikely to be able to effectively
simulate concerns over group membership such as identity or relationships with other
group members. Thus, these experiments are likely to have difficulty determining
whether organizational relationships or identification with decision-makers impacts
assessments of justice.
Contemporary research using distributive justice in fields such as organizational
management has employed a different research design and measurement strategy (e.g.
Clay-Warner et al., 2005; Lim, 2002; Skarlicki & Folger, 1997). These researchers have
22
used scales of perceptions of outcome fairness as a measurement of distributive justice.
These scales ask individuals about the specifics of distributive justice within their
workplace context using questions such as, “Where you work, the amount of pay
employees receive is distributed fairly,” or, “The overall rewards workers receive where
you work are fairly distributed” (Clay-Warner et al., 2005, p. 95). Thus, the measure of
distributive justice is specific to the individual and takes into consideration the difference
in values and relationships that individuals may have. This makes it difficult to determine
how changing the outcome distribution would impact individuals in a broader theoretical
context, but makes it easier to understand how factors such as organizational commitment
or social identity impact these assessments and their consequences. This advantage
becomes especially important when combining distributive justice research with
procedural justice theories that highlight the importance of social identity and
organizational commitment.
Procedural Justice
After social psychology’s early work focused on outcomes, Thibaut and
colleagues (1974) and Thibaut and Walker (1975) began to shift the attention of justice
researchers from outcomes to processes. Lind and Tyler (1988) would later praise
Thibaut and Walker’s (1975) shifting focus, calling it “a seminal event in the emergence
of the social psychology of procedural justice. Although the study of process has a long
history within social psychology, it was Thibaut and Walker who combined the study of
process with an interest in the psychology of justice to initiate the study of procedural
justice” (Lind & Tyler, 1988, p. 5).
23
In arguing for the importance of assessments of process, Thibaut and Walker
(1975) noted that the use of fair procedures is the method most likely to lead to fair
distributions. According to their theory then, the process is a means of achieving
distributive justice. Additionally, this early theory contended that “the just procedure…is
a procedure that entrusts much control over the process to the disputants themselves and
relatively little control to the decision maker” (Thibaut & Walker, 1975, p. 1-2).
Procedural justice is achieved when individuals have greater control and decision makers
have little control. Scholars typically conceptualize control as influence (e.g. Tyler,
Rasinski, & Spodick, 1985). So, greater control is seen where an individual has greater
influence over the process and outcome. For example, being allowed to present evidence
that is perceived to influence a judge’s verdict would be seen as having a high degree of
control. As such, Lind and Tyler (1988) refer to this conceptualization of procedural
justice as the control version of procedural justice.
While Thibaut and Walker’s (1975) emphasis on procedures was a watershed
moment in the study of justice, their theory was not without its limitations. First, the
study focused exclusively on legal proceedings and often measured which types of legal
proceedings would be preferred. This limits the ability to generalize these findings to
other contexts. Second, the emphasis of the study in finding the best process rather than
on finding the consequences of fair or unfair processes limited its appeal to social
psychologists and stood in contrast to the work being done in distributive justice (Adams,
1965; Folger & Konovsky, 1989; Jasso, 1980; Markovsky, 1985). Third, the use of
control as the only predictor of procedural justice is uninformative in situations where
one of the disputants is also the decision maker (e.g. police-citizen interactions, boss-
24
employee disputes). Using the conceptualization of influence, if one of the disputants is
also the decision maker, it is difficult to determine which role (decision maker or
disputant) is influencing the decision. Finally, in assessing procedural justice Thibaut and
Walker relied primarily on responses to questions about which procedure was preferred
and which procedure was viewed as fair, rather than on responses regarding which
aspects of procedures were seen as fair. This limited the ability to draw a more complex
definition of procedural justice beyond control.
Lind and Tyler (1988) attempted to generalize the theory of procedural justice
beyond the specific legal contexts of Thibaut and Walker (1975). The result of their
efforts was the group-value model of procedural justice, a social psychological theory of
procedural justice that emphasized the link between the procedures groups used and the
resulting values and membership implications from experiencing those procedures. Lind
and Tyler (1988, p. 40) begin their discussion of procedural justice by pointing out that
the work of Thibaut and Walker (1975) was “less bold than it should have been.” Lind
and Tyler (1988) follow the lead of Tyler and Folger (1980) in arguing that procedural
justice is important, not just as a means to fair outcomes, but as an end itself.
Furthermore, procedural justice should be seen as subjective. That is, individuals will
evaluate the fairness of procedures differently across situations and that a single “best”
procedure will not always be seen as fair by all parties involved.
Through a series of studies Lind and Tyler (1988) go from these fundamental
arguments to several new considerations regarding procedural justice. Two
considerations in particular would form the basis for the larger implications of criminal
justice studies of procedural justice. First, “procedural justice judgments lead to enhanced
25
satisfaction” (Lind & Tyler, 1988; p. 207), and second, “judgments of procedural justice
enhance the evaluation of authorities and institutions” (Lind & Tyler, 1988; p. 209). A
large portion of justice research in criminal justice has focused on satisfaction with the
police and improved attitudes towards or evaluations of police (e.g. Bradford et al., 2009;
Engel, 2005; Hough et al., 2010; Jackson & Sunshine, 2007; Reisig & Lloyd, 2009;
Sunshine & Tyler, 2003; Wolfe, 2011; Wolfe et al., 2016; Wolfe, McLean, & Pratt,
2017).
Lind and Tyler’s (1988) study also reached several other conclusions with broader
implications for social psychological theories of procedural justice. First, Lind and Tyler
recognized that giving voice was a key factor in procedural justice. This would later be
used as a key component in attempts at defining procedural justice (Tyler, 1988; Tyler,
1990). Additionally, Lind and Tyler noted that procedural justice could result in changes
to behaviors, not just attitudes and beliefs. This represents a similar shift as the work of
Adams (1965) and would be used in both the organizational behavior literature and
research into criminal justice (e.g. Lim, 2002; McLean & Wolfe, 2016; Paternoster et al.,
1997; Reisig & Bain, 2016; Reisig & Lloyd, 2009). Lind and Tyler also concluded that
procedural justice applied to more than just the procedures themselves but also how the
procedures were applied thereby incorporating the developing literature on interactional
justice (Bies, 2001; Bies & Moag, 1986; Bies & Shapiro, 1987). Finally, Lind and Tyler
concluded that procedural justice could apply to a greater range of contextual situations
beyond the courtroom previously studied by Thibaut and Walker (1975).
While these conclusions would shape Lind and Tyler’s (1988) group value model
of procedural justice and continued research on procedural justice in social psychology,
26
organizational management, and criminal justice, its most important contribution may be
in the shift in the consideration of what constitutes a fair procedure. Thibaut and Walker
(1975) had argued that control over the process was the sole predictor of procedural
justice. However, Lind and Tyler’s (1988) conclusion emphasized the need to consider
more than issues of control in measuring assessments of procedural justice including the
quality of interpersonal treatment. This left open the need for more refined measures of
procedural justice in future research. Studies of procedural justice to that point had
primarily measured procedural justice with one or two questions similar to “How fair
were the procedures…” (e.g. Kanfer et al., 1987; Lind et al., 1980; Lind & Lissak, 1985;
Tyler, 1984; Tyler & Caine, 1981). This measurement strategy is problematic for two
reasons. First, relying on a limited number of questions greatly reduces the reliability of
measurement (Carmines & Zeller, 1979). Second, similar to the concerns regarding
Thibaut and Walker (1975), this measurement strategy is unhelpful in determining what
characteristics of procedures are important in assessing procedural fairness.
Prior to Lind and Tyler’s (1988) development of their procedural justice theory,
Leventhal (1980) also noted a distinct lack of definition to the issue of procedural justice.
Leventhal noted that research on procedural justice was relatively new, and as such, there
was little information on what constituted procedural justice other than the somewhat
tautological definition of the use of a fair procedure. Leventhal (1980) then discursively
developed six rules for procedural justice: consistency, bias suppression, accuracy,
correctability, representativeness, and ethicality. In Leventhal’s view, fair procedures
would produce consistent decisions (i.e., the same decision in the same circumstances),
reduce the ability of the decision-makers’ bias to affect decisions, be accurate, have some
27
ability to be corrected if inaccurate, be representative of group values, and adhere to
personal ethics. While these rules were created without any empirical evidence, they do
represent a step forward in procedural justice research as it formed the theoretical basis
for deeper considerations of what constitutes a fair procedure. Lind and Tyler (1988)
recognized this contribution, but criticized Leventhal’s criteria as being too vague and too
broad to operationalize.
Following up on Lind and Tyler’s (1988) recognition that a more complex
definition of procedural justice was needed, Tyler (1988) conducted interviews with over
500 citizens who had recent contact with either the police or courts. In these interviews,
Tyler (1988, p. 128) found seven distinct components of procedural justice: “the
authorities’ motivation, honesty, and ethicality; the opportunities for representation; the
quality of the decisions; the opportunities for error correction; and the authorities’ bias.”
It is important to note that many of these concepts are consistent with the work of
Leventhal (1980, e.g. ethicality and bias come directly from Leventhal’s rules while
quality of the decision contains similarities to Leventhal’s accuracy rule), Thibaut and
Walker (1975; opportunities for representation is similar to control over the process), as
well as the findings of Lind and Tyler (1988; representation contains elements from Lind
and Tyler’s conclusions regarding voice and ethicality contains elements from quality of
interpersonal treatment). Tyler (1990, p. 7) would later include these criteria in Why
People Obey the Law as the characteristics of normative, procedurally-fair experiences.
In further refining the conceptualization of procedural justice, Tyler and Lind
(1992) argued that procedural justice was made up of just three components: trust,
standing, and neutrality. Trust referred to the motive-based trust of the authority, which
28
was characterized by the authority’s perceived concern for the individual’s needs and
consideration of the individual’s views. This reflects some of the same concepts as
Tyler’s (1988) earlier motivation and representation components. Standing referred to
concepts that impact an individual’s conception of his or her status in groups. This
includes aspects like dignity, politeness, and respect for rights. Thus, it draws on the
earlier concept of ethicality. Neutrality consists of honesty, fact-based decision-making,
and the absence of bias. It draws from the earlier concepts of honesty, quality of
decisions, and bias. This new conceptualization of procedural justice reduced the number
of components that make up the concept but had considerable overlap with previous
conceptualizations (e.g. Leventhal, 1980; Tyler, 1988).
Procedural justice would be refined even further by Tyler and Blader (2000) in a
precursor to their group engagement model. In their conceptualization of the concept,
procedural justice consisted of just two components: quality of decision-making
procedures and quality of treatment. The previous concept of motive-based trust was
moved to the quality of treatment component. Finally, Tyler and Huo (2002) refined the
definition of procedural justice once more. In this conceptualization procedural justice
was again composed of quality of treatment and quality of decision-making procedures,
but motive-based trust was removed from both of these components. In Tyler and Huo’s
(2002) model, motive-based trust becomes another important concept considered
separately from procedural justice that is also composed of the quality of treatment and
the quality of the decision-making process.
Conflicting conceptualizations of procedural justice could have resulted in a body
of research with conflicting results and no clear message on the impact of procedural
29
justice. However, despite the various conceptualizations of procedural justice, there is
remarkable consistency in the elements that are considered components of procedural
justice. All models include consideration of dignity, politeness, respect for rights, absence
of bias, honesty, and voice in the decision-making process. This consistency in the
underlying characteristics of a procedurally-just process has made it easier to generate a
body of research on the topic with a consensus on the importance of fair procedures in
group processes.
Interactional Justice – A Brief Note
In addition to distributive and procedural justice, researchers in social psychology
and organizational management have also studied the importance of interactional justice.
While early procedural justice focused on the impact of procedures on individuals’
evaluations of justice, Bies and colleagues noted that individuals were also concerned
with how these procedures were implemented (Bies, 2001; Bies & Moag, 1986; Bies &
Shapiro, 1987). That is, individuals also considered the quality of treatment the individual
received during the implementation of the procedure. This concern over the quality of
treatment was termed interactional justice. Initially, interactional justice was considered
to be an independent form of justice operating separately from procedural and distributive
justice (Bies, 1987; Bies & Moag, 1986). However, conceptualizations of procedural
justice have recognized the importance of interpersonal treatment. The definitions of
procedural justice discussed in the previous section almost all included the quality of
interpersonal treatment somewhere in their components of procedural justice (Tyler,
1988; Tyler & Lind, 1992; Tyler & Blader, 2000; Tyler & Huo, 2002). Bies (2001, p. 99)
highlights the argument that “people can and do distinguish the fairness of formal
30
procedures from the fairness of interpersonal treatment.” This is not, however,
inconsistent with the conceptualizations of Tyler and Blader (2000) or Tyler and Huo
(2002) who consider quality of interpersonal treatment as distinct from the quality of
decision-making procedures with both being components of procedural justice. Still, Bies
(2001) argues that including interactional justice as a separate form of justice is better for
parsimony and allows the two components (interpersonal treatment and decision-making
procedures) to impact attitudes and behaviors differently.
To a certain degree, the discussion over interactional justice as a separate form of
justice is a concern over a technicality. Both sides of the argument recognize that the
quality of interpersonal treatment is an important consideration in individuals’
evaluations of justice and fairness. Furthermore, both consider it to be, to some extent,
separate and distinct from the quality of decision-making procedures. As a result, the best
practice of researchers would be to include questions regarding both components and
utilize statistical techniques such as exploratory factor analysis and confirmatory factor
analysis to determine the degree to which the two components impact the attitudes and
behaviors of interest to the present study. Whether it is its own distinct form of justice or
a distinct component of procedural justice, interactional justice is clearly an important
consideration that must be evaluated in examinations of justice frameworks (Bies, 2001;
Lim, 2002; Skarlicki & Folger, 1997; Tyler & Blader, 2003)
Theories of Procedural Justice
In addition to differences in the conceptualization of procedural justice, the
mechanisms by which evaluations of procedural justice impact attitudes and behavior are
also not agreed upon by scholars. In particular, procedural justice theories typically link
31
procedural justice to key attitudes and behaviors using one of three mechanisms: social
identity, legitimacy, or negative emotions. The roots of both social identity and
legitimacy as mechanisms associating procedural justice to important outcomes lies in
social psychological theories. The primary social psychological theories of interest to
criminologists are the group value model (Lind & Tyler, 1988; Tyler, 1989), the group
engagement model (Tyler & Blader, 2003), and the relational model of authority (Tyler
& Lind, 1992), though other theories have connected perceptions of fairness to identity as
well (Skitka, 2003; Wenzel, 2001). On the other hand, the link between procedural
justice, negative emotions, and behavior lies primarily within criminology’s general
strain theory (Agnew, 1992).
The Group Value Model
In broadening the findings of Thibaut and Walker’s (1975) early procedural
justice theory, Lind and Tyler (1988) developed a new theory that relied on concerns
about the quality of interpersonal treatment and status within the group rather than
control over outcomes. According to this model of procedural justice, a sense of
procedural justice would be strongest whenever the procedures enacted by the group were
consistent with the shared values of the group. In other words, a group that values quick
decision making by a strong leader would consider processes consistent with this value to
be fairer than a due process procedure where procedural technicalities can influence
outcomes. On the other hand, groups that value democracy and giving power to the
common person would likely prefer procedures where authority was given to a group of
common individuals (e.g. a jury) over a single authority. The demonstration of these
shared values through procedural justice would create greater affect for group authorities
32
and leaders. Lind and Tyler (1988) then make two key predictions for the importance of
procedural justice. First, procedures will be more important than outcomes because
procedures represent group values. Second, procedural justice will be most important to
individuals with uncertain status in the group. When individuals experience fair
procedures that conform to the groups values it will reaffirm their identity and status
within the group. In the group value model, then, procedural justice is subject to the
values of the group and is affected by the certainty of the individual’s status within the
group (Lind & Tyler, 1988; Tyler, 1989).
The Group Engagement Model
While identity and status were important considerations in the group-value model,
its importance was limited to how identity and status influenced assessments of
procedural justice (Lind & Tyler, 1988; Tyler, 1989). In contrast, the group engagement
model contends that assessments of procedural justice will impact identity judgments
which will in turn impact attitudes, values, and behaviors within the group (Tyler &
Blader, 2003). Tyler and Blader (2003) use a four factor model of procedural justice
similar to the two factor model they proposed earlier (Tyler & Blader, 2000), but with
quality of treatment and quality of decision making processes split into two components
each of formal and informal. These four components – informal quality of treatment,
formal quality of treatment, informal quality of decision making, and formal quality of
decision making – form procedural justice which impacts identity judgments. Identity
judgments in turn shape psychological engagement in group norms which motivate
individuals to engage in normative behavior (Tyler & Blader, 2003).
While group values and status were an important consideration in the earlier
33
group value model, the group engagement model placed increased emphasis on social
identity. Social identity includes not only status within the group but also the pride and
respect that comes from group membership (Tyler & Blader, 2003). Thus justice or
injustice can increase or decrease the individual’s pride in the group, as well as
questioning their status in the group’s membership. This decreased identification leads to
less engagement in normative behaviors supported by the group. Studies testing the group
engagement model have generally supported the importance of social identity in
engagement in group norms, as well as its role as a mediator in the impact of procedural
justice on these behaviors (Blader & Tyler, 2009; Fuller et al., 2009). Though research
into procedural justice in criminology and criminal justice has tended to focus more on
other theories – particularly the mediating mechanism of legitimacy – studies have shown
identity judgments to be important in understanding the connection between procedural
justice and normative attitudes and behaviors (Bradford, 2014; Bradford, Murphy, &
Jackson, 2014; McLean, 2017; Murphy, 2013).
Legitimacy – The Relational Model of Authority
In their work redefining procedural justice, Lind and Tyler (1988) argued that
procedural justice had the power to improve evaluations of authorities. While this
concept is used in the group value model to demonstrate how procedures are seen as fair
when they are consistent with the values of the group and its authorities, it takes on a
bigger role in theories of procedural justice and legitimacy. In fact, it forms the
foundation for the argument that procedural justice increases the perceived legitimacy of
authorities. That is, legitimacy is considered one of the key evaluations of authority that
can be improved through the use of fair procedures.
34
Legitimacy generally refers to the recognition that an authority has a valid claim
to the power that it wields, though its specific definition and the components of
legitimacy are widely debated (Beetham, 1991; Dornbush & Scott, 1975; Hegtvedt &
Johnson, 2000; Zelditch, 2006; Zelditch et al., 1983). In one of the key works on
legitimacy Beetham (1991, p. 3) states that “where power is acquired and exercised
according to justifiable rules, and with evidence of consent, we call it rightful or
legitimate.” Thus, Beetham’s (1991) definition contains two key elements, consent and
the use of justifiable rules. Exercising power according to justifiable rules could also be
seen as the use of fair procedures. Stated another way, a key component of Beetham’s
(1991) definition of legitimacy is the use of fair procedures in exercising power. Even
outside of the realm of procedural justice theories then, the use of a fair procedure is
critical to obtaining legitimacy as an authority figure. The similarities between justice
frameworks and legitimacy were further reiterated by Hegtvedt and Johnson (2000) who
argued for the inclusion of legitimacy considerations in social psychological justice
research. One attempt at combining these two similar concepts can be seen in Tyler and
Lind’s (1992) relational model of authority.
The relational model of authority asserts that assessments of justice, both
procedural and distributive, impact an individual’s evaluation of the legitimacy of an
authority (Tyler & Lind, 1992). The perceived legitimacy of the authority then impacts
the individual’s normative attitudes and behaviors, such as acceptance of the authority’s
decisions, compliance with the authority’s rules, and satisfaction with the authority.
While both procedural and distributive justice are considered in this model, Tyler and
35
Lind (1992) make it clear that procedural justice will have larger impacts on legitimacy
than distributive justice.
This theoretical model actually arrived in the field of criminal justice prior to
Tyler and Lind’s (1992) relational model of authority or Hegtvedt and Johnson’s (2000)
call for the consolidation of justice and legitimation research. Tyler’s (1990) procedural
justice theory proposed in Why People Obey the Law is consistent with the relational
model of authority. Tyler (1990) argues that individuals consider procedural justice and
distributive justice concerns – with procedural justice being most important – in
evaluating the legitimacy of criminal justice authorities. These evaluations of legitimacy
then shape the individual’s acceptance of criminal justice decisions, compliance with the
law and the orders of criminal justice agencies, and willingness to cooperate with
criminal justice authorities. The key difference between Tyler’s (1990) criminal justice
work and the relational model of authority (Tyler & Lind, 1992) lies in the
conceptualization of procedural justice. While Tyler and Lind (1992) rely on the three
component model of procedural justice discussed earlier (trust, standing, and neutrality),
Tyler’s (1990) conceptualization is more consistent with his earlier work (Tyler, 1988).
The overlap in these theoretical models is considerable, however, as the
conceptualizations of procedural justice of Tyler (1988) and Tyler and Lind (1992) are
very similar and the intervening mechanism between procedural justice and the desired
outcomes (legitimacy) remains the same.
Criminal justice researchers have primarily relied on the intervening link of
legitimacy when conducting procedural justice research (Hough et al., 2010; Jackson et
al., 2012; Murphy, 2005; Reisig & Lloyd, 2009; Reisig et al., 2012; Tankebe, 2013; Tyler
36
& Huo, 2002; Wolfe, 2011; Wolfe et al., 2016; Wolfe et al., 2017). This is due, in part, to
the persuasive arguments of Tyler (1990) in pointing out that improving legitimacy has
other benefits beyond compliance with authorities. Specifically, Tyler and Huo (2002)
argue that in addition to increasing compliant behavior, legitimacy can help to improve
trust and confidence in legal authorities. This has been especially appealing to policing
scholars at a time when increased focus is being placed on community policing and the
relationships between police and communities (President’s Task Force on 21
st
Century
Policing, 2015).
General Strain Theory – Negative Emotions
While theories utilizing the intervening mechanisms of social identity, group
values, and legitimacy, have largely originated from social psychologists, the importance
of negative emotions as an intervening mechanism originates with criminological theory.
In particular, Agnew (1992) argues that experiencing injustice could lead to negative
emotions, which without the proper coping mechanisms could lead to criminal behavior.
Agnew’s theory is primarily focused on identifying potential strains and mechanisms for
coping with these strains. The causal link between experiences of strain and deviant
behavior is negative emotions. However, not all negative emotions are created equal.
Specifically, Agnew cites anger as the primary emotion that leads to deviant or criminal
behavior. Agnew draws from research on justice to describe how experiencing injustice is
one of the most prominent strains that can lead to anger. This argument echoes the
sentiments of Homans’ (1961) original distributive justice theory that emphasized the
emotional impact of experiences of injustice, specifically identifying anger as a likely
reaction to unfavorable conditions of injustice.
37
While a large amount of research has been conducted on general strain theory
providing it a considerable amount of empirical support (e.g. Agnew et al., 2002; Broidy,
2001; Mazerolle & Piquero, 1997; Paternoster & Mazerolle, 1994), relatively little
research has directly combined the propositions of justice theories with the propositions
of general strain theory. The few tests that have combined these two theories have shown
positive results (Barkworth & Murphy, 2015; Murphy, 2009a; Murphy & Tyler, 2008;
Scheuerman, 2013). In addition to research in criminology, the mediating role of negative
emotions has also been explored in other contexts. Specifically, researchers in
organizational justice have noted key relationships between justice and emotions, as well
as a relationship between negative emotions and negative behavioral consequences
(Chebat & Slusarczyk, 2005; Fox, Spector, & Miles, 2001; Krehbiel & Cropanzano,
2000; Weiss, Suckow, & Cropanzano, 1999).
Conclusion
The concepts of procedural justice and distributive justice have extensive histories
dating back several decades. Social psychologists, in particular, have developed detailed
and comprehensive conceptualizations and theoretical models for both concepts. Several
theories, including the group value model, the group engagement model, the relational
model of authority, and general strain theory, provide implications for criminological
research. The extent to which these concepts and theories have been tested in criminology
is discussed in the next chapter.
38
CHAPTER 3
JUSTICE RESEARCH IN CRIMINOLOGY AND CRIMINAL JUSTICE
Justice research in the field of criminology and criminal justice has primarily
focused on the role of procedural justice rather than distributive justice or interactional
justice. This is underscored by the wealth of research that calls Tyler’s (1990) social
psychological theory of legitimacy the “process-based model” (Hough et al., 2010; Reisig
& Bain, 2016; Reisig & Lloyd, 2009; Tyler & Huo, 2002; Wolfe, 2011; Wolfe &
Piquero, 2011). That is not to suggest that issues of distributive justice have not been
explored in criminal justice. Instead, these issues – such as racial disparities in
sentencing, traffic stops, or police use-of-force – have been conducted outside of the
theoretical framework established by researchers in social psychology. Still, justice
processes have been important to the study of criminal justice.
Procedural Justice
Procedural justice research in criminology has typically used legitimacy as the
theoretical link between procedural justice and outcomes of interest. While the role of
social identity and negative emotions has been explored by criminologists examining
procedural justice (e.g. Barkworth & Murphy, 2015; Bradford, 2014; Bradford et al.,
2014), legitimacy stands as the most empirically supported causal mechanism between
procedural justice and important outcomes due to its widespread use in criminological
research. The large amount of research on legitimacy is, in part, a result of its importance
within the policing context. While emotions and social identity may serve a causal link
39
between procedural justice and desired behaviors such as compliance or cooperation with
the police (Barkworth & Murphy, 2015; Bradford, 2014), legitimacy has been argued to
have an impact on other behaviors and key attitudes of interest to law enforcement
agencies, such as trust and support for the police (Hough et al., 2010; Sunshine & Tyler,
2003; Tyler & Huo, 2002).
One of the most important behaviors linked to procedural justice is criminal
offending. A few studies have examined the impact of procedural justice on offending
without examining the previously mentioned intervening factors of legitimacy, negative
emotions, or social identity (e.g. McLean & Wolfe, 2016; Paternoster et al., 1997). In a
key study conducted in the first few years after the publication of Why People Obey the
Law, Paternoster and colleagues (1997) linked improved perceptions of procedural justice
to reduced instances of domestic violence. A larger body of research has linked improved
perceptions of procedural justice to higher evaluations of legitimacy, which in turn, is
associated with less criminal offending (Murphy, 2005; Reisig, Tankebe, & Meško, 2014;
Reisig, Wolfe & Holtfreter, 2011). Overall, studies in criminology and criminal justice
have continued to demonstrate support for Tyler’s (1990) initial finding that perceptions
of procedural justice are related to criminal offending (see also Murphy, Bradford, &
Jackson, 2016; Tankebe, Reisig, & Wang, 2015; Tyler & Jackson, 2014).
Another critical outcome from improved perceptions of procedural justice and
legitimacy is cooperation with the police. Tyler and Fagan (2008), in particular, argued
that it is very difficult, if not impossible, for police to accomplish any of their goals
regarding crime without the help of the community. This argument is further supported
by social disorganization theory which places an emphasis on the role of the community
40
in helping to fight crime (Bursik, 1988; Sampson, 2012; Sampson & Groves, 1989;
Sampson, Raudenbush, & Earls, 1997; Shaw & McKay, 1942/1969). Sampson and
colleagues (1997, p. 919) highlighted the need for collective efficacy or “the willingness
of local residents to intervene for the common good” to reduce crime in neighborhoods.
Tyler’s (1990) model of legitimacy argues that individuals will be more likely to
intervene by calling the police for help or assisting the police in investigations when
perceived legitimacy is higher (see also Tyler & Fagan, 2008). Research into the impact
of procedural justice and legitimacy has strongly supported the connection between
legitimacy and cooperation with the police (Hough et al., 2010; Metcalfe et al., 2016;
Reisig et al., 2007; Reisig & Lloyd, 2009; Reisig et al., 2012; Tankebe, 2013; Tyler &
Fagan, 2008). Furthermore, this finding has been remarkably consistent across a variety
of settings with support coming from research conducted in Europe (Reisig et al., 2012),
North America (Reisig & Lloyd, 2009; Reisig et al., 2007), and even the Middle East
(Metcalfe et al., 2016).
Beyond the key behaviors of compliance with the law and cooperation with the
police, procedural justice and legitimacy have also been associated with a variety of
improved attitudes toward the criminal justice system. Tyler and Folger’s (1980) initial
study of procedural justice and policing revealed that procedural justice increased
satisfaction with the outcome of interactions with police. This result has since been
supported in a number of other studies regarding police satisfaction (Engel, 2005; Hinds
& Murphy, 2007; Murphy, 2009b; Tyler & Wakslak, 2004). Procedural justice and
legitimacy have also been shown to impact public trust, support, and confidence in the
police (Bradford et al., 2009; Hough, Jackson, & Bradford, 2014; Jackson & Bradford,
41
2009; Jackson et al., 2012a, 2012b). Similar to the need for cooperation from the public,
trust, support, and confidence in the police are critical for creating strong bonds between
community members and the police that can help fight criminal behavior in communities.
Eroding ties to the police in the form of decreased legitimacy may also increase the risk
of criminal victimization (Wolfe & McLean, 2017). As a result of the positive benefits
that can be seen from improving these attitudes towards the police, improving police-
community relationships is a key focus of the President’s Task Force on 21
st
Century
Policing (2015).
1
The vast majority of work done on procedural justice has investigated procedural
justice and legitimacy in the context of policing. However, procedural justice has, to a
lesser extent, been explored in other contexts of the criminal justice system. Early studies
of procedural justice emphasized the importance of procedural justice within courtroom
contexts (e.g. Thibaut & Walker, 1975; Lind & Tyler, 1988; Tyler, 1990). Indeed,
Thibaut and Walker’s (1975) initial study of procedural justice focused exclusively on
courtroom interactions. A relatively newer development in procedural justice research
has seen a focus on procedural justice in corrections (Beijersbergen et al., 2015;
Beijersbergen, Dirkzwager, & Niewbeerta, 2016; Reisig & Meško, 2009). Generally,
these studies of procedural justice and corrections support the argument that greater
procedural justice can decrease problems behind bars as well as after release from a
correctional institution.
In addition to being primarily focused on policing, research into procedural justice
in criminal justice has been almost exclusively observational. That is, experimental
1
Agnew (1999) also highlights the importance of community-level indicators in understanding the
importance of injustice and strain.
42
research on procedural justice and attitudes toward the police has only recently begun
(e.g. Johnson, et al., 2017; Lowrey, Maguire, & Bennett, 2016; Maguire et al., 2017; Nix,
et al., 2017). This is an important limitation of the existing research because it questions
the validity of causal claims regarding procedural justice. The observational studies
demonstrate that procedural justice and legitimacy are correlated with each other, as well
as important attitudinal and behavioral outcomes, however, it does not establish the
ability of criminal justice authorities to change evaluations of procedural justice and
legitimacy.
To date, only three randomized controlled trials have attempted to link agency
actions to perceptions of procedural justice: the Queensland Community Engagement
Trial, the Scotland Community Engagement Trial, and the Adana Randomized Controlled
Trial (MacQueen & Bradford, 2015; Mazerolle et al., 2012, 2013; Sahin et al., 2017).
Results from these randomized controlled trials have been mixed. In the Queensland
Community Engagement Trial officers conducting a randomized breath test to crack
down on drunk driving would engage in “business as usual” in the control condition and
employ a script that included key elements of procedural justice in the experimental
condition (Mazerolle et al., 2012, 2013). The results demonstrated that the script
improved procedural justice perceptions of both the specific police-citizen encounter
during the experiment and police more generally. The Scotland Community Engagement
Trial attempted to replicate the Queensland Community Engagement Trial in a different
setting (MacQueen & Bradford, 2015). Results from this study, however, did not support
the findings of the Queensland experiment with no impact on global perceptions of the
police or legitimacy. Finally, the Adana Randomized Controlled Trial utilized a script
43
infused with key elements of procedural justice during traffic stops for excessive
speeding (Sahin et al., 2017). The results of the Adana study showed that the use of the
script improved citizen’s perceptions of the specific encounter with police, but failed to
improve global perceptions of police.
Caution should be exercised before generalizing the findings of these randomized
trials. All three of these studies relied on the use of scripts to introduce procedural justice
into experimental conditions and limited the application of this script to specific traffic
stops. Procedural justice, as described by Tyler (1990), is likely to be more complex than
a script is able to achieve. Engaging in procedurally-just policing necessitates that the
officer be flexible in his or her interactions with citizens to provide them with the proper
procedure for the specific circumstances of the encounter. For example, an individual
stopped for speeding while running late to work may not want as lengthy of a procedure
as an individual stopped for speeding while out on a Sunday afternoon drive, though both
drivers will likely still want a fair procedure. Furthermore, the permanent use of a script
in police-citizen interactions is not a viable long-term method of ensuring citizens have
procedurally-just experiences.
2
As such, randomized experiments evaluating the impact
of procedural justice training given to officers may be more effective in assessing the
practicality of procedural justice as a method of improving perceptions of procedural
justice and legitimacy. This approach would have the added benefit of allowing
procedural justice to be infused into every interaction with a citizen rather than a specific
subset of traffic stops.
2
For a more detailed discussion of the problems researchers encountered in implementing these field trials
see MacQueen & Bradford, 2017.
44
As an alternative, controlled, laboratory experiments may be beneficial in
assessing the ability of procedurally-just practices to shape individuals’ assessments of
police. Within a more controlled setting, researchers will have greater ability to shape the
appropriateness of the procedural justice intervention being used. While the particular
script an officer is using may be more or less appropriate depending on the setting of the
traffic stop in the experiments mentioned previously, researchers can control all aspects
of the interaction in a laboratory setting. Researchers have previously used video clips
(e.g. Johnson et al., 2017; Lowrey et al., 2016; Maguire et al., 2017), as well as vignettes
(e.g. Barkworth & Murphy, 2015; Nix et al., 2017) to achieve greater control. The video
vignettes are typically filmed from the officer’s point of view. Participants are then
assessing vicarious procedural justice, rather than attempting to place themselves in the
vignette and assessing direct procedural justice. Additionally, these analyses have yet to
consider evaluations of distributive justice. These studies are new and confirming their
findings under different conditions is important to increasing confidence in their findings.
Distributive Justice
While the focus of justice research in criminology has been procedural justice, a
small number of studies have included distributive justice measures in their analyses
(Engel, 2005; Reisig et al., 2007; Sunshine & Tyler, 2003; Tankebe, 2013; Tyler, 1990;
Wolfe et al., 2016). Consistent with Tyler’s (1990) initial hypothesis, these studies have
shown that distributive justice has smaller effects on key outcomes (e.g. legitimacy, trust
in the police) than procedural justice. Despite the smaller effect sizes, these studies have
still shown that distributive justice is a significant predictor of these outcomes (Reisig et
al., 2007; Sunshine & Tyler, 2003; Tankebe, 2013; Tyler, 1990; Wolfe et al., 2016).
45
Presently, discussions of distributive justice in theoretical arguments about criminal
justice and legitimacy are often limited to cursory mentions without real substance.
Furthermore despite its significance a growing number of studies exploring legitimacy
and trust in the police using a justice framework have dropped distributive justice from
their empirical models altogether (Hough et al., 2010; Jackson et al., 2012; Mazerolle et
al., 2013; Murphy, 2005; Murphy & Tyler, 2008; Reisig et al., 2014; Sahin et al., 2017;
Wolfe, 2011). Though some of these studies may be ignoring distributive justice out of
practicality (e.g. re-analyzing previously collected data, attempting to shorten omnibus
questionnaires, etc.), Reisig and Lloyd (2009, p. 45) note the importance of including
distributive justice in stating, “at a minimum, however, distributive fairness should be
included in systematic assessments of behavioral cooperation as a control variable.”
Perhaps more concerning than the cursory mentions of distributive justice or its
exclusion from empirical models is the confusion regarding what distributive justice
actually means. Many discussions of distributive justice group it together with
discussions of police effectiveness, sometimes even describing it as an instrumental
concern (Hinds & Murphy, 2007; Metcalfe et al., 2016; Sunshine & Tyler, 2003; Wolfe,
2011). This confusion is undoubtedly the result of Tyler’s reference to distributive justice
as an instrumental concern in one of his key works on procedural justice, legitimacy, and
policing (Sunshine & Tyler, 2003). Tyler’s (1990) initial development of procedural
justice in criminal justice, however, accurately depicted distributive justice as a
normative concern. Instrumental concerns refer to concerns over the favorability of the
outcome, or as Tyler (1990, p. 5) described it, “concern[ed] with winning.” Distributive
justice, as discussed in the social psychological theories of Homans (1961), Adams
46
(1965), Jasso (1978), and Markovsky (1985), are not concerned with outcome
favorability, but with outcome fairness. Jasso’s (1978) justice evaluation specifically
considers how distributive justice varies when outcomes are overly favorable as
compared to overly unfavorable. An instrumental perspective, on the other hand, would
posit that individuals would not be concerned with overly favorable outcomes because
the individual would have won.
3
Confusion over the concept of distributive justice has further complicated matters
in the area of operationalization. In particular, criminal justice scholars have described
distributive justice as being concerned with the distribution of police services (e.g. Reisig
et al., 2007; Reisig & Lloyd, 2009). Although this could be a distributive justice concern,
it is problematic to measure distributive justice as being relevant to the distribution of
police services while measuring procedural justice related to police-citizen interactions.
These questions represent two different decision points as well as two different aspects of
a decision (procedural and distributive justice). By evaluating different decision points
(distribution of services rather than distribution of outcomes of police-citizen
interactions), researchers leave open the possibility that citizens are not more concerned
about procedural justice than distributive justice but rather are more concerned about the
decision regarding police-citizen interactions than the decision regarding the distribution
of services. Ideally, the justice framework is used to evaluate aspects of the same decision
or decision point. If procedural justice measures are concerned with whether or not police
officers on the street engage in fair procedures, distributive justice measures should be
concerned with whether or not police officers on the street distribute outcomes fairly. Put
3
For other discussions of the difference between the instrumental concerns of outcome favorability and the
normative concerns of outcome fairness see Engel, 2005; Tyler, 2001.
47
another way, a better measure of distributive justice in most procedural justice studies
would be to ask individuals about their perceptions of the fairness in the distribution of
outcomes such as tickets, arrests, and searches, rather than fairness in the distribution of
police services, such as frequency of patrols or increased effectiveness.
Despite the limited application and use of social psychological distributive justice
theories in criminal justice, research into distributive justice has flourished in certain
areas of criminal justice research. Specifically, research into racial bias in the criminal
justice system is consistent with the concept of distributive justice though it is conducted
outside of the theoretical frameworks of distributive justice. Claims of racial bias may be
procedural or distributive in nature. For example, a procedural claim of racial bias would
be one that asserts that the process of arrest, prosecution, or sentencing is biased against
individuals of a particular race. Distributive claims of racial bias would assert that the
distribution of outcomes across races in the criminal justice system is unfair.
Research into claims of racial bias typically focus on determining whether or not
distributive injustice exists (Alpert, Dunham, & Smith, 2007; Knowles, Perisco, & Todd,
2001; Lundman, 2004; Lundman & Kaufman, 2003; McLean & Rojek, 2016; Ridgeway,
2006; Rojek, Rosenfeld, & Decker, 2012; Tomaskovic-Devey et al., 2006). In essence,
these studies are attempting to determine the ratio inside the natural log function from
Jasso’s (1978) justice evaluation function. That is, they are attempting to determine the
ratio of actual outcomes to just outcomes. The difficulty in conducting this type of
research is in finding the just outcome that should be placed in the denominator (Fridell,
2004; McLean & Rojek, 2016; Walker, 2001). While these researchers have primarily
been concerned with establishing whether or not objective distributive justice is being
48
achieved, there has been acknowledgment of the fact that feelings of injustice could be as
important as the objective reality when considering police-community relations (Tyler &
Wakslak, 2004). In this scenario, the denominator of just outcomes may be impossible to
determine. As Homans (1961) noted in his initial work, individuals perceive inputs
differently making a universal perception of distribution difficult to achieve.
Concerns over racial bias in policing have been exacerbated recently by incidents
of unarmed African Americans being shot by police officers and an increase in negative
publicity resulting from these incidents (Nix & Wolfe, 2015, 2016; Pyrooz et al., 2016;
Rosenfeld, 2016; Wolfe & Nix, 2016). Similar to research into racial bias in other areas
of criminal justice, research into this issue has concentrated on determining whether or
not the use of force varies by race (Fryer, 2016; Nix et al., 2017). Establishing the
presence of racial bias in the use of force is, of course, an extremely important task.
However, for the purposes of police-community relations, assessing the causes and
consequences of feelings of racial bias may be as important. If feelings of racial bias are
directly linked to the actual presence of bias, the determination of racial bias and finding
its solution is critical to improving police-community relations. If, however, feelings of
racial bias are not linked to the presence of bias but some other predictor of feelings of
injustice, then other tactics may need to be employed to improve police-community
relations. In this manner, research into racial bias in policing may benefit from the
introduction of social psychology’s subjective theories of distributive justice.
Conclusion
There is a considerable body of research in criminology on the relationship
between procedural justice, legitimacy, and a variety of outcomes. However, the
49
conceptualization, operationalization, and analysis of distributive justice has been
underdeveloped. To address this concern, the next chapter lays out an adapted version of
Tyler’s (1990) theoretical model that refines the concept of distributive justice.
50
CHAPTER 4
REVISITING THE THEORETICAL MODEL
The wealth of research conducted in social psychology and criminal justice reveal
that perceptions of justice clearly play an important role in understanding individuals’
reactions to the criminal justice system. Even researchers that doubt the link between
procedural justice and legitimacy acknowledge that procedural justice is important and is
critical for policing (Nagin & Telep, 2017). As such, it is clear that procedural justice and
its related concepts will continue to play a critical role in understanding police-
community relations for years to come.
However, the current treatment of the concept of distributive justice in the field of
criminal justice is troubling. Research has consistently shown distributive justice to be an
important consideration in determining evaluations of legitimacy and trust in the police
(Reisig et al., 2007; Sunshine & Tyler, 2003; Tankebe, 2013; Tyler, 1990; Wolfe et al.,
2016). Yet, flawed conceptualizations and the exclusion of distributive justice are
common in analyses of perceptions of justice in criminology. Improved specifications of
distributive justice coupled with its inclusion in criminal justice research can create a
more complete picture of the theory of justice frameworks. Furthermore, distributive
justice theories can assist criminologists in their examinations of phenomena such as
racial bias in policing.
To address this need, I propose an adapted model of procedural justice,
distributive justice, and legitimacy incorporating guidance from the field of social
51
psychology. In constructing this model I develop a definition of distributive justice and
propose new relationships with other theoretical concepts. However, I do wish to note
that this model is not a refutation of Tyler or any other procedural justice scholars.
Rather, it is a refinement that builds on the Tylerian model that reconsiders the
importance of distributive justice.
Definition
I define distributive justice as the perceived fairness in outcomes delivered by the
police. While assessing fairness in the distribution of police services is also an outcome
judgment in line with distributive justice, it focuses on a different step in the decision-
making process from most procedural justice research. As such, I do not include it here as
a distributive justice consideration, though other studies may wish to do so if they are
focused on that point in the decision making process.
Additionally, my conceptualization of distributive justice considers it to be a
normative concept, consistent with Tyler’s (1990) original theory. That is, distributive
justice is not concerned with the favorability of outcomes or the threat of punishment.
Rather, perceiving distributive fairness results in a normative bond that encourages
individuals to have positive attitudes toward the police and to engage in compliant
behaviors. That is not to say that instrumental concerns, especially outcome favorability
(e.g. Brockner et al., 1997; Brockner & Wiesenfeld, 1996), may not play a role in
forming perceptions of distributive justice. Rather the concept itself is normative and
distinct from outcome favorability.
52
Theoretical Propositions
Teachable Moments
The practical implications for theories of procedural justice and legitimacy are
limited if officer behavior cannot impact an individuals’ perceptions of fairness and
legitimacy. In fact, Nagin and Telep (2017) have argued that procedurally just treatment
at the hands of criminal justice authorities is desirable, not because it is linked to other
evaluations, but because it is the right thing to do. From this perspective, procedural
justice should be emphasized as the ethical thing to do, but plays no role in impacting
individuals’ attitudes towards criminal justice figures. This argument rests largely on the
fact that most studies of procedural justice and legitimacy in the field of criminal justice
have been observational and cross-sectional in nature. Thus, the link between legitimacy
and procedural justice may be operating in the opposite direction of what is hypothesized.
That is, individuals’ perceptions of legitimacy may be influencing procedural justice
perceptions, rather than the other way around.
Still, the results from randomized controlled trials of procedural justice provide
considerable evidence that changes in officer behavior do influence individuals’
perceptions in a given situation (Johnson et al., 2017; Lowrey et al., 2016; Maguire et al.,
2017; Mazerolle et al., 2012, 2013; Sahin et al., 2017). This finding forms the basis of the
first theoretical proposition of this dissertation:
P
1
: Perceptions of procedural justice will be higher when police officers behave in
a manner consistent with the principles of procedural justice.
That is, officers who are polite, respectful, and listen to the citizen’s side of the story will
be viewed as procedurally just.
53
Fair Process Effect
The theories of distributive justice established by social psychologists rely on a
subjective evaluation of distributive justice, but that evaluation is directly linked to the
actual fairness of the outcome the individual received (Adams, 1965, Homans, 1961;
Jasso, 1978, Markovsky, 1985). These theories also typically rely on an available
reference to determine the fairness of outcomes. In the context of criminology and
criminal justice, such references are likely difficult as contact with the police is relatively
rare and individuals are likely to find out about contact with the police from other people
like them, rather than individuals of different races and social classes. Information on
others dissimilar to the individual would be needed to form an accurate reference. Given
the difficulty in establishing a reference, distributive justice evaluations in criminal
justice likely occur by processes other than the reference evaluations established by these
theorists (Adams, 1965; Homans, 1961; Jasso, 1978; Markovsky, 1985).
In evaluating how individuals determine outcome fairness without an available
reference, van den Bos and colleagues (1997) turned to the fair process effect. Their
research revealed that individuals make assessments regarding the distributive fairness of
their outcomes based on the fairness of procedures when references are unavailable. That
is, procedural justice shapes perceptions of distributive justice in the absence of an
available reference. Applying this finding to the field of criminal justice would suggest
that individuals make assessments regarding distributive justice based on procedural
justice. This argument forms the basis of the next theoretical proposition in this
dissertation:
54
P
2
: Higher perceptions of procedural justice will lead to higher perceptions of
distributive justice.
This proposition will be contrasted with a more traditional view of distributive justice. If
the fair process effect were not occurring, distributive justice would logically be related
to manipulations of the outcome the individual receives. That is, altering the outcome an
individual receives would be the primary method for manipulating distributive justice if
the fair process were not occurring in the field of criminal justice. As a result, the third
theoretical proposition is:
P
3
: Outcome favorability will have a smaller impact on perceptions of distributive
justice than perceptions of procedural justice.
Tyler’s Model
Adding in the fair process effect to our understanding of procedural justice and
legitimacy in criminal justice should not negate the wealth of literature on the subject that
has already been conducted. As such, the new theoretical model will also contain the
assertions made by Tyler (1990) in Why People Obey the Law and supported by the
decades of research reviewed in Chapter 3. Specifically, the theoretical model still
expects that procedural justice and distributive justice will impact individuals’
perceptions of the legitimacy of criminal justice authorities. However, the fair process
effect will be included in this analysis. Thus, procedural justice will positively impact
distributive justice, and both constructs will positively impact legitimacy. This results in a
condition of partial mediation that is specified in the fourth theoretical proposition:
P
4
: Perceptions of distributive justice will partially mediate the relationship
between procedural justice and legitimacy.
55
Justice-Restoring Responses
Revisiting the role of distributive justice also provides an opportunity to utilize
the work of social psychologists to expand criminologists’ understandings of the impact
of justice perceptions. The first promising area for expansion is the inclusion of the
concept of justice-restoring responses (Markovsky, 1985). This concept refers to specific
actions taken by an individual in an attempt to rectify situations of injustice. Markovsky
(1985) argues that greater perceptions of injustice will make justice-restoring responses
more likely. This forms the next theoretical proposition:
P
5
: Lower perceptions of procedural justice, distributive justice, and legitimacy
will increase the likelihood of behavioral responses to officer-citizen interactions
(e.g. justice-restoring responses).
Within the context of criminal justice, justice-restoring responses will be conceptualized
as including behaviors like filing a complaint with the police department regarding an
officer-citizen interaction. Thus, this theoretical model could assist in providing a
framework for understanding complaints filed against police officers and departments.
Outcome Favorability
Finally, social psychologists have explored the possibility that outcome
favorability also plays a role in relationships involving justice evaluations. Namely,
research by Brockner and colleagues has found that the relationship between justice
evaluations, such as procedural justice and distributive justice, and attitudinal and
behavioral outcomes, such as legitimacy and justice-restoring responses, is stronger when
outcome favorability is low (Brockner et al., 1997; Brockner & Wiesenfeld, 1996). In
other words, individuals’ perceptions of fairness will play a more important role in
56
predicting legitimacy and behavioral responses to officer-citizen interactions when
outcome favorability is low. When outcome favorability is high, perceptions of justice
will matter less. These findings inform the last set of theoretical propositions:
P
6
: Outcome favorability will moderate the relationship between procedural
justice, distributive justice, and legitimacy.
P
7
: Outcome favorability will moderate the relationships between procedural
justice, distributive justice, and behavioral responses to officer-citizen
interactions.
The Full Theoretical Model
The combination of all of these theoretical propositions results in the theoretical
model presented in Figure 4.1. The application of these propositions to officer-citizen
interactions is largely consistent with the conditions under which their sources intended
them to be used. However, this specific model should only be applied to interactions
between police officers and citizens. Officer-citizen interactions represent a unique
situation as the officer is likely to represent both an authority figure and an adversary.
Furthermore, there is no clear method for dealing with errors in the decision-making
process. Decisions by judges can be appealed and corrections personnel have supervisors
that are often located within the correctional facility. However, improper conduct on the
part of a police officer is likely harder to reconcile because the officers’ superiors are not
easily contacted and errors in decisions may not be corrected until after sufficient harm
has been caused. Lastly, distributional information is at its most rare in officer-citizen
interactions. Individuals may do research before appearing in court on possible outcomes
and can talk to others in a correctional facility about the outcomes they receive. With
57
police officers, though, the contact is not expected, so research on the potential outcomes
will not have occurred. Some of these principles may apply to situations involving other
criminal justice officials, but the model would likely need to be adapted first.
Figure 4.1. Theoretical Model
There is a clear consensus between the research of social psychologists (see e.g.
Clay-Warner et al., 2005; Skarlicki & Folger, 1997; van den Bos et al., 1997), the
research of criminologists including distributive justice in their analyses (e.g. Reisig et
al., 2007; Sunshine & Tyler, 2003; Tankebe, 2013; Tyler, 1990; Wolfe et al., 2016), and
the claims of activists concerned with racial bias in policing – distributive justice matters.
Refining Tyler’s model to obtain a more thorough understanding of distributive justice,
as attempted here, could help create a more complete theory, as well as assist in
understanding and addressing the “Ferguson Effect.” For example, incorporating
distributive justice as an intervening mechanism between procedural justice and
legitimacy supports the claims of Tyler’s legitimacy theory, as well as the claims of
individuals that feel racially profiled. Procedural justice still exists as a predictor of
legitimacy; however, legitimacy is also impacted by feelings of distributive justice which
are influenced by perceptions of procedural justice.
58
Furthermore, misspecifying this relationship could result in underestimating the
impact of both procedural justice and distributive justice on legitimacy. Including these
concepts as co-variates in a regression model will cause the models to compete for
explanation in their direct effects. However, treating them as a mediating relationship, if
correct, will cause distributive justice to have a larger direct effect. Additionally, while
the direct effect of procedural justice may decrease, the total effect (the direct effect plus
the indirect effect) will likely be higher than the originally estimated direct effect.
Current Study
The present study will attempt to further criminological understanding of
individuals’ justice evaluations by testing the new theoretical model of individual
perceptions of justice in police officer-citizen interactions. To do so, this study will
utilize a 2x2 factorial design to assess the impact of variations in procedural justice and
outcome favorability on perceptions of procedural justice, distributive justice, legitimacy,
and hypothetical behavioral responses to these perceptions. The factorial design will be
achieved by randomly assigning one of four vignettes to participants completing a survey
on perceptions of police officer-citizen interactions. Varying conditions of procedural
justice will allow for the examination of how police officer behavior impacts individual
perceptions of procedural fairness, as opposed to respondents’ pre-existing attitudes
regarding the police. Additionally, variations in outcome favorability will help address
whether outcome favorability or procedural justice is more influential in determining
perceptions of distributive justice. Furthermore, manipulating outcome favorability
allows for the evaluation of a potential interaction between outcome favorability and
perceptions of justice when predicting reactions to justice evaluations.
59
The theoretical propositions developed throughout this chapter have been
operationalized using this methodology to create the following hypotheses corresponding
to each proposition:
H
1
: Respondents’ perceptions of procedural justice will be higher in the
procedural justice condition than the procedural injustice condition.
H
2
: More positive perceptions of procedural fairness will be associated with more
positive perceptions of outcome fairness.
H
3
: Unfavorable outcomes will be associated with more negative perceptions of
distributive justice, but outcome favorability will have a smaller impact on
perceptions of distributive justice than perceptions of procedural fairness (H
2
).
H
4
: Perceptions of distributive justice will partially mediate the relationship
between procedural justice and legitimacy.
H
5
: Negative perceptions of procedural justice, distributive justice, and legitimacy
will make behavioral responses to officer-citizen interactions (e.g. justice-
restoring responses) more likely.
H
6
: Outcome favorability will moderate the relationships between procedural
justice, distributive justice, and legitimacy such that perceptions of justice are
more important when outcome favorability is low.
H
7
: Outcome favorability will moderate the relationships between procedural
justice, distributive justice, and behavioral responses to officer-citizen interactions
such that perceptions of justice are more important when outcome favorability is
low.
60
CHAPTER 5
METHODS
Data and Procedure
Data for the present study comes from a national convenience sample of 2,084
adults in the United States. The sample was obtained by Qualtrics Labs, Inc and had
quotas established for the sample to be nationally representative on age, race/ethnicity,
and gender. No response rate for the survey was available due to the variety of
methodologies used to recruit participants.
4
However, when recruiting for the study no
information was provided regarding the content of the study. Additionally, the sample
had a completion rate of 68.7%. This means that 68.7% of people entering the survey
provided data that was retained for the final sample. Individuals could have been
excluded if their inclusion violated a quota (n=782) or if they provided invalid data
(n=167).
5
This data shows that individuals did not drop out once they found out about the
content of the survey or once they experienced the vignette manipulations (discussed
later). Research on response rates demonstrates that response rates are problematic when
there is reason to believe that response is linked to variables of interest in the study
4
Qualtrics Labs, Inc. recruits participants through emails and banner ads. Due to its use of panel partners
and advertisement recruiting it is impossible to know how many people saw the link to complete the
survey. However, this is not uncommon for this type of study. The methodology is easily comparable to
laboratory-style experiments conducted in psychology departments where participants are recruited through
a departmental website used by undergraduate students.
5
Invalid data includes violating Qualtrics quality checks for answering questions too quickly, not agreeing
to provide best answers, or indicating that the individual was less than 18 (a violation of IRB and Qualtrics
policies).
61
(Pickett, 2017). The characteristics of this sampling procedure – specifically, the lack of
information on the content of the study prior to entering the Qualtrics website and the
lack of individuals dropping out of the study after the content was revealed – make it
unlikely that willingness to respond was linked to any variables included in the study.
Still, convenience samples are argued to have limited generalizability (Bachman
& Schutt, 2014; Peterson & Merunka, 2014). This sample, in particular, is limited to
individuals who have internet and were willing to participate in online surveys in
exchange for remuneration in the form of cash and gift cards. So, there is some question
as to whether the results apply to individuals who do not meet these criteria. Despite
these concerns there is significant reason to appreciate the value of a convenience sample
for this study.
First, this study takes place entirely in the realm of theory testing; it tests a
theoretical framework to determine if the tenets of the theory are supported. When
considering theory testing, Lucas (2003) argues that generalizability should more
appropriately be considered a characteristic of a theory, rather than a study. Unless a truly
random sample of the entire world’s population is achieved, the findings of any study are
limited to the context under which the study was conducted and any generalization
beyond this context is conjecture regardless of the sample method. A theory, however,
can be supported by a variety of studies in a variety of contexts. Thus, a theory is
generalizable to the extent that it is supported by a number of studies conducted in
different contexts, regardless of the generalizability of the individual studies. Scholars
should strive more for generalizable theory rather than a generalizable study. A
convenience sample then, could serve as an important first context to examine the theory.
62
Second, generalizability, or external validity, can only be assessed in studies with
adequate internal validity. “Internal validity is the basic minimum without which any
experiment is uninterpretable” (Campbell & Stanley, 1963, p.5). In other words, there is
little point to generalizing findings, if those findings are not even valid within the study
context (Weisburd, 2010). This study employs the random assignment of vignettes to
manipulate conditions under which procedural justice, distributive justice, and legitimacy
are evaluated. Random assignment provides for strong internal validity. Sacrifices in
external validity that are made when using a convenience sample are, therefore,
overcome by the added internal validity achieved through random assignment.
Design
Participants were asked for their demographic information, whether they had any
previous contact with police, and for their global perceptions of procedural justice and
distributive justice. After completing this portion of the survey, participants were then
asked to read a randomly assigned vignette. In each vignette, the individual is running
late for work. In an attempt to arrive at work on time, the individual exceeds the speed
limit. While speeding, the individual is pulled over by a police officer.
The vignettes then varied across conditions of high or low procedural justice, and
favorable or unfavorable outcomes. High procedural justice was achieved by having the
officer in the scenario speak politely, give voice to the citizen, and give reasons for
pulling the individual over related to public safety. Low procedural justice was achieved
by having the officer use profanity, not give voice, and give reasons for pulling the
citizen over related to personal biases rather than the law. A favorable outcome was
presented by having the officer let the individual off with a warning. An unfavorable
63
outcome was presented by having the officer write the individual a ticket (see Appendix
A for each vignette). Accordingly, this study used a 2x2 design with the following
possible randomized vignettes: high procedural justice and favorable outcome; high
procedural justice and unfavorable outcome; low procedural justice and favorable
outcome; low procedural justice and unfavorable outcome. These vignettes were
randomly assigned using Qualtrics’ random presentation function.
After reading the vignette, individuals were then asked about their perceptions of
procedural justice and distributive justice regarding the specific scenario they read, and
their likelihood of engaging in a justice-restoring response following the interaction with
the police officer. Following these questions, participants were then asked a series of
questions regarding their evaluations of police legitimacy.
Pilot Study
The vignettes and measures to be utilized in this study were piloted on 128
students at the University of South Carolina in September 2016. Participants in the pilot
study were recruited from two classes in the Department of Criminal Justice.
Participation in the study was voluntary, anonymous, and no compensation was given by
the survey administrator for participation in the study. The survey was administered to
everyone in attendance at both classes and participants were given time in class to
complete the survey. Two sections of the same course were chosen for administration so
as not to sample the same students twice. Additionally, vignettes were randomly assigned
using R’s uniform distribution random number generator. R’s random number generator
was limited to randomly generating a 1, 2, 3, or 4 corresponding to one of each of the
possible vignettes. After a list of 200 numbers was generated, vignettes were placed in
64
the order corresponding to the list. In class, students were then handed a survey and
vignette packet in order, effectively achieving random assignment. In total, responses
were received from 128 of the 147 students enrolled in the two courses for a response rate
of 87.1%. This response rate, however, does not account for any non-responses due to
absences because attendance was not taken in either class. Analyses of the pilot data are
presented in Appendix B. These analyses provided guidance for how concepts were
measured in the larger data collection and present the validity of both the measures and
the manipulations used to create the factorial design.
General Analytic Strategy
The analyses conducted in this study will fall under the umbrella of latent variable
analysis. That is, analyses will be conducted using confirmatory factor analysis and
structural equation modeling. As discussed in the pilot study analysis (see Appendix B),
these methods rely on the variance-covariance matrix to create latent variables that
minimize the impact of measurement error (Kline, 2016). Throughout the analyses, the
primary estimation method will be the robust maximum-likelihood (MLR) estimator
provided by R’s lavaan package. The maximum-likelihood estimator is the most common
and widely accepted estimator used in latent variable analyses and the robust estimator
attempts to account for small amounts of non-normality in the data (Finney & DiStefano,
2013). While the measures to be utilized in this study did not provide any indications of
univariate non-normality, the robust estimator is still appropriate given the potential for
multivariate non-normality. If multivariate non-normality is present, the estimator will be
able to adjust for it. If there is normality in the data, the estimator will reduce down to
65
provide the same estimates that would be seen using the traditional maximum likelihood
estimator.
To assess the fit for all of the models, four global fit indices will be utilized: the
Tucker-Lewis Index (TLI), the Comparative Fit Index (CFI), the standardized root mean
residual (SRMR), and the root mean squared error of approximation (RMSEA). Hu and
Bentler (1999) note that using a single-index to assess model fit is problematic because
indices are able to detect different aspects of model misfit (e.g., structural fit compared to
measurement fit). As such, it is recommended that scholars use the SRMR – due to its
unique advantages for detecting fit – and at least one other fit index. SRMR values below
.08 are generally recognized to represent good fit (Hu & Bentler, 1999). RMSEA values
above .10 indicate poor fit (Browne & Cudeck, 1993). Finally, TLI and CFI values above
0.95 indicate good fit (Hu & Bentler, 1998; 1999). The chi-square test of exact fit was not
utilized in this study because it is highly susceptible to inflation in large sample sizes.
Thus, when sample sizes are large enough for structural equation models it becomes
extremely difficult to achieve a sufficient chi-square value (Kline, 2016). Where model
fit on the four indices suggested changes were needed, modification indices were
examined to suggest the best changes.
The analysis proceeds in two stages. First, the measures for each construct are
presented. This analysis contains details regarding the measures, as well as tests for
satisfactory fit in the measurement model. Once the measures have been constructed, the
analytic strategy for the tests of the theoretical propositions are presented.
66
Measures
The measures section of this dissertation is split into three sections: vignette
manipulations, theoretical variables, and control variables. While all measures were
tested in the pilot study (see Appendix B), changes to most of the measures were made
between the pilot and the larger data collection. However, these changes were consistent
with the recommendations from the pilot study and were necessitated by the practicalities
of funding and survey length for the large data collection. Given that the measures were
guided by the results of the pilot study, exploratory factor analysis is not used on any of
the measures presented below. Rather, confirmatory factor analysis, was utilized to test
the measurement strategy. All variables were measured on a five-point scale, unless
otherwise indicated.
Manipulations
The first set of variables contains the manipulations to the vignette. As stated
previously, the vignette was manipulated by having the officer behave in a procedurally
fair or procedurally unfair manner and by having the subject receive a ticket or a warning
(see Design for a description or Appendix A for the full vignettes). Each of these
manipulations are treated as separate variables. Officer behavior is a binary variable with
1 indicating that the officer behaved in a procedurally fair manner and 0 indicating the
officer behaved in a procedurally unfair manner. Outcome is also a binary variable where
1 indicates that the subject received a ticket and 0 indicates that the subject received a
warning.
67
Theoretical Variables
The next set of theoretical variables contains measures of the concepts contained
in the hypotheses of the study. A full list of all of the items contained within this section
can be found in Appendix C. Situation specific procedural justice consisted of seven
items presented immediately following the vignette that assessed individuals’ perceptions
of the quality of treatment (e.g., “The police officer in the scenario treated me with
respect”) and the quality of the decision-making process (e.g., “The police officer in this
scenario gave me a chance to tell my side of the story before they decided what to do”).
Thus, both of Tyler & Huo’s (2002) components of procedural justice were represented
in the measure. The initial single-factor model suggested by the pilot study had a slight
problem with fit (RMSEA=0.11). Modification indices indicated that the error variances
for two items (“The police officer in the scenario: explained their decisions and actions in
a way that I understood” and “…made their decisions based on facts, rather than their
own personal opinions”) were correlated.
Correlating error variances is slightly problematic because it violates some of the
assumptions of Classical Test Theory, but it is necessary in some cases. In those
situations, it is important to consider the source of the error and not correlate errors
merely to improve model fit (Gerbing & Anderson, 1984). In this case, the source of the
error appears to be the similarities in the two items. It is logical to assume that believing
the officer made her decision based on facts could result from the officer explaining her
decisions and actions – the content of these two items. Thus, the correlated variance
seems to be a result of indicator specific commonalities, which is not problematic for the
measurement model. Additionally, specifying this correlated error variance did not
68
violate the rules for nonstandard confirmatory factor analysis models with correlated
errors developed by Kline (2016). Thus, the model was re-estimated with the error
correlation specified. With this change, the model achieved excellent fit (TLI=0.99,
CFI=0.99, RMSEA=0.07, SRMR=0.01).
Situation specific distributive justice was assessed by three items also presented
immediately following the vignette. These items asked individuals to assess the fairness
of the outcome they received in the scenario (e.g., “The police officer in this scenario
delivered a fair outcome”). Standard rules for confirmatory factor analysis require three
indicators for a single factor model, as is the case here. While meeting this rule will allow
the model to be estimated, fit measures cannot be inspected as the model is just-identified
(df=0). Thus, the specific distributive justice model was combined with the already tested
procedural justice model to generate a correlated factor model with higher degrees of
freedom. Given that the specific procedural justice model had already been tested, poor
fit in the correlated factor model would have to be caused by the fit of the specific
distributive justice factor. Fortunately, the correlated factor model had good fit
(TLI=0.97, CFI=0.97, RMSEA=0.09, SRMR=0.04).
Legitimacy was measured using six items presented after the vignette that
assessed respondents’ evaluations of the police (e.g., “You should support the decisions
of police officers even when you disagree with them”). A single-factor model, as
suggested by the pilot study, did not fit the data (TLI=0.70, CFI=0.82, RMSEA=0.27,
SRMR=0.10). The factor loadings estimated by the model clustered items into two
groups, one with standardized loadings around 0.55, and one with standardized loadings
around 0.87. A content inspection of these items revealed that one cluster focused on
69
concepts described in other studies as obligation to obey (e.g., “You should do what the
police tell you even if you do not understand or agree with the reasons”) and trust in the
police (“The police in my community care about the people in my community”; see
Hamm, Trinkner, & Carr, 2017). This suggested that there may be two correlated factors
present in the data. A model specifying this relationship was estimated and achieved good
fit (TLI=0.96, CFI=0.98, RMSEA=0.10, SRMR=0.05).
Justice-restoring responses was measured using seven items assessing an
individual’s willingness to file a complaint about the police officer in the scenario
through a variety of venues (e.g., “If the events in the scenario happened to you, how
likely would you be to file a complaint with the police department regarding the officer’s
behavior?”). This measure also did not achieve acceptable fit in the suggested single-
factor model (TLI=0.81, CFI=0.88, RMSEA=0.16, SRMR=0.05).
Modification indices indicated that several of the items had correlated error
variances. Rather than attempt to specify each of these correlations, which may have led
to a violation of Kline’s (2016) rules for such models, the correlations were examined for
possible commonalities in the items. This revealed that the items were forming two
clusters; one regarding complaints about the officer’s behavior – procedural justice-
restoring responses (see above example for this measure) and one regarding complaints
about the outcome – distributive justice-restoring responses (e.g. “If the events in the
scenario happened to you, how likely would you be to file a complaint with the police
department regarding the outcome you received?”). One item did not clearly fit into
either of these categories and was, therefore, removed from the measure. A re-estimation
of the model as a correlated factor model based on these groups had significantly
70
improved the model fit, but still not to acceptable levels. Modification indices were re-
examined and suggested that two items using the same method of complaint
(“…complain about the officer’s behavior to your friends or family?” and “…complain
about the outcome you received to your friends or family?”) had correlated error
variances. As with the specific procedural justice measure, specifying this correlated
error did not violate Kline’s (2016) rules and was justified given the similarities in the
question. Specifying this correlated error also allowed the model to achieve acceptable fit
(TLI=0.97, CFI=0.99, RMSEA=0.11, SRMR=0.01).
6
Control Variables
All control variables were assessed prior to the participant reading the vignette.
Measures in this category include global attitudes towards the police, whether the
individual had previously been stopped by the police for a traffic violation (dummy
coded, 1=previously stopped), the participant’s age, the participant’s gender (dummy
coded, 1=female), and the participant’s race/ethnicity. The latter variable was measured
by a series of dummy variables where the participant identified as Black/African
American, Hispanic/Latino, or any other race category. Participants identifying as
White/Caucasian served as the reference category for these dummy variables.
Global attitudes toward the police were assessed by two measures. The first, was
the participant’s global procedural justice evaluations. This was assessed by an
adaptation of the seven items used in the specific procedural justice measure worded to
be broader in nature (e.g., “The police treat citizens with respect”). As with the specific
6
Note that RMSEA is still not below the originally stated cutoff (RMSEA<0.11), however, it approached
this cutoff and all other measures indicated excellent fit. Given the steps taken to improve fit already, this
model was deemed acceptable.
71
procedural justice measure, the initial single-factor model did not achieve acceptable fit
(TLI=0.94, CFI=0.96, RMSEA=0.13, SRMR=0.03). Modification indices suggested that
this was the result of correlated errors between the first two procedural justice items.
These items contained similar content by asking participants about their evaluations of
the quality of interpersonal treatment by police. Thus, correlating the errors was once
again justified. A re-specified model with correlated errors was estimated and achieved
acceptable fit (TLI=0.98, CFI=0.99, RMSEA=0.07, SRMR=0.07).
The second measure of global attitudes towards the police was a measure of
participants’ global distributive justice evaluations. This measure contained five items
that asked individuals for their perceptions of fairness in the distribution of outcomes by
the police (e.g. “How often do the police deliver fair outcomes to the citizens they
interact with?”). It should be noted that the pilot analysis (see Appendix B) suggested that
this model might need some alterations. As predicted, a single-factor model had very
poor fit (TLI=0.24, CFI=0.62, RMSEA=0.43, SRMR=0.18). As with the pilot, two items
that asked about the fairness of outcomes generally (see example already provided) and
three items asking about the fairness of outcomes by group (e.g. “How often do the police
deliver different outcomes to individuals because of their race?”) appeared to group into
separate factors. Two different methods effect models where these groupings were caused
by similarities in question wordings were evaluated, but did not achieve acceptable local
fit – that is, individual items did not load on factors as hypothesized – and were rejected.
Models examining a second-order structure and a correlated factor structure, as
hypothesized in the pilot study, were also examined and achieved poor local fit. Despite
several attempted modifications to the measurement structure, satisfactory fit could not
72
be achieved by any models including all five items. Given their consistency with the later
measure of specific distributive justice, two items measuring global distributive justice
broadly (as opposed to the group-based items) were assessed for their ability to form a
satisfactory measurement model. As mentioned above, confirmatory factor analysis
requires at least three items to be fully identified, so the confirmatory factor analysis was
run as a correlated factor model with the global procedural justice factor and the two
global distributive justice items. This model achieved acceptable fit and was retained
(TLI=0.99, CFI=0.99, RMSEA=0.05, SRMR=0.02).
Final Measurement Model
As a final assessment of the measurements conducted in this dissertation a full
confirmatory factor analysis of all items was constructed. This model included each of
the latent constructs that have been discussed and included them in a correlated factor
model with no structural regressions. The model achieved acceptable fit (TLI=0.96,
CFI=0.96, RMSEA=0.05, SRMR=0.06), indicating that the measures constructed here fit
the data well enough to allow for a structural equation model to be estimated.
Structural Analytic Strategy
With the measurement model constructed, MIMIC models were run to test for
differences in latent means on global procedural justice, global distributive justice,
specific procedural justice, and specific distributive justice. MIMIC models regress latent
variables – constructed using CFA – on observed indicators, in this case the vignette
conditions. This allows for the assessment of differences in latent means across these
vignette conditions. MIMIC models assume strict measurement invariance, which is to
say that it assumes the latent variable (in this case, procedural justice and distributive
73
justice) is evaluated exactly the same across the groups identified by the observed
variables (Thompson & Green, 2013; Kline, 2016). For this reason, MIMIC is limited in
its utility to compare latent means across groups such as race or gender where
measurement may vary for the same reason the mean varies. However, the present study
uses random assignment to vary vignette conditions. As a result, there is little reason to
suspect that measurement would vary by randomly assigned condition. Thus, MIMIC is
the best option for assessing simple latent mean differences by condition.
In assessing differences in the latent means by vignette conditions, preliminary
support can be provided for a number of hypotheses and assumptions. First, the MIMIC
models for global procedural justice and global distributive justice provide a test of the
assumption of pre-test balance. That is, random assignment should have created groups
with equivalent views of the police prior to the introduction of the vignette. If this is true,
the MIMIC models for global procedural justice and global distributive justice would be
expected to show no significant effects across vignette conditions. Second, the MIMIC
models for specific procedural justice and specific distributive justice provide a
preliminary test of the fair process effect specified in Hypotheses 2 and 3, as well as the
relationship between officer behavior and perceptions specified in Hypothesis 1.
Structural regressions were run to test the theoretical propositions described in the
new model. The first SEM, labelled the situation specific SEM, examined the
relationships between outcome, officer behavior, specific procedural justice, and specific
distributive justice as shown in Figure 5.1. This tests the first, second, and third
theoretical propositions and corresponding hypotheses. If officer behavior impacts
individual perceptions, a relationship between the officer behavior manipulation and
74
specific procedural justice perceptions would be seen. If the fair process effect is
operating in officer-citizen interactions, a significant relationship between perceptions of
procedural justice and perceptions of distributive justice would be expected. At this point,
there is no need to add any control variables to the structural model, as all measures are
specific to the vignette which contains random assignment. If pre-test balance is shown in
the first MIMIC models, including control variables would be unnecessary.
Figure 5.1. Situation Specific Structural Equation Model
For comparison, however, another structural equation model – labelled the global
structural equation model – that includes the control variables listed in the Measures
section (global procedural justice, global distributive justice, previously stopped by
police, age, gender, race/ethnicity) will be estimated. The relationships observed in the
situation specific model are not expected to change, because of random assignment, but
this model will allow for the examination of the influence of global attitudes on the
hypothesized relationships. That is, the impact of pre-existing attitudes can be compared
to the impact of officer behavior on evaluations of procedural justice in specific officer-
citizen interactions. This will provide greater confidence in the findings from the
situation specific structural equation model.
The legitimacy SEM will explore the fourth hypothesis. To do so, it will take the
global structural equation model (described above) and add in the measures of
75
legitimacy; obligation to obey and trust in the police. Global justice evaluations, specific
justice evaluations, and the control variables will be used to predict the added measures.
In so doing, a mediating relationship will be specified between specific procedural
justice, specific distributive justice, and legitimacy, consistent with the fourth hypothesis.
This model will then estimate the direct and total effects of these theoretical variables to
assess the degree to which the effect of procedural justice on the legitimacy variables is
mediated by distributive justice. The model, excluding control variables, is depicted in
Figure 5.2.
Figure 5.2. Legitimacy Structural Equation Model
The next model, the justice-restoring responses SEM, further builds upon the
legitimacy model by incorporating behavioral responses to conditions of injustice. This
model evaluates the fifth hypothesis, that feelings of injustice and low legitimacy will
increase justice-restoring responses. As with the previous model, this model simply adds
a new endogenous variable to the previously constructed equation. The hypothesis is then
confirmed or rejected based on the regression coefficients between perceptions of justice,
the legitimacy variables, and justice-restoring responses. The model, excluding control
variables, is depicted in Figure 5.3. It should be noted, however, that due to the
measurement issues discussed above, justice-restoring responses will be measured by two
correlated factors, rather than one factor. For ease of interpretability, Figure 5.3 depicts
just one of the factors.
76
Figure 5.3. Justice-Restoring Responses Structural Equation Model
Interaction Effects
The final set of analyses assess potential interaction effects between outcome
favorability, perceptions of fairness, perceived legitimacy, and justice-restoring responses
consistent with hypotheses six and seven. To conduct this analysis the data will be split
into two samples: one subsample of individuals who received a warning and one
subsample of individuals who received a ticket. Structural equation models three and four
(see Figures 5.2 and 5.3) will then be replicated on each subsample. As suggested by
Paternoster and colleagues (1998; see also Brame et al., 1998), the Clogg and colleagues
(1995) formula will be used to determine if significant differences in the regression
coefficients between samples exist.
77
CHAPTER 6
RESULTS
The first set of analyses are MIMIC models examining pre-test balance on global
procedural justice and global distributive justice. Keeping in mind that these items were
assessed prior to respondents reading the vignettes, these analyses serve to provide
evidence that sufficient random assignment was achieved to be confident in the non-
spuriousness of the relationships between the vignette condition and situation specific
evaluations of procedural justice and distributive justice. The results from these MIMIC
models are presented in Table 6.1.
For both pre-test variables and both vignette conditions, all coefficients are not
significantly different from zero. Thus, there are no significant differences in global
perceptions of procedural or distributive justice across vignette conditions and greater
confidence can be placed in the assumption of pre-test balance.
Table 6.1. MIMIC Models Testing Pre-Test Balance
Completely
Std. Coeff.
Latent Std.
Coeff.
Standard
Error
Global Procedural Justice
Officer Behavior 0.02 0.04 0.05
Outcome -0.01 -0.01 0.05
Global Distributive Justice
Officer Behavior 0.02 0.05 0.05
Outcome 0.02 0.04 0.05
TLI=0.99, CFI=0.99, RMSEA=0.04, SRMR=0.02; *p<0.05, **p<0.01
With pre-test balance established, Table 6.2 examines the coefficients for another
set of MIMIC models looking at differences in situation specific procedural and
78
distributive justice evaluations across vignette conditions. Three coefficients in this
analysis are significantly different from zero. When examining a model with a
dichotomous variable, it is typically more informative to examine the latent standardized
coefficient. This coefficient is standardized on the latent variable (e.g. situation specific
procedural justice) but not the dichotomous variable (e.g. officer behavior).
In this model, the officer behavior condition is associated with situation specific
perceptions of procedural justice. The latent standardized coefficient can be interpreted to
mean that in the procedurally fair officer behavior condition, subjects had situation
specific procedural justice perceptions 1.57 standard deviations higher than subjects in
the procedurally unfair officer behavior condition. Similarly, subjects in the procedurally
fair officer behavior condition had situation specific perceptions of distributive justice
0.99 standard deviations higher than subjects in the procedurally unfair officer behavior
condition. Finally, the coefficient for changes in situation specific procedural justice
perceptions by the outcome condition is significant, but is substantively small (β=-0.07).
As such, caution should be utilized in drawing too many inferences from this
relationship. Still, it is worth noting that this relationship is such that when the subject
received a ticket in the vignette, respondents rated the procedural fairness of the officer to
be slightly lower.
Table 6.2. MIMIC Model Testing Vignette Manipulations
Completely
Std. Coeff.
Latent Std.
Coeff.
Standard
Error
Specific Procedural Justice
Officer Behavior 0.78** 1.57 0.09
Outcome -0.07** -0.14 0.05
Specific Distributive Justice
Officer Behavior 0.50** 0.99 0.05
Outcome -0.01 -0.02 0.05
TLI=0.96, CFI=0.97, RMSEA=0.09, SRMR=0.04; *p<0.05, **p<0.01
79
Since there is confidence in the pre-test balance assumption, the differences
shown in these coefficients are assumed to be due to the vignette rather than pre-existing
attitudes towards the police. As such, this test provides preliminary evidence in support
of hypotheses one, two, and three. First, the coefficient between officer behavior and
situation specific procedural justice evaluations is consistent with the proposition that
perceptions of procedural justice will be higher when police officers behave in a manner
consistent with the principles of procedural justice. Second, the coefficient between
officer behavior and situation specific procedural justice evaluations is consistent with
the second proposition that is based on the fair process effect. That is, it shows that
distributive justice evaluations likely are partially based on perceptions of officer
behavior. The coefficient between outcome and distributive justice was not significant,
supporting the third proposition that outcome favorability will play a smaller role than
perceptions of procedural justice in forming distributive justice evaluations.
Additional evidence regarding these propositions was found in the next step of the
analysis, the situation specific structural equation model (Figure 6.1). This model is a
simple path model of the same variables included in the previous MIMIC model (Table
6.2). The previous MIMIC model only established a link between perceptions of
distributive justice and the officer behavior condition, not perceptions of the officer’s
behavior. This structural model on the other hand, has the advantage of specifying a path
between situation specific perceptions of procedural justice and situation specific
perceptions of distributive justice to directly test the second theoretical proposition. As
these variables are all still specific to the vignette and pre-test balance was established,
there is no need to include control variables in this model.
80
Consistent with the MIMIC models, the structural model in Figure 6.1 supports
the theoretical propositions. Officer behavior is strongly and significantly associated with
situation specific perceptions of procedural justice (β=0.78). With regards to the fair
process effect, situation specific perceptions of procedural justice are associated with
situation specific perceptions of distributive justice (β=0.70). Finally, outcome is
significantly associated with situation specific perceptions of distributive justice, though
this relationship is small and in the opposite direction of what would be expected
(β=0.04). That is, situation specific distributive justice perceptions are higher when the
subject was in the ticket condition, as compared to the warning condition. Again,
however, given the small substantive size of the relationship, caution should be taken in
drawing inferences from this result.
TLI=0.97, CFI=0.96, RMSEA=0.09, SRMR=0.05; *p<0.05, **p<0.01
Figure 6.1. Situation Specific SEM
It is important to note that several alternative models to the one presented in
Figure 6.1 are possible. That is, the paths presented in Figure 6.1 were specified by the
author and the model could be re-estimated with different paths. In fact, the author did
81
estimate several alternative models.
7
Each of these models had similar fit statistics,
providing no clear indication for which model should be chosen. Thus, the author chose
the model presented in Figure 6.1 because it was most consistent with the theoretical
model presented in this manuscript. While the fact remains that alternative models are
just as plausible as the model presented here, each of these alternative models are
generally consistent with the theoretical propositions. That is, each model showed
significant relationships between officer behavior and procedural justice perceptions,
officer behavior or procedural justice perceptions and distributive justice perceptions, and
smaller or no relationship between the outcome condition and distributive justice
perceptions. As such, despite indeterminacy in the appropriate model, confidence can still
be placed in the argument that the structural model supports the theoretical propositions.
Before moving on to examine legitimacy, a second structural equation model was
examined that incorporated the control variables that would be included in the legitimacy
analyses (Table 6.3).
8
This model also contained a path between officer behavior and
situation specific distributive justice evaluations, as well as a path between outcome and
situation specific procedural justice evaluations. These paths were not included in the
model presented in Figure 6.1 but were considered in the alternative models. They are
presented here and included in future models in an effort to include the most
comprehensive models possible.
7
For example, one alternative model contained all of the paths depicted in Figure 6.1, as well as direct
paths from outcome to procedural justice and officer behavior to distributive justice.
8
While structural equation models are typically depicted in graphical form, the models presented from this
point forward are presented as tables. They contain the same data as the graphical depictions, but achieve
greater clarity due to the large number of paths being specified.
82
As expected, the inclusion of covariates and additional paths did not significantly
change the relationships observed in the situation specific structural model. In particular,
the coefficient for the path between officer behavior and situation specific perceptions of
procedural justice is exactly the same, and the coefficient for the path between outcome
and situation specific perceptions of distributive justice is nearly the same. The
relationship between situation specific procedural justice and situation specific
distributive justice is again strong and significant (β=0.80). The coefficient for the direct
path between officer behavior and situation specific distributive justice is also significant
and negative. This is the opposite of what is expected as it means that individuals had
lower perceptions of distributive justice when the officer behaved in a procedurally fair
manner. However, this ignores the fact that there is an indirect path between officer
behavior and situation specific distributive justice that operates through situation specific
procedural justice. The total effect that includes both this indirect path as well as the
direct path, is estimated to be β=0.49 (p<0.01). Thus, the true relationship between
officer behavior and situation specific distributive justice is positive and significant.
9
Outside of the relationships already examined, it is also worth noting that the
relationships between global perceptions and situation specific perceptions were also
significant. That is, global perceptions of procedural justice are associated with situation
specific perceptions of procedural justice (β=0.13) and global perceptions of distributive
justice are associated with situation specific perceptions of distributive justice (β=0.09).
These relationships are expected as global attitudes likely influence the way an individual
9
It is difficult to interpret the meaning of the officer behavior/situation distributive justice direct
relationship. It would mean that officer behavior is associated with lower evaluations of distributive justice
holding evaluations of officer behavior (situation procedural justice) constant.
83
perceives a specific situation. Several demographic variables also have significant, but
weak relationships with the situation specific perceptions.
Table 6.3. Global SEM
Dependent Variable: Procedural Justice Distributive Justice
Std. Coeff. Std. Error Std. Coeff. Std. Error
Situation Procedural Justice -- -- 0.80** 0.04
Officer Behavior 0.78** 0.09 -0.13** 0.08
Outcome -0.07** 0.05 0.05* 0.05
Global Procedural Justice 0.13** 0.05 -0.01 0.05
Global Distributive Justice 0.08* 0.05 0.09* 0.05
Age -0.06** 0.01 0.06** 0.01
Female -0.04** 0.05 0.05* 0.05
Black 0.03 0.09 -0.03 0.08
Hispanic -0.01 0.07 -0.04 0.07
Other -0.01 0.09 -0.04* 0.09
Stopped by the Police 0.01 0.05 0.07** 0.05
TLI=0.96, CFI=0.97, RMSEA=0.05, SRMR=0.05; *p<0.05, **p<0.01
Legitimacy and Justice-Restoring Responses
The analyses to this point have established support for the first three theoretical
propositions, all of which deal with how officer behavior and outcome impact situation
specific evaluations. The next set of analyses attempts to connect these concepts with the
concept of legitimacy using the obligation to obey and trust in the police measures, and
justice-restoring responses using the procedural and distributive justice-restoring
response measures. In estimating the structural models, all of the paths from the previous
structural equation model are included. However, since these paths are unchanged from
the previous models they are not presented again because the coefficients are identical.
Still, they are used to estimate the total effects (direct effect plus indirect effect) for
certain relationships so their presence should be kept in mind.
Table 6.4 presents the first of these analyses with obligation to obey and trust in
the police as the dependent variables of interest. Situation distributive justice has a
84
significant relationship with both legitimacy variables. Situation procedural justice does
not have a significant direct relationship with either obligation to obey or trust in the
police, but since it is closely related to situation distributive justice (see Table 6.3), it
does have a significant total effect for obligation to obey (β=0.25, p<0.01) and trust in the
police (β=0.19, p<0.01). Thus, situation procedural justice is related to legitimacy
through the intervening mechanism of situation distributive justice, consistent with the
fourth theoretical proposition.
Table 6.4. Legitimacy SEM
Dependent Variable: Obligation to Obey Trust in the Police
Std. Coeff. Std. Error Std. Coeff. Std. Error
Situation Procedural Justice 0.02 0.04 0.06 0.04
Situation Distributive Justice 0.29** 0.04 0.17** 0.03
Officer Behavior -0.07 0.09 -0.09* 0.10
Outcome -0.01 0.05 0.01 0.05
Global Procedural Justice 0.30** 0.06 0.54** 0.07
Global Distributive Justice 0.16** 0.05 0.15** 0.06
Age 0.14** 0.01 0.11** 0.01
Female -0.08** 0.06 0.05* 0.06
Black -0.01 0.10 -0.08** 0.10
Hispanic 0.05 0.08 -0.03 0.07
Other -0.02 0.09 -0.06** 0.10
Stopped by the Police 0.03 0.05 -0.03 0.05
TLI=0.96, CFI=0.96, RMSEA=0.05, SRMR=0.06; *p<0.05, **p<0.01
In addition to these situation specific relationships, global procedural justice and
global distributive justice also have significant relationships with the legitimacy
variables. In fact, global procedural justice has the strongest standardized coefficient of
any of the variables included in the model for both obligation to obey and trust in the
police. As noted previously, the majority of the literature reviewed in this manuscript has
been cross-sectional in nature. Thus, the measures of procedural and distributive justice
were predominantly global measures. These results are consistent with Tyler’s model and
prior research, in that global procedural justice is the strongest predictor of legitimacy.
85
The last coefficients of note are the relationships between officer behavior and the
legitimacy variables. As with procedural justice, these coefficients are slightly misleading
as officer behavior has a number of indirect paths to the legitimacy variables. In fact, the
total effects of officer behavior on obligation to obey (β=0.09, p<0.01) and trust in the
police (β=0.04, p<0.05) are significant (Table 6.5). Further, the direction of these
coefficients suggests that individuals have higher legitimacy evaluations when an officer
behaves in a manner consistent with the principles of procedural justice.
Table 6.5. Legitimacy Total Effects
Dependent Variable: Obligation to Obey Trust in the Police
Std. Coeff. Std. Error Std. Coeff. Std. Error
Situation Procedural Justice 0.25** 0.03 0.19** 0.04
Officer Behavior 0.09** 0.05 0.04* 0.05
Outcome -0.01 0.05 0.01 0.05
TLI=0.96, CFI=0.96, RMSEA=0.05, SRMR=0.06; *p<0.05, **p<0.01
Table 6.6. Justice-Restoring Responses SEM
Dependent Variable: Procedural JRR Distributive JRR
Std. Coeff. Std. Error Std. Coeff. Std. Error
Situation Procedural Justice -0.23** 0.04 0.17** 0.04
Situation Distributive Justice -0.26** 0.03 -0.47** 0.03
Officer Behavior -0.21** 0.10 -0.24** 0.11
Outcome 0.10** 0.05 0.12** 0.05
Global Procedural Justice 0.09* 0.06 0.11* 0.06
Global Distributive Justice 0.06 0.05 0.06 0.05
Obligation to Obey 0.01 0.03 -0.01 0.03
Trust in the Police 0.07* 0.03 0.03 0.03
Age -0.17** 0.01 -0.20** 0.01
Female -0.06** 0.05 -0.08** 0.05
Black 0.09** 0.09 0.10** 0.09
Hispanic 0.07** 0.07 0.07** 0.08
Other 0.04 0.09 0.04* 0.09
Stopped by the Police -0.03 0.05 -0.04* 0.05
TLI=0.95, CFI=0.95, RMSEA=0.05, SRMR=0.07; *p<0.05, **p<0.01
Table 6.6 presents the results of the analysis expanded to include justice-restoring
responses. At this point, however, there is a large number of mediating relationships
complicating the direct effects presented here. The most easily interpreted relationships
86
are those between legitimacy and justice-restoring responses. Of these coefficients, only
the path between trust in the police and procedural justice-restoring responses is
significant and is in the opposite direction of what would be expected. That is, individuals
who have greater trust in the police are more likely to engage in a procedural justice-
restoring response.
The total effects of situation procedural justice (β=-0.43, p<0.01) and situation
distributive justice (β=-0.24, p<0.01) are significant and in the expected direction (Table
6.7). That is, when subjects viewed the police in the scenario more favorably they were
less likely to indicate that they would engage in a procedural justice-restoring response.
Additionally, situation procedural justice is more strongly associated with procedural
justice-restoring responses than situation distributive justice, as would be expected given
their similar areas of concern. The total effect of officer behavior on procedural justice-
restoring responses is also significant (β=-0.51, p<0.01). When an officer behaves in a
manner consistent with the principles of procedural justice, individuals are less likely to
engage in a procedural justice-restoring response. Finally, there was a relationship
between outcome and procedural justice-restoring responses such that when an individual
received a ticket, he or she was more likely to engage in a justice-restoring response
(β=0.11, p<0.01).
Table 6.7. Justice-Restoring Responses Total Effects
Dependent Variable: Procedural JRR Distributive JRR
Std. Coeff. Std. Error Std. Coeff. Std. Error
Situation Procedural Justice -0.43** 0.03 -0.21** 0.03
Situation Distributive Justice -0.24** 0.03 -0.46** 0.03
Officer Behavior -0.51** 0.07 -0.34** 0.06
Outcome 0.11** 0.05 0.11** 0.05
TLI=0.95, CFI=0.95, RMSEA=0.05, SRMR=0.07; *p<0.05, **p<0.01
87
The distributive justice-restoring response coefficients are similarly complicated
by the large number of indirect effects not included in the table. The total effect for
situation procedural justice is again significant (β=-0.21, p<0.01). Individuals who view
the police as having acted in a procedurally fair manner are less likely to engage in a
distributive justice-restoring response. The total effect for situation distributive justice is
also significant (β=-0.46, p<0.01) and in a similar direction. Consistent with their similar
content areas, the total effect of situation distributive justice on distributive justice-
restoring responses is stronger than the total effect of situation procedural justice on
distributive justice-restoring responses. The total effect of officer behavior on distributive
justice-restoring responses suggests that when officers behave in a manner consistent
with the principles of procedural justice, individuals are less likely to engage in
distributive justice-restoring responses (β=-0.34, p<0.01). Additionally, as with the
procedural justice-restoring responses, there was a significant relationship with outcome
such that when an individual received a ticket, he or she was more likely to engage in a
justice-restoring response (β=0.11, p<0.01).
Interaction Effects
The final set of analyses examined potential interaction effects due to individuals
receiving different outcomes in the vignette. To conduct this analysis separate models
were run on individuals that received a warning in the vignette and individuals that
received a ticket in the vignette. The differences in the coefficients produced by these
models were then assessed for significance using the Clogg and colleagues (1995) test
recommended by Paternoster and colleagues (1998) and Brame and colleagues (1998).
88
The first analysis in this set is presented in Table 6.8 and examines interaction effects for
the obligation to obey dependent variable.
Table 6.8. Obligation to Obey Interaction Effects
Warning Ticket Diff
Std. Coeff. Std. Error
Std. Coeff. Std. Error Z-score
Direct Effects:
Situation
Procedural Justice 0.11 0.05 -0.07 0.06 1.73
Situation
Distributive Justice 0.22** 0.05 0.33**
0.05 -1.50
Officer Behavior -0.14** 0.11 0.01 0.15 -2.03*
Global Procedural
Justice 0.30** 0.08 0.30**
0.08 -0.07
Global Distributive
Justice 0.14* 0.07 0.16**
0.08 -0.31
Age 0.14** 0.01 0.14**
0.01 0.00
Female -0.10** 0.08 -0.05 0.08 -1.16
Black -0.01 0.13 -0.01 0.15 0.02
Hispanic 0.04 0.11 0.05 0.11 -0.13
Other -0.02 0.12 -0.03 0.15 0.32
Stopped by the
Police 0.06* 0.07 0.01 0.08 1.38
Total Effects:
Situation
Procedural Justice 0.29** 0.04 0.19**
0.05 1.21
Officer Behavior 0.05** 0.07 0.12**
0.08 -1.64
*p<0.05, **p<0.01; Warning only model fit: TLI=0.96, CFI=0.97, RMSEA=0.05,
SRMR=0.06 ; Ticket only model fit: TLI=0.96, CFI=0.96, RMSEA=0.05, SRMR=0.06
The only coefficients that are significantly different are the coefficients for the
direct path between officer behavior and obligation to obey. This effect is a small part of
a larger total effect between officer behavior and obligation to obey that includes indirect
paths through situation procedural justice and situation distributive justice. This total
effect is not significantly different between the warning and ticket subsamples (warning
β=0.11, p<0.01; ticket β=0.16, p<0.01; Clogg=-0.18). Thus, in both samples, when an
officer behaves in a manner consistent with the principles of procedural justice, the
89
individual reports a stronger obligation to obey the police. This relationship is invariant
between outcome conditions.
Table 6.9. Trust in Police Interaction Effects
Warning Ticket Diff
Std. Coeff. Std.
Error
Std. Coeff. Std. Error Z-score
Direct Effects:
Situation Procedural
Justice
0.14* 0.05 -0.02 0.06 1.78
Situation Distributive
Justice
0.11* 0.05 0.21**
0.05 -1.59
Officer Behavior -0.14** 0.13 -0.03 0.16 -1.67
Global Procedural
Justice
0.52** 0.10 0.56**
0.10 -0.34
Global Distributive
Justice
0.18** 0.08 0.13* 0.08 0.77
Age 0.12** 0.01 0.09**
0.01 0.55
Female 0.05* 0.08 0.04 0.08 0.18
Black -0.05 0.14 -0.10**
0.15 0.95
Hispanic -0.05* 0.10 -0.01 0.11 -1.28
Other -0.07** 0.13 -0.05 0.15 -0.48
Stopped by the Police -0.04 0.07 -0.02 0.08 -0.57
Total Effects:
Situation Procedural
Justice
0.23** 0.05 0.14* 0.05 1.27
Officer Behavior 0.02 0.07 0.06* 0.07 -1.08
*p<0.05, **p<0.01; Warning only model fit: TLI=0.96, CFI=0.97, RMSEA=0.05,
SRMR=0.06 ; Ticket only model fit: TLI=0.96, CFI=0.96, RMSEA=0.05, SRMR=0.06
Table 6.9 presents the same comparisons with the trust in police measure as the
dependent variable. None of the coefficients for either the direct or indirect paths are
significantly different across outcome conditions. Thus, the relationships observed in the
analysis presented in Table 6.4 are not affected by the outcome condition. That is, no
interaction effects are present between these predictors and the outcome received.
The results for the models including justice-restoring responses present a similar
story. Table 6.10 shows only one significant interaction for procedural justice-restoring
90
Table 6.10. Procedural Justice-Restoring Responses Interaction Effects
Warning Ticket Diff
Std. Coeff. Std.
Error
Std.
Coeff.
Std. Error Z-score
Direct Effects:
Situation Procedural
Justice
-0.16* 0.06 -0.26**
0.05 1.08
Situation Distributive
Justice
-0.29** 0.05 -0.27**
0.04 0.06
Officer Behavior -0.20** 0.15 -0.24**
0.15 0.68
Global Procedural
Justice
0.07 0.08 0.08 0.09 -0.18
Global Distributive
Justice
0.05 0.07 0.07 0.07 -0.29
Obligation to Obey -0.05 0.04 0.08* 0.04 -2.40*
Trust in the Police 0.07 0.04 0.07 0.05 -0.02
Age -0.18** 0.01 -0.17**
0.01 0.35
Female -0.04 0.07 -0.08**
0.07 1.28
Black 0.11** 0.12 0.07* 0.13 0.73
Hispanic 0.06* 0.11 0.08* 0.10 -0.45
Other 0.06* 0.13 0.01 0.14 1.32
Stopped by the Police -0.05 0.07 -0.01 0.07 -0.85
Total Effects:
*
Situation Procedural
Justice
-0.40** 0.05 -0.44**
0.04 0.62
Situation Distributive
Justice
-0.29** 0.04 -0.23**
0.04 -0.65
Officer Behavior -0.47** 0.10 -0.56**
0.11 2.54**
*p<0.05, **p<0.01; Warning only model fit: TLI=0.95, CFI=0.96, RMSEA=0.05,
SRMR=0.07 ; Ticket only model fit: TLI=0.94, CFI=0.95, RMSEA=0.05, SRMR=0.07
responses. The effect of obligation to obey is weak and positive in the ticket vignette
condition, but insignificant in the warning vignette condition. Thus, in the ticket vignette
condition higher obligation to obey evaluations are associated with an increased
likelihood to engage in a procedural justice-restoring response. One total effect, the
combined direct and indirect effect estimate, had significant differences across outcome
conditions. This was the officer behavior vignette condition (warning β=-0.47, p<0.01;
ticket β=-0.56, p<0.01); Clogg=2.54). In both conditions, the officer behaving in a
91
procedurally fair manner decreased the likelihood that an individual would engage in a
procedural justice-restoring response. However, the relationship was stronger in the ticket
condition. Thus, when an individual receives a ticket, the manner in which the officer
behaves has a stronger impact on the likelihood of an individual engaging in a procedural
justice-restoring response.
Table 6.11. Distributive Justice-Restoring Responses Interaction Effects
Warning Ticket Diff
Std. Coeff. Std.
Error
Std.
Coeff.
Std. Error Z-score
Direct Effects:
Situation Procedural
Justice
0.22** 0.05 0.15* 0.05 0.73
Situation Distributive
Justice
-0.45** 0.05 -0.53**
0.05 1.32
Officer Behavior -0.26** 0.15 -0.24**
0.15 -0.08
Global Procedural
Justice
0.12 0.08 0.10 0.09 0.13
Global Distributive
Justice
0.01 0.07 0.11 0.07 -1.23
Obligation to Obey -0.10* 0.04 0.09* 0.04 -3.01**
Trust in the Police 0.04 0.04 0.01 0.05 0.43
Age -0.19** 0.01 -0.21**
0.01 0.83
Female -0.09** 0.07 -0.07* 0.07 -0.40
Black 0.12** 0.12 0.06 0.14 1.16
Hispanic 0.09** 0.11 0.06 0.11 0.51
Other 0.06* 0.13 0.03 0.13 0.75
Stopped by the Police -0.07* 0.07 -0.02 0.07 -1.34
Total Effects:
Situation Procedural
Justice
-0.17** 0.05 -0.24**
0.05 0.73
Situation Distributive
Justice
-0.47** 0.05 -0.49**
0.05 0.72
Officer Behavior -0.31** 0.07 -0.37**
0.08 1.62
*p<0.05, **p<0.01; Warning only model fit: TLI=0.95, CFI=0.96, RMSEA=0.05,
SRMR=0.07 ; Ticket only model fit: TLI=0.94, CFI=0.95, RMSEA=0.05, SRMR=0.07
Finally, Table 6.11 presents the comparison in coefficients for the distributive
justice-restoring response dependent variable. As with the procedural justice-restoring
92
response coefficients, the only significant difference concerns the relationship between
obligation to obey and distributive justice-restoring responses. In the warning outcome
condition, a greater perceived obligation to obey has a small and negative relationship
with distributive justice-restoring responses. In the ticket outcome condition, a greater
perceived obligation to obey has a small and positive relationship with distributive
justice-restoring responses. None of the combined direct and indirect effect estimates
were significantly different across outcome conditions.
93
CHAPTER 7
DISCUSSION AND CONCLUSIONS
This dissertation developed and tested seven theoretical propositions for
integrating social psychological research on distributive justice into criminal justice
research on procedural justice and legitimacy. Those seven propositions are reviewed
here to discuss their status following empirical testing.
P
1
: Perceptions of procedural justice will be higher when police officers behave in
a manner consistent with the principles of procedural justice.
Nagin and Telep (2017) noted that the first step in confirming the validity of
Tyler’s (1990) legitimacy theory was to verify that perceptions of procedural justice are
shaped by officer behavior and not just other contextual factors, such as pre-existing
attitudes towards the police. This dissertation took a similar first step in establishing that
differences in the way an officer behaved in a vignette were associated with differences
in procedural justice perceptions. Several analyses were run relevant to this proposition
including a MIMIC model and two structural equation models. All three models
supported the proposition. Thus, Nagin and Telep’s concern can be refuted here;
differences in perceptions of procedural justice were seen across officer behavior
conditions.
P
2
: Higher perceptions of procedural justice will lead to higher perceptions of
distributive justice.
94
Consistent with the fair process effect put forth by van den Bos and colleagues
(1997), this dissertation proposed that individuals would use procedural cues to form
their perceptions of distributive justice. Again, a number of models (MIMIC and
structural) were relevant to this proposition and all supported it. Regardless of model
specification, individuals used their perceptions of officer behavior to help formulate
their perceptions of the fairness of the outcome the officer delivered. Thus, there is reason
to believe that the fair process effect does operate in a criminal justice context.
Participants in this study formed judgments of outcome fairness based on their
perceptions of the officer’s behavior during the traffic stop in the vignette.
P
3
: Outcome favorability will have a smaller impact on perceptions of distributive
justice than perceptions of procedural justice.
In contrast to the fair process effect, the possibility that distributive justice
perceptions were based on outcome favorability was also considered. Consistent with this
theoretical proposition, however, the models analyzed here all demonstrated that
perceptions of distributive justice were more closely linked to officer behavior and
procedural justice than the outcome delivered. In fact, outcome played little no role in the
formation of distributive justice perceptions in this study. However, the vignettes
presented here contained low-risk outcomes. That is, individuals drove away from the
traffic stop regardless of the vignette condition. Other interactions may not be so benign.
Officer-citizen interactions can end in citizens having force used against them or being
incarcerated. When the outcome has more serious consequences, its role in shaping
distributive justice perceptions may be increased. Still, these analyses provide sufficient
evidence to believe that the fair process effect is present in officer-citizen interactions and
95
procedural justice perceptions will be an important predictor of distributive justice
evaluations.
P
4
: Perceptions of distributive justice will partially mediate the relationship
between procedural justice and legitimacy.
The fourth theoretical proposition integrated Tyler’s (1990) theory of legitimacy
by linking perceptions of distributive justice and procedural justice to legitimacy
evaluations. The legitimacy structural model tested this proposition and demonstrated
that the effect of situation specific procedural justice perceptions on legitimacy
evaluations were mediated by situation specific distributive justice perceptions. That is,
situation specific perceptions of procedural justice shape situation specific perceptions of
distributive justice, which in turn, impact legitimacy evaluations. This supports Tyler’s
argument by confirming that procedural justice is an important predictor of legitimacy
evaluations. However, this is due to its importance in shaping distributive justice
evaluations. Thus, the finding does not refute Tyler’s argument, but provides context for
why his propositions are accurate.
In addition to the situation specific findings, global procedural justice and global
distributive justice had significant relationships with legitimacy as well. These findings
also supported Tyler’s (1990) theory and were consistent with previous literature on
procedural justice, distributive justice, and legitimacy. When examining global attitudes,
procedural justice is more important than distributive justice in predicting legitimacy
evaluations, though distributive justice is still important.
96
P
5
: Lower perceptions of procedural justice, distributive justice, and legitimacy
will increase the likelihood of behavioral responses to officer-citizen interactions
(e.g. justice-restoring responses).
The social psychological concept of justice-restoring responses was also
considered in this dissertation. In social psychology, justice-restoring responses is a
critical outcome in research on distributive justice (e.g. Markovsky, 1985), though it has
remained relatively unexplored in the criminal justice context. In applying the concept to
the field of criminal justice, there was a need to split the concept into two constructs –
procedural justice-restoring responses and distributive justice-restoring responses –
depending on the focus of the individual’s response. Consistent with the theoretical
proposition, situation specific perceptions of procedural and distributive justice were
primary predictors of an individual’s likelihood of engaging in justice-restoring
responses. However, legitimacy evaluations did not operate as expected. Though two
measures of legitimacy were employed, only one had a relationship with one of the
justice-restoring response measures. Further, this relationship was in the opposite
direction of what was expected. That is, individuals with greater trust in the police
evaluations were more likely to engage in procedural justice-restoring responses. The
fifth theoretical proposition, then, received partial support.
P
6
: Outcome favorability will moderate the relationship between procedural
justice, distributive justice, and legitimacy.
The sixth theoretical proposition proposed moderation effects with legitimacy as
the dependent variable. This proposition was tested with a series of models across
different outcome conditions, which were then compared to determine if the effects were
97
equivalent. In these models, however, none of the proposed moderation effects were
found and the proposition was unsupported. This suggests that it may be best to drop this
theoretical proposition. However, given the strong findings in social psychological
research regarding the moderating effect of outcome favorability (e.g., Brockner et al.,
1997; Brockner & Wiesenfeld, 1996), this finding may be context specific. As mentioned
earlier, the outcomes used in these vignettes were all relatively low risk. That is,
regardless of whether the individual received a ticket or a warning, the subject still drove
away after the officer-citizen interaction. Other interactions may have more serious
outcomes, such as arrest. When the outcome is more serious, the proposed moderation
may be more likely to occur as the subject encounters greater consequences from
variations in the outcome. Still, the findings here cannot speak to this possibility.
P
7
: Outcome favorability will moderate the relationships between procedural
justice, distributive justice, and behavioral responses to officer-citizen
interactions.
Finally, the last proposition proposed moderation effects with justice-restoring
responses as the outcome of interest. This proposition also proved problematic as only the
obligation to obey measure of legitimacy demonstrated significant differences. Further,
this difference was such that individuals with higher perceptions of obligation to obey
were more likely to engage in justice-restoring responses, the opposite of what one would
expect. On the other hand, there was a significant interaction effect for the officer
behavior condition such that the total effect of the officer’s behavior on procedural
justice-restoring responses was stronger in the ticket condition than in the warning
condition. This effect is in the direction anticipated by the work of Brockner and
98
colleagues (1997). The presence of this interaction in the total effect, but none of the
other direct effects, may be due to the strength of the effect. That is, the total effect for
officer behavior, because it encompasses a lot of the variation in both procedural and
distributive justice, is larger than the effects of either concept individually, and therefore,
easier to detect. Still, the evidence is not strong and the validity of this proposition is
dubious.
Revisiting the Theoretical Model
The theoretical model proposed in this dissertation received varying levels of
support depending on the specific propositions examined. Propositions related to the fair
process effect and the importance of distributive justice received considerable empirical
support from a variety of analyses. On the other hand, propositions regarding the
proposed moderation effects received no support and at times ran counter to the
expectations of the theory. Finally, the incorporation of justice-restoring responses had
mixed support, with the relationship between perceptions of fairness and justice-restoring
responses supported and the relationship between legitimacy and justice-restoring
responses rejected. Further testing is needed to confirm these findings before making
permanent changes to the theoretical model tested in this dissertation. However, for
clarity, a theoretical model consistent with the evidence presented in this dissertation is
pictured in Figure 7.1.
99
Figure 7.1. Final Theoretical Model
Practical Implications
This dissertation provided practical information for law enforcement in three
areas. First, the evidence regarding the fair process effect suggests that law enforcement
agencies should look more carefully at their treatment of individuals when claims of
outcome unfairness, such as racial bias, are made. Certainly, it may be necessary to
examine whether or not racial bias is occurring, but the analyses here demonstrated that
perceptions of outcome unfairness are driven by perceptions of officer behavior.
Therefore, changing officer behavior is likely necessary to reduce claims of outcome
unfairness.
Second, the analyses provided a deeper understanding of when citizens will file
complaints regarding an officer-citizen interaction. Police agencies may rely on a
convenient excuse regarding complaints – that individuals complain when they receive an
outcome they do not like. While the analyses here did show that complaints are more
likely when individuals receive an unfavorable outcome, officer behavior was still a
strong predictor of justice-restoring responses. Thus, complaints should be taken
seriously, as changes in an officer’s behavior can reduce the likelihood that a citizen will
100
be so unsatisfied with their experience that they file a complaint with the police
department or the city.
Finally, the theoretical propositions tested here provide a foundation for
understanding large-scale police protests, such as the Black Lives Matter movement.
These recent protests have largely focused on complaints of outcome unfairness with
respect to police use of force, arrests, and traffic stops. Individuals are more likely to
engage in a distributive justice-restoring response – a response that attempts to correct a
situation of perceived outcome unfairness – when individuals perceive more distributive
injustice. Further, individuals form their distributive justice evaluations, in part, using
their perceptions of the officer’s behavior. Thus, when an officer behaves in a manner
consistent with procedural justice principles, distributive justice-restoring responses are
less likely. What remains unexamined, however, are the conditions under which
individual justice-restoring responses, such as those examined here, become group-level
justice-restoring responses.
Theoretical Implications
From a theoretical perspective, the analyses presented here support the
overarching message of this dissertation: distributive justice is important and should not
be neglected in studies of the legitimacy of criminal justice institutions. The role and
meaning of distributive justice has been neglected in criminology for too long and should
be reconsidered given its importance in this theoretical model. Distributive justice is not
merely a competing normative component for procedural justice, but is critical to
understanding why procedural justice is important to our understanding of legitimacy.
Individuals base their judgments of outcome fairness on cues from an authority’s
101
behavior and decision-making process. These judgments of outcome fairness then impact
the legitimacy the individual grants to the authority.
This finding does not contradict the model proposed by Tom Tyler (1990). In fact,
the role procedural justice plays is, if anything, more important in this theoretical model.
Distributive justice is often an outcome of importance in and of itself. That is,
criminologists are often concerned with individuals’ perceptions of the fairness of
outcomes delivered by criminal justice authorities (e.g., perceived racial bias in traffic
stops; see e.g., Tyler & Wakslak, 2004). With the model proposed here, procedural
justice becomes a primary predictor of these perceptions. Moreover, in the models
proposed here, though the direct effect of situation specific procedural justice was, at
times, washed out, it’s total effect remained as large as or larger than the total effect of
distributive justice. In this way, the importance of procedural justice is not reduced, but
contextualized by the proposed theoretical model.
Finally, a wave of criticism over procedural justice and legitimacy in criminal
justice has risen to popularity (e.g., Nagin & Telep, 2017). These criticisms point out that
most research in the area has been cross-sectional and neglects the possibility that
legitimacy evaluations are shaped by a larger sociological context. Proponents of this
argument suggest that the relationships found in procedural justice research are spurious
and broader sociological concepts should be emphasized over the role of procedural
justice. Counter to this argument, however, the results presented here demonstrate that
officer behavior can influence perceptions of the police. Further, it joins a growing body
of experiments to support the argument that officer behavior matters in forming these
perceptions (e.g. Johnson et al., 2017; Lowrey et al., 2016; Mazerolle et al., 2012, 2013;
102
Maguire et al., 2017). Certainly, there is a need for continued rigorous research in this
area to determine the full extent to which officer behavior does impact behavioral
outcomes, but disregarding the theory at this point would ignore a growing body of
promising research in criminology and decades of strong research in social psychology.
Still, the critics of procedural justice are likely correct in asserting that a broader
sociological context matters in determining individuals’ legitimacy evaluations. In fact, a
number of recent studies have shown that factors other than procedural justice likely
influence an individuals’ legitimacy evaluations (e.g. Cavanagh & Cauffman, 2015; Fine
et al., 2016; Wolfe et al., 2017). However, often ignored in these arguments, is the fact
that procedural justice is the one component that the criminal justice system can control.
There is little that a police officer can do about parental upbringing, neighborhood
conditions, or poverty, but it is well within the ability of every police officer to behave in
a manner consistent with the principles of procedural justice.
In sum, though this dissertation focused on the neglected concept of distributive
justice, it further reinforces the role of procedural justice in forming attitudes towards the
police. As a result, it joins a chorus of scholarly literature in encouraging police
departments to embrace procedural justice as a method of improving citizens’ attitudes
and improving police-community relationships.
