ChapterPDF Available

Pauwels, L., Ponsaers, P. , Svensson, R. (2010). “An analytical perspective on the study of crime at multiple levels”, in Social disorganisation, offending, fear and victimisation – Findings from Belgian studies on the urban context of crime, Pauwels, L., Hardyns, W., Van de Velde, M. (eds.), The Hague: BJu Legal Publishers, Reeks Het groene gras, 195-211.

  • Ghent University, Faculty of Law and Criminology Co-director Research Group IRCP


In this contribution we want to draw attention to and stimulate the theoretical study of crime causation, i.e. the etiology of crime occurrence and offending. One could argue that there are too many theories on these subjects already. Criminologists are familiar with the ‘classics’ such as Merton’s Anomie Theory or Hirschi’s Control Theory. Yet, despite a proliferation of theories, theoretical improvement has not always been the result. To stimulate theoretical improvement, we want to present an approach to the study of crime at multiple levels that is deeply rooted in the fundamentals of sociology. We are convinced that this approach not only serves the scientific study of etiological research, i.e. the study of crime as the commitment of acts, but it also leads to a balanced relationship between theory, empirical research and policy. Although this approach is applied to studying the causes of offending, it could also be applied to other theoretical frameworks since we put forward a general style of theorising.
Pauwels, L., Ponsaers, P. , Svensson, R. (2010). “An analytical perspective on the study of crime at multiple levels”,
Social disorganisation, offending, fear and victimisation Findings from Belgian studies on the urban context of
, Pauwels, L., Hardyns, W., Van de Velde, M. (eds.), The Hague: BJu Legal Publishers, Reeks Het groene gras,
10 An analytical perspective on the study of crime at multiple levels
Lieven Pauwels, Paul Ponsaers and Robert Svensson
10.1 Theoretical improvement in the study of crime as social fact and individual act
Why are some societies more crime-prone and why do some individuals commit acts of crime
more often than others? These questions remain fundamental for criminology and criminal
policy. The answers to such questions determine our judgments on crime as a social
phenomenon, and the individuals who commit crime (Wikström, 2004). Beyond that, the
answers to these questions also influence how we react to crime through criminal policy
measures, crime prevention and the criminal justice system. Policy measures need to be
accurate and effective, i.e. they must succeed in reducing crime as a social phenomenon.
Therefore policy measures need to influence factors that are causally related to crime.
For many decades developing etiological theories was done relatively independently of
empirical analysis. Classic etiological theories provided explanations for the differences in
crime at the level of societies, neighbourhoods and ultimately individuals. At the same time,
empirical analyses conducted at that time had many flaws characterized and this lead to a gap
between theory and empirical analysis. Fortunately, more recently formulated theories are the
result of new research findings and empirical developments (Bruinsma & Weerman, 2007).
This contribution aims to reduce the gap between theory and empirical research even more.
After all, what is the point of theories that lack empirical foundations? Theories of crime
causation should have strong empirical grounds, i.e. they should be rooted in social reality. The
role of theories is to improve our understanding of the social world around us. ‘Without a
theory, empirical research often lacks wider significance, and without empirical research,
sociological theory easily turns into fictitious storytelling’ (Hedström, 2005, p. 114, see also
Elster, 1989, p. 7-8).
Figure 17: The relationship between theory, empirical research and policy
Our starting point is that one always needs to find a balance between theory, empirical research
and criminal policy. If one puts too much emphasis on one of the aspects in Figure 17,
reification can be the result.100 If we emphasise theory too much, and thus neglect empirical
research and criminal policy, scientific activity will be reduced to ‘practicing science for its
own sake’: criminology will be reduced to prosaic texts. A theory-driven criminology that lacks
empirical grounds and policy relevance puts the scientist in his ‘Ivory Tower’. Putting too much
empha sis on empirical research, and thereby ignoring theory and policy, leads to blind
empiricism, the reification of science as number crunching and fetishism of the number. Science
is then reduced to a collection of data that is carefully ranked and reported in a book of results
but without meaning or social and policy relevance. The result is a patchwork of observations
that grows endlessly, without direction, meaning and use. Science that is too narrowly related
to policy is also reification. In this situation, science is reduced to a technical instrument of
criminal policy that adopts unverified theories without trying to empirically test their underlying
hypotheses. Science then is merely a technical instrument to confirm hypotheses that were
formulated a long time ago by policy makers. Criminological research should be relevant for
policy, not per se supportive of policy.
In this contribution we want to draw attention to and stimulate the theoretical study of crime
causation, i.e. the etiology of crime occurrence and offending. One could argue that there are
too many theories on these subjects already. Criminologists are familiar with the ‘classics’ such
as Merton’s Anomie Theory or Hirschi’s Control Theory. Yet, despite a proliferation of
theories, theoretical improvement has not always been the result. To stimulate theoretical
improvement, we want to present an approach to the study of crime at multiple levels that is
deeply rooted in the fundamentals of sociology. We are convinced that this approach not only
serves the scientific study of etiological research, i.e. the study of crime as the commitment of
acts, but it also leads to a balanced relationship between theory, empirical research and policy.
Although this approach is applied to studying the causes of offending, it could also be applied
to other theoretical frameworks since we put forward a general style of theorising.
Our point of view is that mechanism-based action orientated explanations can be applied in the
field of criminology. The key philosophy behind this approach is that social phenomena (such
as the amount of crime occurring in a society, or offending by individuals) need to be explained
from a constellation of social factors, participants and their actions and reactions, that are related
to the dependent variable in such a way that they usually bring about the social phenomenon or
action that we try to explain (Hedström, 2005). The concept of action refers to observable
actions, behaviours and attitudes of actors in the social reality. The idea is that actions of
individuals shape a society. Micro-micro statements are about the effects of individual level
conditions or attitudinal characteristics on individual level actions or behaviours, while macro-
macro statements explain differences in the occurrence of acts of crime at (multiple) societal
levels, such as cities, neighbourhoods and schools.
Within the field of sociology, it was Coleman (1986) in particular who brought macro-micro
relationships and statements to the attention of scholars. Coleman (1986) strongly supported
what he referred to as ‘methodological individualism’. Methodological individualism means
that aggregate level correlations or societal outcomes should be seen as consequences of acts
of individuals that are part of the aggregate. Coleman’s methodology is an important anchoring
point in the study of crime at multiple levels of analysis.
We refer to this approach as analytical criminology. This way of theorising is a scientific realist
approach to crime that is rooted within the tradition of analytical sociology (for a discussion
see Hedström & Swedberg, 1998; Hedström, 2005). ‘Analytical sociology seeks to explain
complex social processes by carefully dissecting them and then bringing into focus their most
important constituent components. It is through dissection and analytical abstraction that the
important cogs and wheels of social processes are made visible and intelligible’ (Edling &
Hedström, 2005, p.1). An analytical approach to crime differs from classical etiological studies
in that many more attempts are undertaken to visualise and expose the social processes that are
actually at work when an explanation of crime is formulated.
The analytical approach to crime as action is influenced by scientific realism, today’s answer
to the positivist approach (Bunge, 2004). Let us therefore clarify the difference between
scientific realism and the positivist approach to crime.
10.2 From a positivist approach to scientific realism
Within the field of criminology, several philosophical traditions or paradigms exist. Some
scholars have argued that paradigms can be considered as either ‘positivist’ or ‘symbolic
interactionist’ (Edling, 2007). We do not agree with this oversimplification of reality. In fact,
some theories of crime causation share propositions of symbolic interactionism, such as
Sampson and Laub’s theory of cumulative disadvantage (1997). In that theory, crime is not
exclusively explained by the characteristics of individuals involved, but attention is also paid
to the relatively persistent character of serious offending by small groups of adolescents. That
is where influences of symbolic interactionism come into the picture. The theory states that a
harsh reaction (e.g. by the State or the police) can explain the ongoing character of offending
among recidivists. In our view, this theoretical dichotomy or ‘clash between theoretical
paradigms’ can be put down to tensions between researchers that preferred causal
(‘nomological’) traditions and those that prefer hermeneutic (‘interpretative’ or ‘verstehende’)
traditions in criminology. These tensions peaked in the 1960s and 70s. Fortunately tensions and
mutual misunderstandings as a result of extreme interpretations of positivism and
constructivism have by and large declined. We are convinced that it is of great importance that
at least some healthy tensions should exist between paradigms, as long as they contribute to a
fruitful discussion and an increase in knowledge relevant to criminology.
Unfortunately such strong tensions did not completely disappear. In some aca demic circles
(within criminology and sociology) research on causes of crime is still considered ‘not done’
(Edling, 2007). Scholars that conduct etiological research are often depicted as being unwilling
to admit that crime is a socially constructed reality (Jupp, 1989; Radzinowics, 1966). It has
nevertheless been underscored many times that researchers who focus on moral rule breaking
do not per se see crime as a characteristic that is exclusively attributable to people (Olaussen,
2004; Pauwels, 2007). Such a view is referred to as the ontological point of view on crime. We
are convinced that an ontological view should be abandoned as well because it does not
recognise the sociological view on ‘social facts’ developed by Durkheim (1969), i.e. the view
that the social reality exists on its own.
In contemporary criminology, causal theories are mainly developed from a scientific realist
point of view. Realism is of interest because it may bridge a gap between paradigms that are
conflicting in their dogmatic application. Realism refers in this context to the idea that society
exists independent of scientists and the theories that scientists produce
. From this angle,
realism has its roots in positivism. Realism, although influenced by positivism, does not
disapprove of constructivism. Realism explicitly recognises the Thomas’ theorem (“If men
define situations as real, they are real in their consequences”- this is the basic idea of symbolic
interactionism), and even considers Thomas’ theorem as one major explanation of why
objective conditions are perceived differently. One important task of science is to discover how
the world functions by looking for explanations and searching for causes. Therefore we strongly
prefer a realist approach to crime when the main aim is to discover the causes of crime and how
people get involved in moral rule breaking. To us it seems that a non-constructivist approach to
social construction is the most fruitful way to study the causes of crime at various levels of
aggregation, i.e. from Durkheim’s social facts to offending by individual members of a society.
This approach is useful because crime does not exclusively exist in the minds of criminologists,
but also happens in the real world outside. In the real world, crime leads to the suffering of
victims, offenders and society.
From a realist approach, causes and explanations are not invented but discovered through
induction and deduction (Bunge, 2006a). A realist approach will therefore never strive to
produce grand theories nor can it be seen as equivalent of ‘blind empiricism’: the eclectic and
arbitrary variables-based
approach to crime. One might wonder why Durkheim is considered
so important to analytical criminology as Durkheim’s view may seem to contrast with
Coleman’s methodological individualism. Indeed, Durkheim exclusively focused on the
analysis of social facts as rates of behaviour, and by doing so he developed the foundations for
macro-level theory. He also emphasised that ‘social facts’ have a much more specific meaning
than the phenomena that occur in society, and have some general societal importance.
Furthermore he underscored that sociology should be organised around the kind of facts that
share special characteristics. Durkheim named three characteristics: (1) objective
, (2) forced
, and (3) specific
. According to Durkheim, crime should unequivocally be seen as a
social fact, as a reality of it own or ‘sui generis’, that also exists outside the mind of the
The founding fathers of sociology, such as Durkheim, Weber, Parsons, but also Marx, described
the consequences of denying the existence of social facts independent of ourselves. Denying
that social facts exist independent of ourselves is typical of constructivistic approaches of
science, such as marxist criminology, radical criminology and abolitionism (Christie, 2004).
Such extreme interpretations of constructivism, like the extreme interpretation of positivism,
undermine the very essence of science. If every aspect of the social world around us were
Today it is quite generally accepted that the hypothesis, and thus the nature of the knowledge that has to be
produced, determines whether one chooses an etiological or symbolic interactionist approach. Both points of view
lead to a different kind of knowledge and the researcher needs to be explicit when making a choice.
2 Variable-based criminology refers to the a-theoretical multiple factor approach or risk factor approach to crime
and delinquency that has made the study of empirical correlates an end in itself.
Durkheim (1969) talks about the ‘extérieur’, the exterior context. Durkheim discusses obligations that an
individual does not create himself, but that he is subjected to, and thus exist outside of the individual.
Durkheim also uses ‘impérative et coercitive’ to refer to sanctions. Social facts do have a coercive and ordering
power, which means that social facts are imposed on individuals. The coercion is intensified when one tries to
resist it.
With this Durkheim means that they are determined by society. Durkheim introduced the group or groupings. It
deals with facts that are compelling (norms and opinions), that are specific to a group, a society. This is the origin
of the term social facts.
interpreted as social constructions, then science would be of no use, as it would be impossible
to learn something about the world and how things works (Edling, 2007). At best we would
learn how a social construction is constructed. Scientific debate would no longer be possible
because every argument about how social reality functions would be equally valid, there would
be no standard (social reality itself) against which to measure scientific thoughts. The concern
that criminology from an extreme social constructivist approach would undermine its own
legitimacy, is the main argument for preferring a realist approach to crime, at least when the
key objective is to get to advanced theory-based knowledge. Such an approach has both feet in
empirical reality, and has the potential to effectively deal with the prevention of crime.
As Wikström (2004) stated, it is probably more fruitful to study crime causation as a separate
research question (and thus to answer the question of why some people commit crimes knowing
that punishment will follow) without simultaneously having to explain why some behaviour
has been outlawed while other behaviour has not
. By doing so, theories about crime remain
embedded within social reality and are not at risk of becoming grand theories that can never be
adequately tested and therefore remain middle-range theories. In short, we are convinced that
the further development of middle-range theories from a scientific realist point of view provides
the key to developing theories of crime causation. A scientific realist approach to crime means
that attention is paid to mechanisms and processes leading to crime under certain societal
conditions. Analytical criminology fits the theoretical framework of scientific realism and is
put forward in this contribution as a useful tool in the study of causes of crime as action at
different levels of aggregation
10.3 Action-oriented explanations and the rational choice perspective
‘Humans have agency, that is, a capability to act upon their environment in a
purposeful way. Action, the expression of agency, may be defined as behaviour
(acts) performed under the persons’ guidance. A theory of action specifies what
moves people to act. (Wikström, 2004)’
Wikström (2004) argues that criminology knows little about causes of crime, in contrast with
the huge number of crime correlates. This contrast is due to the lack of general theories about
action in criminology. Within other fields, like sociology and economy, theories of action do
exist. In adapting a general framework of action-oriented theory, Wikström argues that
behaviour of individuals should be seen as a consequence of the behavioural options that actors
have and the actual choices that they make. Those behavioural options are not chosen in a
vacuum, they are a consequence of the interaction between moral attitudes and the social
milieus to which actors belong. The concept of choice refers to the judgment of situations
(Wikström, 2004). Individual characteristics and characteristics of social settings can act as
direct or indirect causes of actions that are undertaken, but only if they influence the behavioural
It is not per definition necessary to answer the question ‘Why is it forbidden to smoke in a public space?’, if one
wants to answer the question ‘Why do some individuals knowingly break that law?’ This does not mean that
answering the question of criminalisation is less important, it is merely a different issue altogether.
Typical of both early positivism and the empiricism of a variable driven criminology, is that no attention is paid
to not-directly observable phenomena. Processes and mechanisms are not directly observable, but are a part of
core business of science. Positivists only studied relation ships between directly observable characteristics, and
used a rather deterministic interpretation of causality. In addition, positivists, such as Tarde, Ferri, Lombroso
unconditionally questioned rationalists’ belief in the free will of individuals.
options of individuals, such as the perception of behavioural alternatives and processes of
Action-oriented theories are often a priori classified as theories that belong to the rational choice
perspective (Edling & Stern, 2003). The rational choice perspective serves only as a framework
allowing for the study of action and is indeed useful because it can explain why people make
particular choices from a broader variety of options (Kleemans, 1996; Verwee, 2007). It has
been stated that one shortcoming of the rational choice perspective is that it is not possible to
scientifically measure and study every behavioural option or choice that an individual
con siders (Elffers, 2005). But in a less rigorous interpretation of the rational choice
perspective, as is familiar in the Scandinavian tradition (Edling & Stern, 2003), action is not
merely seen as the outcome of pure rational choices, but additionally as a consequence of moral
choices and values. From that point of view, attention is paid equally to the interaction between
moral values, desires and opportunities that can be seen as important motives for the actions
that actors bring about.
An analytical approach to crime can benefit from the rational choice approach by taking into
account both rational and moral criteria in the analysis of acts of crime.
10.4 Principles of an analytical criminological approach towards crime as action
An analytical approach to criminology as a social science can provide researchers with more
theoretically guided, and therefore more profound, insights into the social world around us. This
approach is rooted in analytical sociological theorising. It explains complex social phenomena
related to crime by carefully dissecting the problem and emphasising the components central to
these phenomena. Following Edling & Stern (2003), analytical criminology emphasises the
study of the social structures within which mechanisms operate to explain individual acts as
outcomes of a series of social processes. An analytical criminological approach can use the
analytical sociological approach to the study of social facts by focusing on following principles:
explanation, dissection and abstraction, precision and clarity.
10.4.1 Explanation
Explanations have an added value compared to descriptions because they answer why-
questions. Analytical criminology brings the message that we need to bridge the gap between
grand theories and the empiricism of a variables based criminology. A variables-based
criminology runs the risk of getting bogged down in the swamp of the crime and crime-related
correlates. A mere theoretical criminology pretends to tell the truth by ‘story-telling’. Such
‘truth’ quickly becomes dogma because it is not subjected to empirical tests. Analytical
criminology needs to explain the ‘why’s’ criminal behaviour and events, and ‘why’ social
factors differ in time and location, between societies and communities. Elster (1989) argued
that the focus should be on explanation by mechanisms while Hedström & Swedberg (1998, p.
25) argued that a ‘mechanism-based explanation seeks to provide a fine-grained as well as tight
coupling between explanans (the explaining causally linked concept) and explanandum (that
what has to be explained’. The tighter the link, the higher the level of clarity will be. This
principle of tightness is very important. If we want to explain why and how it is that adolescents
that have low self-control offend more frequently, then we have to explain this link. The
explanation should be clarifying (see below: ‘Precision and clarity as third principle’). One such
explanation could be that low self-control makes adolescents more vulnerable in criminogenic
settings, particularly those adolescents that tolerate moral rule breaking.
10.4.2 Dissection and abstraction
Dissection refers to the division of both social phenomena and their causes, and causal
mechanisms that explain the outcome. These dissections should result in the reduction of social
phenomena to their essence. The researcher should abstract this essential information about a
social phenomenon theoretically. The dissection of a social phenomenon deals with the
measurement of concepts, the translation of theoretical concepts into observable, operational
elements. This empirical translation deserves more attention in contemporary criminology.
Hedström (2005, p.3) argued that ‘dissection and abstraction are two aspects of the same
activity, and they are core components of the analytical approach’. Through dissecting and
abstracting, our understanding of processes becomes more transparent and clear. One example
is the dissection of morality as a central concept of criminology. What does the theorist actually
mean by that concept? Is morality a general concept referring to broader societal norms that are
related to conforming behaviour, or is morality a multidimensional concept that refers to levels
of tolerance of moral rule breaking or legal cynicism towards the moral principles stated in
criminal law? A careful theoretical dissection of this concept is necessary to improve our insight
into the phenomenon and its role in the explication of crime as action.
10.4.3 Precision and clarity
Precision and clarity should characterise the theory-driven hypotheses that are about to be tested
by empirical findings. The creation of theory in criminology needs to demystify, not reify or
add confusion. Thus, theoretical concepts need to be clearly interpretable, as do the
relationships between concepts, and mechanisms that bring about effects. This distinction
should already be present at the abstract and theoretical level, i.e. before starting an empirical
analysis, but also at the empirical level. Precision and clarity therefore imply a transparent way
of measuring concepts in empirical studies.
We are convinced that an analytical approach that takes into account these principles can
improve our knowledge because it forces the theorist to carefully consider why the phenomenon
under study should be studied, which conditions and mechanisms should be taken into account
and why. The theorist should have a clear vision of the broader and more abstract concepts that
deserve attention in an analysis. From this theoretical attention it follows that empirical tests of
hypotheses can be interpreted more directly, so that results of analyses are less ambiguous and
less ambivalent. By doing so, studies of crime and crime causation at multiple levels of
aggregation become more fruitful and provide more profound insights for science and policy
makers. Thus, there is less room for discussions about which theoretical concepts really do
matter in the explanation of crime and which do not. The principles of analytical criminology
could be considered to be acting against another principle of empirical research: the parsimony
principle. Following the principle of parsimony, one would only focus on those concepts that
increase theoretical knowledge. Scientists that follow the principle of parsimony use as few
mechanisms as possible: only those mechanisms that ‘directly affect a dependent variable and
increase the explained variance in a dependent variable’. We argue that parsimony may impede
theoretical ideas of how concepts are related to each other in nomological networks. Indeed,
the principle of parsimony may also encourage the criminologist to overlook mechanisms that
do not directly operate and influence a dependent variable. We argue that, from a theoretical
point of view, it is just as important to gain insights into what is referred to as the ‘causes’ of
‘the causes of crime’, i.e. theoretical concepts that have influence on the direct causes of crime
at several levels of analysis. An analytical approach to crime can therefore be best conducted
from the principle of sequential integration. This means that attention is paid to the complex
networks of relationships that link causes of crime.
10.5 Individual and collective action or the logic of an action oriented analytical
Figure 18: Linking macro and micro-levels of crime
Macro-micro questions in criminology are a consequence of correlations between social facts,
such as structural characteristics of ecological settings and crime, and offence rates of these
settings as social outcomes. The arrow that goes from A to D in Figure 18 represents this macro-
level relationship. If the correlation between A and D in Figure 18 is caused by underlying
characteristics measured at a lower level, i.e. at the level of individual actors and their actions,
the ecological correlation is spurious and the methodological individualism of Coleman is
justified. This is also the case if there is a true contextual effect of macro-level characteristics
on individual level outcomes (see the arrow from A to C). In that case causal mechanisms that
explain individual differences and changes partially operate at the aggregate level. The
contextual effect then is not caused by the differential composition of settings, and is partially
responsible for social outcome D. The development of micro-level theories is then useful for
explaining macro level phenomena. Methodological individualism serves as a means to
critically reflect on the relationships between aggregate level observed relationships, also when
people’s actions are partly influenced by context itself. In the field of criminology, attention to
micro-macro problems has been growing as a consequence of increases in urban segregation
because of poverty, and fear of the negative consequences of ghettos. The integration between
micro level and macro level is called cross-level integration (Liska, Krohn & Messner, 1989).
It has been acknowledged that the process of theoretically integrating different explana tory
levels of offending is extremely complex. Scholars consider the issue of how and why
individual characteristics and settings interact to be of major importance to improve our
knowledge of juvenile offending. Baerveldt (1998) before us has argued that Coleman’s
methodological individualism could be very useful in the refinement of macro-level theories.
One of Baerveldt’s most important statements was that macro-level theories should be
improved through theoretically dissection by detecting causal links between mechanisms at the
macro- and micro level. However, he did not use the principle of mechanisms. From that point
of view Baerveldt can be seen as an analytical criminologist ‘avant-la-lettre’. Scholars such as
Sampson, Raudenbush and Earls (1997) consider aggregates to be contextual units that have
their own characteristics, and are more than the sum of their parts. Also, according to Bunge
(2006b) the action of individuals as social actors cannot be understood without taking into
account contexts, networks and even broader structures of individuals. On the contrary, these
contexts should be studied together with the motives of individual actors. The criminologist
that has an environmental view is wrong because he does not take individual characteris tics
into account, whereas the individualist is wrong for ignoring context i.e. the settings to which
individuals are exposed. Individuals are connected with society and its structures. Following
Baerveldt (1998), macro-level units that form ecological systems in society should be carefully
chosen from a theoretical point of view.
It is an important theoretical task for analytical criminology to state why a macro level unit
plays an important theoretical role. Mario Bunge (2004, 2006b) can fill a theoretical gap in
explaining why this is so. Bunge adheres to the systemic approach to crime in which one needs
to make a distinction between:
(1) the composition of the system, i.e. the collection of parts of the system;
(2) the environment of the system or the collection of entities that are not a part of the
system but do have an influence on parts of the system;
(3) the structure of the system, i.e. the whole of structural relationships between the parts
of which a system comprises
(4) the mechanisms that function within a system.
As mechanisms play such an important role in evaluating theoretical improvement, we devote
a separate paragraph to the study of mechanisms.
From a methodological point of view, using the logic of analytical criminology remains the
only way to avoid aggregation bias. Aggregation bias is the consequence of wrong assumptions
about the relationships between characteristics measured at different levels of analysis and this
is the very essence of micro-macro debates in social sciences. Since Robinson (1950) it has
been known that the paying exclusive attention to macro-level relationships and analyses is not
without dangers. Robinson (1950) documented the well-known ‘ecological fallacy’ and in
doing so he referred to the lifework of Shaw and McKay (1942), prominent scholars of
ecological (or macro-level) criminology at the time.
The discussions of the time dealt with the effects of exposure to ecological settings on
individuals. Shaw and McKay were convinced that exposure to ghetos had negative
consequences for children and adolescents. That statement was derived from the ecological
observation of a correlation between neighbourhood poverty and school truancy. The authors
did not question if valid statements could be made about contextual influence of neighbourhood
poverty on individual level outcomes observed among the neighbourhood residents, such as
school truancy or offending. Some were convinced that effects of ecological settings could be
put down to setting composition rather than setting characteristics, such poverty lev els of the
individual or household. Hauser (1970) warned against the contextual fallacy in the 1970s in
his famous article ‘context and consex: a causionary tale’. The contextual fallacy is thus the
incorrect interpretation of a contextual effect. This happens when one mistakes an individual
Structure should not be confused with the composition of the system. Structure refers to the relations between
actors of a system.
level characteristic for its eco logical counterpart. Valid causal judgments about characteristics
of macro-level units first need to be identified at the theoretical level and must be empirically
measured and exist independently of intervening mechanisms operating at the individual level.
We are convinced that this can only be achieved by taking Coleman’s methodological
individualism seriously as an anchor point for analytical criminology.
10.6 The clarifying role of mechanisms in an action-oriented analytical criminology
Wikström (2004; 2007) has made strong arguments to support the notion that the problem of
causation has not always been well treated in criminological research. The etiology of crime
does not understand very well the causal mechanisms and processes that are at work when
individual differences and changes in crime are to be explained. That is why there are too many
correlates of crime (e.g. biological, psychological and sociological), which merely obscure a
causal interpretation (Farrington, 2002). As a consequence some scholars have stated that
nothing really matters while others have stated that everything matters in the explanation of
crime (Wikström, 2004). This situation does not help theorists, nor does it help practitioners
such as probation officers and social workers in their attempts to learn about crime causation.
The basic philosophy behind a mechanism-based explanation is that social phenomena are
explained by those mechanisms that bring about differences and changes in a condition (the
explanandum or the dependent variable in a study). The Norwegian sociologist Jon Elster
(1989, p. 3-4) argues that the explanation of an event or act deals with plausible reasons why
the event or the act took place.
The attention to mechanisms is not new to our proposed style of theorising and is rooted in the
ideas and theories of very influential sociologists such as Robert K. Merton (°1910-2003) and
Emile Durkheim (°1858-1917). To this day we can learn important lessons from these founding
fathers of sociology, as Edling & Hedström (2005) stated earlier, and these lessons go beyond
historical issues. These early theorists were not always theoretically specific and they did not
have the most sophisticated tools of analysis to test their hypotheses. Yet early theorists deserve
respect for their approach and vision on science. Emile Durkheim is, from a historical point of
view, the analytical sociologist avant-la-lettre because he was interested in social mechanisms
that operated at the societal level. Durkheim (1951) described mechanisms of self-regulation in
his analysis of the balance between individual and groups, and the effects of this balance on the
suicide rate in his classic book Le Suicide: étude de sociologie. Durkheim never used the word
‘social mechanism’. The sociological contribution of Robert K. Merton (1968) was a
mile stone when it comes to mechanism-based explanations. Merton strongly emphasised the
fact that theory and empirical research need to be very closely related in his groundbreaking
work Social Theory And Social Structure. Merton talked about social mechanisms as
fundaments of sociological middle-range theories and defined the word ‘mechanism’: as ‘social
processes having designated consequences for designated parts of the social structure (1968, p.
43)’. In contemporary sociology the concept of ‘social mechanism’ is prominent in the writings
of Mario Bunge (2006a), James Coleman (1986) and Peter Hedström (2005). In the field of
criminology Robert Agnew (1993), Per-Olof Wikström (2004), Robert Sampson and Stephen
Raudenbush (1999) can be seen as key authors that favour mechanism-based explanations.
According to Agnew (1993), a mechanism explains how something works, and it therefore
gives insight into the relationship between social structure and crime. Criminologists see
mechanisms as intermediate variables that are necessary to make an observed correlation
between characteristics plausible. Blalock (1964) described this relationship as the indirect
effect. There is however one danger that one should be aware of when following this
interpretation. One cannot consider a variable that represents a mediator of another variable to
be equal to the underlying mechanisms behind a known association, without having given
serious thought to cause-and-effect relationships. One clear example of how cause-and-effect
relationships have been given serious attention from a theoretical point of view can be found in
Wikström and Sampson (2003).
Having read Bunge (2006b), Hedström (2005), Hedström & Swedberg (1998), Elster (1989)
and Wikström (2007) we are forced to conclude that no consensus exists about precise
definitions of mechanisms. Both philosophers and social scientists give plenty of definitions.
Table 37 illustrates a selective anthology of interpretations of the concept of mechanism by
leading sociologists / criminologists.
Although these definitions differ from each other they do share one common theme that we
identify as a major task of an analytical criminology. These definitions clearly emphasise the
need to explain observed regularities and they do that by pointing to how they happen in detail.
Mechanism is often used as a synonym for process. The latter term refers actually more to the
dynamic character or time-aspect that is so typical of ‘to bring about’ in its causal meaning.
Hence, the study of mechanisms adds theoretical value to the quantitative explanatory research
into individual differences and changes in the actions of actors. We clarify our statement by
looking at the objectives of research into crime causation. Research into crime causation often
tries to make statistical inferences from observed phenomena in samples. Statistical inference
can never be seen as an equivalent for causal inference. Both forms of inference do have one
common denominator. Both try to use localised information to draw conclusions for more
general phenomena (Hedström, 2008). Statistical inference deals with the chance that an effect
parameter in a population does not differ from a ‘regression coef ficient’ as sample statistic.
From a pure analytical criminological perspective the aim is to go beyond the original sample.
What we ultimately intend is the generalisation of the process that brought about an act. The
focus is not on the relationship between variables, but on actors, and their acts as outcome of
diverse processes. Of course, we need to measure acts as variables. The issue at stake here is
that causality does not operate at the variable level. It is not because causal relationships are so
hard to detect using quantitative analysis (often even further restricted by the non-experimental
design and cross-sectional character of most studies) that such studies are restricted by this.
Each study has its own restrictions. It is important to realise that correlations never prove
causality convincingly. The opposite is true: causality implies the existence of empirical
correlations. Therefore it is possible to eliminate some statistical models that do not fit the
observed data (Blalock, 1964). For that reason, working with path diagrams that identify
relationships is imperative for analytical criminologists. By doing so, one clarifies what is to be
tested and one is obliged to provide a clear and well-motivated analytical argumentation of all
causal arrows, well-founded in its thinking about cause and effect. In variables-driven
criminology, the theoretical arguments as to why some variables are considered as causes of an
outcome in a quantitative analysis are lacking. Variables-driven criminology simply includes
variables in statistical models without carefully considering why they should even matter. As a
consequence, this type of research risks overcontrolling for the effects of other ‘variables’, and
therefore makes it difficult to assess influences of theoretically relevant mechanisms. The
explanations of why, and under which conditions, mechanisms bring about an effect are crucial
for both theory and policy. One other flaw in many variables-driven studies is that background
variables, that can never be interpreted as causal, are entered into equations for no good reason.
Gender, immigrant background and family structure can never be seen as the foreground of
actions undertaken by individuals. At best, they are entered into an equation because the actual
mechanism at work, which is associated with the background characteristic and the outcome
variable, is not measured at all. Thus, our style of theorising with its focus on mechanisms helps
to theoretically explain why actors do what they do.
10.7 Methodological consequences of an analytical approach
An analytical approach to crime as action that takes into account the macro-micro context of
crime as a social phenomenon is essentially quantitative. One possible explanation for this is
that it is a consequence of Coleman’s methodological individualism itself. Effect parameters in
quantitative analyses can easily be interpreted using advanced multilevel models that are
statistically able to make a clear and precise distinction between the aggregate level, contextual
effect and the macro level compositional effect
. Advanced theories require advanced
methodological tools. The study of crime as action requires that one starts from plausible cause
and effect relationships and that one needs to clearly conceptualise and create operational
variables, and develop hypotheses that explain how these relationships are connected to each
other. Methodological innovations and theoretical innovations should go together. It cannot be
denied that the increasing popularity of multilevel modelling has awoken the theoretical interest
of some criminologists, just as method problems have stimulated methodologists to find new
solutions for old theoretical problems. An analytical approach to crime as acts cannot
degenerate into a merely empirical matter, where the quality of empirical data used for complex
analysis becomes of minor importance. That would be the opposite of what is intended, namely
It is not our intention to suggest that qualitative analysis cannot be used for the macro-micro level discussion.
theoretical progression. It will not do to collect data with a sledgehammer and consequently
analyse it with the finest tools of the trade. Analytical criminology requires a balance between
10.8 Discussion and conclusion
To stimulate and improve the further development of advanced theories of crime causation and
research on effects of individual and ecological level mechanisms, we have put forward an
analytical approach to crime in this contribution. We have named this approach ‘analytical
criminology’. This approach is rooted in the current field of analytical sociology. The tradition
itself is not new but can be seen as an elaboration of the visions of founding fathers of sociology
such as Durkheim and Merton. In this tradition, crime is seen as the result of the action of
individuals, as is any behaviour. The study of causes, and not just correlates of crime as action,
is prioritised within the analytical approach. The theoretical explanation needs to make a
systematic study of the explanandum, i.e. that what is to be explained, and the explanans, i.e.
the explaining concept. Thereby theoretical concepts should be stated clearly and precisely. We
need to carefully study the mechanisms that are connecting causes and effects. This approach
to the study of crime also has roots in classic etiology of crime although it differs significantly
from that classic approach. The early positivist approach only took directly observable
phenomena into account and had a deterministic connotation. Indeed, this is a consequence of
the inheritance of classical positivism.
From this contribution it should be clear that narrow-mindedness from both a positivist and
constructivist point of view does not provide us with a key to the scientific study of crime
causation at several levels. Bunge’s scientific realist approach was chosen because it allows for
integrating aspects of both positivism and constructivism. From positivism Bunge argues that
the social world can be studied independently of the scientist, and that the study of crime
causation is perfectly possible without simultaneously paying attention to processes of
criminalisation and decriminalisation. From constructivism, Bunge’s realism recognises
validity in Thomas’ theorem by stating that objective conditions are perceived differentially. A
realist approach takes science seriously and wants to increase theoretical knowledge of crime
by recognising the parts of the aforementioned paradigms. Ultimately theoretical innovations
should be the result. The ultimate goal of a realist approach is to increase our knowledge on
social phenomena, which we argue should be the meaning of all science. The study of crime as
acts at multiple levels of analysis is gradually being influenced by such a mechanism-based
approach. This can be seen from the writings of internationally renowned scholars such as
Robert Agnew (1995), Per-Olof Wikström (2004), Sampson and Raudenbush (1999). The study
of social processes that connect causes and effects at different levels of analysis are being
studied within this approach. These levels should be explicated and the theoretical relevant
mechanisms that operate at each level should be clearly dissected and defined. The answer to
these ‘why’ questions can best be given when four principles are taken into account: the
principle of explanation, dissection, precision and clarity. These principles force criminologists
to study a social phenomenon such as crime and give an explanation that goes beyond the
relations between variables. We should keep in mind, also when it comes to crime prevention,
that causality does not operate at the variable level. Advanced theoretical development in the
study of crime causation is necessary if we want to improve our insight into the complexity of
a social phenomenon such as crime, and if we want to develop prevention projects that achieve
crime reduction.
One strong point of this approach is that it appeals for the study of crime as action at multiple
levels of analysis from a theoretical point of view. To illustrate the existence of crime at multiple
levels, Coleman’s approach towards the study of micro-macro relations was given as example.
Methodological individualism can help to determine the nature of observed correlations at
multiple levels. This framework can easily be extended to the study of other topics in
criminology while still using the same principles of analytical precision. Our message therefore
is that analytical criminology should not be interpreted as one more paradigm, but as a highly
useful method for both theorising and conducting research.
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.