A Reconceptualization of Social Bond Theory to Predict Change Sequences in Offending

Abstract

We argue that reconceptualizing social bond theory (SBT) through incorporation of dual agency and change can identify unique causal change sequences, improve its ability to explain offending, and generate new questions about it. The reconceptualization recognizes that individuals and those with whom they interact play an ongoing role in contributing to the bond. It shows that changes in the bond can contribute to changes in delinquency through three sequences, each with a unique over-time pattern that depends on how bond agents respond to delinquency. We identify implications for SBT—highlighting that theoretical arguments about static effects do not necessarily equate to straightforward predictions about change effects—and, more broadly, efforts to advance theories of offending.

Introduction

Criminological theories of offending have been critiqued for insufficient attention to causal change processes (Agnew, 2005; Bruinsma, 2016; Dooley & Goodison, 2020). Many criminological theories, for example, present models of offending that assume unidirectional causation. Missed in these accounts is the possibility that causal effects may be reciprocal (i.e., bidirectional). Social bond theory (SBT; Hirschi, 1969)—alternatively referred to as social control theory and one of the most widely cited theories of offending (Costello & Laub, 2020)—is illustrative. It assumes that higher levels of the social bond are associated with a lower probability of delinquency, but ignores the potential for bidirectional change effects, such as an increase in an adolescent’s offending that might decrease the social bond.

Scholars have highlighted the potential for crime causation to involve reciprocal effects and the need for theories to take such possibilities into account (Agnew, 2005; Liska & Reed, 1985; McGee & Farrington, 2019; Thornberry, 1987; Thornberry & Krohn, 2019). They have highlighted as well the potential for individual agency (i.e., the ability to influence one’s life conditions) to play a causal role (Sampson & Laub, 2005; see also Paternoster & Bachman, 2017). Even so, there remains a need to develop theoretical accounts of dual agency, that is, the potential for different participants in relationships to influence one another in a dynamic interactive process that can only be identified and assessed through a focus on change over time. Doing so may address the relative inattention to causal change processes in criminological theories (Bruinsma, 2016) and improve these theories’ predictive accuracy (Weisburd & Piquero, 2008). It also may help to address concerns about theoretical stagnation that can occur when theories are not reconceptualized to account for new ideas (Dooley & Goodison, 2020).

We argue that SBT can be reconceptualized to incorporate these ideas in a manner consistent with the original formulation, and that doing so allows for identification of unique causal sequences, an improved ability to explain offending, and the generation of new questions, insights, and testable predictions. We draw on a model from the field of ecology (Wangersky, 1978), which sets out a model of how interactions between different populations may influence one another, to identify a series of causal processes that stem from dual agency, change effects. We ground this argument in research that highlights ways in which social forces affect youth and in which youth interact with and affect their social relationships in a dynamic process that can lead to a cycle of increases and decreases in the social bond and in delinquency. We then discuss implications of this extension of SBT not only for the theory but also for how other theories of offending might be improved through a focus on dual agency and change.

Background

Despite substantial advances in the development of theories of offending (Cullen et al., 2008), a recurring critique is that the theories typically are limited in their predictive accuracy (Jeffery, 1959; Weisburd & Piquero, 2008). A related critique is that the theories frequently suffer from insufficient specification of causal mechanisms, including change and reciprocal causation processes (Agnew, 2005; Bernard, 1990; Bruinsma, 2016; Mears & Stafford, 2002). One response is to work to specify more precisely how a theory can be adapted to include predictions about change. Doing so may simultaneously provide clarity about the theory’s causal logic, provide guidance about hypotheses that emanate from the logic, improve predictive accuracy, and lead to new questions and avenues of inquiry.

To advance this argument and work on SBT, we reconceptualize SBT to include an account of dual agency and change. SBT, which Hirschi (1969) presented in Causes of Delinquency, proceeds from a Hobbesian view of individuals as perpetually driven to pursue their own interests. For Hirschi (1969), the question is, What restrains individuals from offending? Hirschi’s (1969) answer was that the social bond—comprised of the four dimensions of attachment, commitment, involvement, and belief—inhibits this drive. These elements bind the individual to others and so create the equivalent of a mechanical governor or brake against actions that would risk adversely affecting others’ views or support. The theory’s central hypothesis, then, is that the greater the level of the bond, the lower the likelihood of offending.

Despite support for the theory (Costello & Laub, 2020)—what Akers et al. (2021, p. 133) review characterized as “moderate to low” support—scholars have highlighted limitations in the theory (Agnew, 1985, 1991; Gibbs & Erickson, 1975; Greenberg, 1999; Hirschfield & Gasper, 2011; see, generally, Oleson & Costello, 2020). Some of the limitations center on details about different elements of the bond and ways in which other theories, such as general strain theory or social learning theory, may be related to SBT.

One additional limitation centers on an incomplete account of individual agency and change. Although Hirschi’s (1969) account contemplated the possibility that the social bond requires interaction with others, and so can increase or decrease depending on the amount of these interactions, it did not go further to consider the effects of changes in the bond or the salience of dual agency—the idea that adolescents and those with whom they interact may influence one another—in creating change. Research highlights, however, that individuals and those with whom they interact can and do shape each others’ views and actions (e.g., Laub & Sampson, 2003; Paternoster & Bachman, 2017; Wikström, 2006). In addition, studies suggest that social bonds may be influenced by offending through reciprocal causation, suggesting a causal role of offending in influencing youth or those with whom they interact (Hirschfield & Gasper, 2011; Thornberry et al., 1991). Thornberry’s (1987) interactional theory, for example, emphasized that bonds may influence delinquency but also that delinquency may influence bonds. In a related vein, Sampson and Laub’s (1993) age-graded informal social control theory pointed to the role that life events may have in introducing changes to the bond and thus to offending. These accounts underscore the potential salience of agency, interactions, and change processes, but have not systematically examined their implications from within a social bond theory logic. They also have not addressed the assumption in SBT that bond agents have a limitless capacity to bond with youth, and the corresponding inattention to the implications of a limited bonding capacity for predictions about change effects.

Different strategies exist to improve theory. One avenue is to rely on formal theory, but this can be difficult and has not been widely embraced (Gibbs, 1972; Hage, 1994; Homans, 1964). Another approach, and the one we adopt here, is to focus on elaborating on process (Bruinsma, 2016) and context (Abbott, 1997). This approach accords with Merton’s (1968, 1973) observations about the nature of theoretical advances, which may come in many ways, including thinking systematically about process and change, and the ways in which they arise (see also Dooley & Goodison, 2020; Mears & Stafford, 2002).

Our starting point, then, centers on process and change in SBT. For the theory, it is clear that the individual’s bond to others matters and that for youth, parents, and teachers are primary social bond agents. Others (e.g., community leaders and coaches) might be as well. Hirschi’s (1969) discussion assumed a largely one-way child-to-parent (or other bond agent) view of bonding processes, but these processes may be dynamic, involve multiple parties, and may be shaped by and through ongoing interactions. For example, greater exposure to more bond agents, such as two parents rather than one (Rankin & Kern, 1994) or parents, coaches, and so on, creates more opportunities for interactive and reciprocal effects to arise, depending on the extent to which parents and other bond agents want to or have the capacity to invest in the child.

Put differently, the bond necessarily must be created, sustained, or changed through interactions. As Mead (1934) and the subsequent development of symbolic interactionism (Blumer, 1969) and labeling theory (Becker, 1963) highlighted, individuals think about themselves in the context of others. Their identity and sense of value and place derives from how others see us, and we in turn shape how they view themselves. Interactions perforce involve two-way dynamics: Person A can influence person B, and person B can also affect person A.

For SBT, consideration of this two-way dynamic highlights that relationships involve distinct possibilities that depend on what each person brings to them. To illustrate, an adolescent may feel attached (+) or not (−) to a parent, and the parent may feel attached (+) or not (−) to the adolescent. Four possible situations then arise: The two are both attached to one another (++); the adolescent is attached to the parent but the parent is not attached to the adolescent (+−); the adolescent is not attached to parent but the parent is attached to the adolescent (−+); and neither is attached to the other (−−). McGloin (2009) has emphasized a similar point in showing that youth may adjust their offending relative to those of the peers with whom they associate. These examples highlight that the relationships can change because of the actions of one or both parties, such as how adolescents or their parents act or feel.

These considerations highlight, not least, that change—arising from actions and reaction of individuals and those with whom they interact—is central to the social bond and, indeed, to relationships. A static view misses changes in the bond and, by extension, causal processes that may underlie or differ from what SBT contemplates. A focus on dual agency necessarily leads to a focus on change, and the insight that any given state of affairs does not “just happen.” There are processes that must give rise to changes and their effects.

Dual Agency, Change, and the Lotka-Volterra Model

Consideration of dual agency and change (i.e., changes in and interactions between agents) allows for a range of causal possibilities. A focus on dual agency recognizes that individuals affect others and others affect these individuals, while a focus on change recognizes that individuals can and do change. Taking both possibilities can yield predictable patterns of outcomes that necessarily can only be observed over time.

One example of this observation—the Lotka-Volterra model (Wangersky, 1978)—comes from ecology. It argues that, all else equal, competition between prey (A) and predator (B) populations creates a series of lag effects that result in each generating population changes in the other. For example, in the first stage, prey populations increase (A_t1) and then predator populations eventually increase (B_t2). They do so due to a variety of mechanisms, the main one being attraction to a food source that is increasingly available. At the second stage, the increase in predator populations leads eventually to decline in prey populations (B_t2 → A_t3). The predators compete with one another for the prey; larger numbers of predators translate into greater consumption of available prey. At the third stage, prey populations decrease, which translates into less available food to support the predator population. Accordingly, the predator populations decrease, whether through death, moving to new areas for other prey, or other mechanisms (A_t3 → B_t4). Finally, as the predator populations decrease, prey populations eventually increase because of the reduced predation (B_t4 → A_t5). In short, the sequence involves lag effects that alternate from positive to negative:

1. As prey populations increase, predator populations eventually increase [A_t1 → B_t2 (+)].

2. As predator populations increase, prey populations eventually decrease [B_t2 → A_t3 (−)].

3. As prey populations decrease, predator populations eventually decrease [A_t3 → B_t4 (+)].

4. As predator populations decrease, prey populations eventually increase [B_t4 → A_t5 (−)].

The model has many applications in ecology (Abrams, 2000; Hold & Bonsall, 2017) but also to areas outside of ecology, such as economics (Becker & Leopold-Wildburger, 2020; Wang & Wang, 2016). Like any model, it simplifies reality through assumptions. In what follows, we extend the same logic to SBT. Our argument is not that SBT’s account of offending should center on a model of competition. Rather, we argue that the model’s focus on the agency of different groups and on change, including reciprocal causation—and, most importantly, its identification of predictable sequences that flow from specific changes—can be incorporated into SBT to identify three distinct sets of causal sequences.

Dual Agency, Change, and a Reconceptualization of Social Bond Theory

Hirschi’s (1969) theoretical account did not take into account dual agency or change, and that has resulted in a missed opportunity to refine the theory in ways that can identify particular patterns of offending and that may improve the theory’s predictive accuracy. In our reconceptualization, the main tenets of SBT remain, but these issues are addressed. We assume that (i) agency involves not only youth but also those with whom they can bond, (ii) offending may influence youth, social bond agents, and, in turn, the social bond, and (iii) change processes entail specific sequences that vary depending on the nature of youth and bond agent interactions.

These points warrant elaboration because they establish the foundation for reconceptualizing SBT. First, youth interactions with social bond agents (e.g., parents, teachers, and neighbors) need not always be positive forces in the production of the social bond but also can be negative, thus causing a higher rather than a lower likelihood of offending. Research on child-rearing and on SBT and reciprocal causation highlights this possibility (e.g., Kiriakidis, 2010; Liska & Reed, 1985). At the same time, a long literature from symbolic interactionism underscores that identities and meaning are negotiated through interactions (Matsueda & Heimer, 1997). Parental responses, or those of other bond agents, to children may vary depending both on their relationships with the children and on social and cultural contexts. Incorporating that insight changes how one understands the social bond. Specifically, the child’s view is not the only one that matters. Other bond agents’ views and how they respond do so as well.

Second, as prior research establishes, reciprocal causation can arise in many crime causation processes (Agnew, 2005), not least SBT (Hirschfield & Gasper, 2011; Thornberry et al., 1991). That is, the social bond may affect offending, and offending may affect the social bond. This possibility highlights the potential for interactive processes to vary in whether they enhance or weaken the social bond. Research on parent-child relationships shows that parents influence their children and that children influence their parents in a dynamic interactive process (Kiriakidis, 2010). Much the same can be said about interactions that youth not only have with parents but also with others in their schools, peer networks, and community. In every instance, the nature of how children or social bond agents respond to particular behaviors may vary. For example, while some parents might respond to delinquency through an increase in attachment, others might respond by becoming less attached.

Third, this set of considerations means that social bond or offending changes can generate several predictable sequences. Each sequence, however, depends on the content, or nature, of the youth’s response to bond agents and bond agents’ responses to the youth. The reconceptualization recognizes this possibility rather than assume that bond processes only occur in one direction and that only positive bonding processes occur. In this way, it differs from Thornberry’s (1987) interactional theory, which emphasized that interactions among causal forces can occur in a bi-directional manner but did not identify causal sequences, presented below, specific to a social bond theory logic. Similarly, it differs from Sampson and Laub’s (1993) age-graded informal social control theory by identifying causal sequences not tied to age or particular life events but to the logic of social bond change processes.

Building from the above assumptions, we reconceptualize SBT as including dual agency and change. Core tenets of the theory remain, but with recognition of interactive change processes and of the potential for lagged effects, similar to the logic suggested by the Lotka-Volterra model of competition. Specifically, we expect that changes in the social bond contribute to changes in delinquency through three possible sequences, each with a unique over-time pattern that depends on how social bonding interactions unfold and how bond agents respond to delinquency. Each sequence logically flows from a dynamic, interactive view of the social bond and delinquency, but not the static one implied by the theory.

The set of lag effects and logic underlying each are discussed below. The first sequence is termed, for shorthand, the “best of all possible worlds” model because it results in situations when bonds constantly increase and delinquency constantly decreases. The second is the “squeaky wheel investment” model because it results in situations where parents or other bond agents alter their investment according to the level of attention that they view as necessary. If delinquency occurs or is increasing, they may invest more in the child; if it does not occur or is decreasing, they invest less. The third is the “squeaky wheel rejection” model, which recognizes that an alternative response to a child’s delinquency may be to pull back or reject the child.

Sequence 1: “Best of All Possible Worlds” Model

This sequence represents the conventional view from SBT that social bond agents are a positive force and never a negative one, and it takes into account the logical extension of this idea, namely, that their responses to decreased delinquency will be positive. The end result is an endlessly positive cycle. For illustrative purposes, we focus on attachment, which arguably represents the central element of the social bond. In this “best of all possible worlds” sequence, the specific anticipated lag effects are as follows:

1. As the bond increases, delinquency eventually decreases [B_t1 → D_t2 (−)].

2. As delinquency decreases, the bond increases even more [D_t2 → B_t3 (−)].

3. As the bond increases, delinquency decreases even more [B_t3 → D_t4 (−)].

4. As delinquency decreases, the bond increases even more [D_t4 → B_t5 (−)].

In the first stage, the bond (B) increases, and so delinquency (D) can be expected to decrease by time 2. SBT does not explicitly state what happens when attachment increases or decreases or, by extension, what the lag periods or processes are that determine a given period. Per the logic of the theory, however, it can be anticipated that an increase in attachment would lead to an eventual decrease in delinquency due to greater restraints against offending. In the second stage, the decrease in delinquency can be expected to result in further increases in the bond by time 3 (see, e.g., Kerr et al., 2012). Bond agents may be more likely to be attached to and involved with prosocial youth. In the third stage, this increase in the bond results in further decreases in delinquency by time 4. In turn, the bond increases still further by time 5.

Observe first that in each stage, the change effects accord with the general logic of SBT. That is, more social bonds generate less delinquency. The reconceptualization, then, generates a set of predictions that align with what SBT predicts.

Second, a logical flaw in the theory surfaces when we consider change. The theory implicitly, and unrealistically, assumes that parents and others have limitless amounts of attachment, involvement, and so on that they can offer a child. They clearly do not, which does not mean that this “best of all possible worlds” sequence cannot exist. But the fact of limits in bonding capacity highlights that other sequences may be possible.

Third, this sequence ignores the possibility that a response to decreased delinquency may not always be to invest even more in a child. SBT does not actually state how bond agents will respond to a child’s delinquency. However, the theory necessarily must rest on assumptions about how bond agents react to children when they act in different ways. This sequence recognizes one possible reaction, while the other sequences recognize alternatives.

Fourth, the sequence highlights a logical incongruity in SBT that hinders recognition of different sequences of reciprocal causation. Hirschi (1969) puts the views of youth at center stage, not the views or actions of bond agents. Consider, then, that this sequence makes little sense when viewed from the perspective of a child. For example, at stage 2, why would a decrease in delinquency increase a child’s attachment to their parents? As we have argued, that only makes sense when we consider social bonding processes as involving a two-way interaction between youth and others, where, for example, parents might respond in a favorable manner to a child’s decreased offending and where the child responds by increasing their attachment.

Sequence 2: “Squeaky Wheel Investment” Model

This sequence represents an alternative set of possible responses that arise from recognition that parents and other bond agents may have a limited capacity for bonding. In contrast to some discussions of reciprocal causation that assume that delinquency necessarily reduces parental or other agents’ bonding, it recognizes as well that that parents and other bond agents do not necessarily pull away from or reject youth who engage in delinquency, and actually may invest more in them. The specific sequence involves these lagged effects:

1. As the bond increases, delinquency eventually decreases [B_t1 → D_t2 (−)].

2. As delinquency decreases, the bond eventually decreases [D_t2 → B_t3 (+)].

3. As the bond decreases, delinquency eventually increases [B_t3 → D_t4 (−)].

4. As delinquency increases, the bond eventually increases [D_t4 → B_t5 (+)].

The process in stage 1 is the same as for sequence 1. In the second stage, however, the social bond decreases, thus resulting in a positive correlation (whereas all associations are negative in sequence 1). Why? A parent or any bond agent may turn their attention elsewhere once they feel that they have traversed a problematic period in a child’s life. They may do so because they need to focus on additional life priorities, including other children (Hanushek, 1992), which may inadvertently reduce attachment over time. This process can happen in all relationships, as when one party in a relationship comes to take the other for granted.

In the third stage, and consistent with sequence 1, the expectation, per SBT, is that delinquency should increase given a decrease in the bond. However, in the fourth stage, a completely different prediction arises. Here, as delinquency increases, the bond eventually increases. Why? Youth may act out, but that does not mean that parents only respond by rejecting them (Kiriakidis, 2010). When children have problems, including acting out, parents and other bond agents may see that as a signal—a “squeaky wheel”—that calls for intervention, thus increasing attachment and other dimensions of the bond. As Williams and Steinberg (2011) have noted: “Having an adolescent who is charged with a serious felony offense may lead some parents to rethink their parenting strategies and adopt more responsive, authoritative styles” (p. 634). The more common response may be to become less responsive or authoritative (see, e.g., Kerr et al., 2012), but the possibility remains that some parents respond differently to situations when their children act out or engage in externalizing behaviors.

Although sequence 2 diverges markedly from sequence 1, it remains grounded in the logic of SBT. The divergence arises from recognition of different patterns of responses to changes in the bond and to delinquency on the part of children and bond agents, respectively. In addition, sequence 2 incorporates an arguably more realistic appraisal of some parents’ or others’ bonding capacity. Perhaps for some individuals, that capacity is limitless. For most, though, it is a finite resource that must be expended carefully. From this perspective, it is reasonable to anticipate that parents and other bond agents may decrease their attention to youth once they have helped the youth to get over a difficult period. It arguably represents a more realistic view of how life events may introduce disruptions to social bonds, with attendant consequences, as Sampson and Laub’s age-graded informal social control theory (1993) highlighted. To illustrate, attending college may be clear evidence of an individual following a prosocial path, but it might simultaneously be accompanied by a decline in parental attachment and involvement. Such changes might give rise to an increase in offending, as SBT would anticipate.

Sequence 3: “Squeaky Wheel Rejection” Model

The third possible sequence can be described briefly. Here, instead of parents or other bond agents responding to delinquency by investing more in the child and thus increasing the social bond—as a response to the “squeaky wheel”—they react negatively. This sequence is suggested by research on reciprocal causation and SBT, and anticipates that parents or other bond agents may withdraw from or reject the child. The specific sequence of lagged effects is:

1. As the bond increases, delinquency eventually decreases [B_t1 → D_t2 (−)].

2. As delinquency decreases, the bond eventually decreases [D_t2 → B_t3 (+)].

3. As the bond decreases, delinquency eventually increases [B_t3 → D_t4 (−)].

4. As delinquency increases, the bond eventually decreases [D_t4 → B_t5 (−)].

The main difference here is that in stage 4, parental and other bond agents reject the youth rather than increase their investment in them (Gault-Sherman, 2012; Liska & Reed, 1985; Williams & Steinberg, 2011). As with sequence 2, this reaction is a potential response. Neither the sequence 2 or sequence 3 bond agent responses can likely be determined a priori, and any effort to do so would require a different theoretical focus. What the reconceptualization identifies is simply that the different responses are possible and that, when viewed from the perspective of a sequence of changes, multiple sequences of bond and delinquency associations can result.

Implications of the Reconceptualization of Social Bond Theory

Several implications flow from a reconceptualization of SBT that recognizes the salience of dual agency and of change. First, reconceptualizing SBT this way, and through identification of three possible sequences, allows for additional SBT hypotheses. It also allows for identifying and explaining causal complexities involving social bonds and delinquency. Importantly, the sequences include more than the one outcome anticipated by Hirschi (1969) that was assumed to be positive: “If the bond to the parent is weakened, the probability of delinquent behavior increases; if this bond is strengthened, the probability of delinquent behavior declines” (p. 88). As we have demonstrated, alternative sequences and outcomes, not all of which are positive, are possible and yet still align with the theoretical argument that social bonds influence delinquency.

Second, a key to understanding the reconceptualization centers on recognition of the dynamic nature of social bonds, and underscores the importance of theory and research that focuses not only on the causes of offending but also on the causes of the causes (Agnew, 2005). Hirschi’s (1969) conceptualization relied on a relatively static view of social bonds, which in part may be why he argued for cross-sectional research and against longitudinal research (Gottfredson & Hirschi, 1990; Hirschi & Gottfredson, 1983). That is understandable given that, at the time he developed his theory, longitudinal data were not common. Even so, a dynamic view of the social bonding processes arguably reflects reality more accurately. Sampson and Laub’s (1993) age-graded informal social control theory illustrates a step in that direction by focusing on the effects of life events that may change individuals and their circumstances and, in turn, the bond. There are, however, other types of change processes that do not require or entail dramatic turning points. For example, in everyday life, change is normal. People get in arguments, difficulties arise at work or home, families or friends move, accidents happen, and so on. These are all events that, over time, can influence individuals and those with whom they interact.

Third, the reconceptualization highlights the need for research on which sequences are most likely. Hirschi’s (1969) theory appears to assume a primarily positive view of parents and other social bond agents, which in turn would suggest that sequence 1, the “best of all possible worlds” model, is most common. It assumes only beneficial effects of changes in social bonds and delinquency. However, other sequences clearly are logically possible. Although some might argue that these sequences might be viewed as disconfirmation of Hirschi’s (1969) theory, we would argue instead that they support the idea that social bond processes may contribute to delinquency. The difference simply is that there are different logical and predictable sequences, when one considers dual agency and change processes.

Fourth, there is, at the same time, a need to identify what contributes to different sequences. One factor might be the contextual circumstances in which families find themselves. For example, a family with more resources might be more likely to fall into sequence 1 or 2, while a family with fewer resources might be more likely to fall into sequence 3. This type of contingency could explain why there have been inconsistent findings in the family size literature, with some studies finding that family size is positively associated with delinquency (Farrington, 2010) and others finding that it is not (Tygart, 1991). Hirschi (1969), in fact, noted the potential variability in what family size could represent, observing that the presence of older children allows for assistance in supervising and parenting other children. Such observations underscore that family dynamics, and those of social bond agents and youth more generally, constitute a critical avenue of inquiry on their own and may shed light on the bond-delinquency relationship.

Fifth, the reconceptualization points to the importance of investigating the role of “floor” or “ceiling” effects. In contrast to what SBT assumes, parents do not have an infinite amount of bonding to offer to their children, and youth do not commit an infinite amount of delinquency. There must be a “ceiling” for how much bonding or delinquency can occur. That in turn must translate into the identified sequences as well as limits in the effects of changes in social bonds and delinquency. Similarly, there can be “floor” effects. As attachment increases, delinquency should decrease, but it can only decrease to the point of zero offenses. SBT does not explain what could or should happen in this instance. The reconceptualization highlights this issue and suggests one way in which the theory could dovetail with desistance scholarship. For example, it could be argued that sustained non-delinquency creates an ever-stronger social bond, one that may help insulate youth from diverse criminogenic forces. This idea would be ironic, given Hirschi’s opposition to seemingly atheoretical longitudinal studies. Yet, it flows directly from the idea that delinquency, or its absence, may influence bond agents. This idea and the notion of floor and ceiling effects point to additional lines of inquiry. What, for example, contributes to variation in how parents or other bond agents respond to a child’s first delinquent act? Or to how a child responds to their responses to first-time delinquency?

Sixth, reconceptualizing SBT through a focus on dual agency and change creates opportunities for generating insights about elements of the bond. Many accounts of SBT treat each element as static. The level observed in a study somehow just happened to exist, and it, in turn, affects delinquency. This approach diverts attention away from processes that may illuminate important nuances in the logic of SBT. By contrast, a focus on dual agency and interactive change allows for identification of causal processes. It not only anticipates reciprocal causation but also how a cyclical process—of bond influences on delinquency and delinquency influences on bonds—may arise. To illustrate, Liska and Reed (1985) found that delinquency causes weak attachment to parents and school more than the other way around. That pattern may be interpreted to mean that teachers withdraw from or even reject problem students. However, the pattern could change as the student commits less delinquency. Such change might alter teachers’ views of and reactions to the student. Similarly, some parents might actively work to strengthen attachment to their children who commit and then later refrain from delinquency. A reconceptualization that retains SBT’s core theoretical logic but that places agency at center stage can address the “static” assumption oversight in the theory and research on it—and possibly Hirschi’s (2004) attempt to reconcile or integrate SBT with self-control theory—while also drawing attention to the need to understand how elements of the bond develop and change. Paternoster et al. (2015) have made a similar argument about the need to incorporate insights about agency into criminological theory.

Seventh, the reconceptualization points to the need for theory and research on whether the identified processes unfold the same for all elements of the bond. The SBT literature is not clear about which elements matter more than others (Costello & Laub, 2020). According to Hirschi (1969), SBT derives from Durkheim’s (1897/1951) notion of egoistic suicide. Attachment is the closest to that notion. Integration, which is the cause of egoistic suicide, refers to the strength and durability of social relationships, which is similar to attachment. There is, then, justification for viewing attachment as the primary causal force of SBT, and the one most relevant for considering the dual agency and change reconceptualization presented here. But perhaps involvement, commitment, and belief operate in different ways. On the face of it, an argument for similar processes can be made as those for attachment. For example, under the sequence 2 (“squeaky wheel investment”) model, the second part of the causal process involves delinquency decreasing, which in turn reduces a given element of the bond. How might a decrease in delinquency contribute to a decrease in commitment or involvement? One possibility is that it may be difficult to stay committed to achieving conventional goals or to be involved in various prosocial activities, especially without reinforcement for it from social bond agents. Such reinforcement may decline as parents and other bond agents perceive an adolescent to be on a seemingly prosocial track. A similar argument can be made for an adverse effect on beliefs. A central insight from the correctional intervention literature is that beliefs can change, and that reinforcement is critical to change. When problem behaviors decline, the risk arises that individuals may revert to earlier beliefs in the absence of reinforcement from parents and other bond agents. Such examples aside, agency may matter for some bonds more than others. Or perhaps the directionality of agency effects may vary for some elements of the bond. Changes in maternal attachment, for example, might influence youth more so than changes in paternal attachment (Craig, 2015). These possibilities warrant investigating.

Eighth, continued work in developing and testing SBT should center on longitudinal predictions and analyses of change, given that cross-sectional tests may be masking a dynamic process that SBT, as originally formulated, does not anticipate (Gault-Sherman, 2012). Understanding the causes of these changes, one of which may be changes in delinquency, is central to predicting the different sequences. At the same time, it may help to improve the predictive accuracy of SBT. Scholars have highlighted the extensive body of work supporting SBT’s predictions (Costello & Laub, 2020). Others, though, have critiqued its weak predictive accuracy (see, e.g., Akers et al., 2021, pp. 99, 133; see also Greenberg, 1999). One reason for any limitations in its predictive accuracy may stem from a failure to take into account the dynamic processes that shape the social bond and the interaction between the bond and delinquency.

Ninth, the predictions about unique sequences of bond and delinquency lag effects do not contradict the notion that there can be or is a cross-sectional negative correlation between bonds and delinquency. A cross-sectional correlation represents an averaging of different types of dual agency, reciprocal effects, and of the three sequences. The nature of those effects could theoretically result in an average negative, null, or positive relationship. Consistent empirical support for the negative cross-sectional relationship anticipated by SBT suggests that the average effect is negative. But this finding does not provide evidence about the sequences and processes in which delinquency affects the bond (Liska & Reed, 1985).

Tenth, the reconceptualization of SBT highlights the need for research on the timing of changes in social bonds, beyond the age-graded ones identified by Sampson and Laub (1993), and the likelihood of committing delinquency. Specifically, what determines the lag time? For example, how long does a youth need to commit or show signs of being willing to commit delinquency before bond agents might feel compelled to respond? What is the level of acceleration or deceleration in the lag? For example, when a youth has consistently gone months without committing delinquency, this duration may suffice to signal to social bond agents that the youth is the equivalent of a sure bet and so no longer requires close attention. At that point, it might take very few missteps (e.g., offending) by the youth to draw attention from social bond agents. However, it might be the case that the agents have turned their attention elsewhere. In that case, youth may need to substantially increase their delinquency before bond agents focus on them. Investigating these and other possibilities will likely require more granular lags (e.g., month-level) than is typical of many longitudinal surveys. However, advances in technology and survey administration make it possible to create data that would permit investigating them.

Eleventh, a larger theoretical implication of the reconceptualization of SBT is that a focus on cyclical processes may identify a similar phenomenon for many causes of offending. That possibility suggests a need to revisit theories with an eye toward understanding if cyclical lagged effect processes align with the causal process articulated by the theories. This idea echoes in part the calls from other scholars to investigate reciprocal and interactive relationships between delinquency and various causes of delinquency (Agnew, 2005; McGee & Farrington, 2019; Thornberry & Krohn, 2019).

Finally, the reconceptualization highlights opportunities to improve the predictive accuracy of crime theories by focusing on agency and change. In some accounts, such as Greenberg’s (1999) re-analysis of Hirschi’s (1969) Richmond Youth Survey data, SBT has been found to have weak predictive accuracy. One reason may be that, unlike social learning theory (Akers et al., 2021), SBT does not incorporate notions of dual agency and change. Indeed, many theories of offending provide little to no account of change, the effects of different levels of change, or reciprocal causation. Modifications of the theories to take account of these possibilities may improve their predictions. Doing so will require incorporation of individual agency (Paternoster et al., 2015), the ways in which interactions influence causal variables, and how offending affects these causes (Agnew, 2005). The relative inattention to agency is, we should emphasize, not specific to SBT. For example, labeling theorists sometimes characterize individuals as being fully influenced by labels, as if the individuals lack the ability to reject them (see, however, Maruna, 2001). They also tend to ignore change within individuals or those with whom they associate. Revisiting theories and reconceptualizing them with a focus on dual agency and interactive processes might improve their predictive accuracy and lead to new questions about causal processes (Healy, 2013; Paternoster & Bachman, 2017; Thornberry & Krohn, 2019).

Policy Implications of the Reconceptualization of Social Bond Theory

SBT’s central implication for policy has been that any interventions that improve an individual’s bond may reduce offending. As the reconceptualization highlights, additional implications flow from consideration of dual agency, change, and sequences of lagged effects. One is that interventions could and perhaps should focus on both youth and bond agents and how they respond to each other. Programing that targets only juveniles, as if the central cause of their offending lies within them, may fail if they do not incorporate strategies for ensuring that bond agents engage in activities that may increase their own and children’s bonds.

Another implication is that formal social control agents (e.g., police and courts) may want to be vigilant about consistent investment in youth over time and avoid the potential tendency to focus only on higher risk youth or to assume that these youth will remain high risk. Research suggests that interventions may be more effective when they target higher-risk individuals (Latessa et al., 2014; MacKenzie, 2006). The reconceptualization, however, highlights how lower-risk youth can become higher-risk. There is the risk of net-widening when focusing on lower-risk youth, but a middle ground can include diversion programs that focus on intervening with youth and their parents, as well as other potentially important social bond agents in the youths’ lives.

Still another implication is that juvenile justice system responses might be improved by viewing youth in light of dual agency and change. When law enforcement and the courts see an adolescent, they do so on a given day and not how the youth’s behavior—or the causes of the behavior—has been trending. Their delinquency might be increasing or decreasing. And the causes might stem from a myriad of causes, one of which could consist of how their parents or other social bond agents interact with them. For example, some parents might withdraw from their children while others might invest more in them during difficult periods. Such possibilities suggest that an adolescent’s behavior should be understood within the context of their past behavior and the role that a variety of social bond agents may play in their current behavior.

Not least, the reconceptualization of SBT underscores the potential importance of efforts that empower youth in gaining a greater sense of agency. Changing their physical or social surroundings can be one way to remove negative bonding experiences and transition to ones that may be more positive. Parents and other bond agents can assist with such efforts. But when they cannot, youth need strategies that they can independently adopt to improve their lives.

Conclusion

Since the 1940s, the field of criminology has accumulated many different theories of offending, with SBT being one of the most prominent, but they still have weak predictive accuracy (Weisburd & Piquero, 2008) and have not fully developed or incorporated an account of causal change processes. Some, like general strain theory and social learning theory (Agnew, 2005; Akers et al., 2021), have gone well beyond their original formulations and made advances by explicitly seeking to introduce agency or change or both. We have argued that a reconceptualization of SBT that centers on dual agency and change, and that articulates the sequences of change that can arise from reciprocal effects of bonds and delinquency, provides a way to advance the theory, improve its predictive accuracy, and open new lines of inquiry. The reconceptualization aligns with the original articulation in some predictions, but it also points to three distinct causal sequences and bond-related explanations of them. It is, we acknowledge, less parsimonious, but it also allows for identifying new patterns and explanations of them.

We submit that a similar approach can be taken to revisiting the logic of other theories. This work begins with taking seriously the notion of dual agency and the process and consequences of changes in causal variables and offending (Paternoster & Bachman, 2017). In every instance, a crucial test of whether doing so improves the theories is whether empirical studies can demonstrate greater predictive accuracy as a result (Dooley & Goodison, 2020; Gibbs, 1997). Another is whether it identifies new patterns or explanations that may better capture the causes and effects of crime. But such steps cannot be taken, or will not likely be especially productive, without first carefully articulating the role of individual agency and change in a given theory.

Improving theory to create a better understanding of phenomena constitutes a central goal of science (Bruinsma, 2016). At the same time, doing so holds the potential to inform policy. That many criminological theories of offending do not directly engage with notions of agency and change diminishes their salience for efforts to prevent or reduce crime. Advances that address this situation have the potential for providing insights into the causes of offending and possibly new, more effective policies, programs, and interventions.