Introduction

Conduct problem behaviour such as lying, cheating, disobedience, and fighting can have a lasting negative impact on individuals during childhood and adolescence, increasing the likelihood of their involvement in early onset and persistent offending (Farrington, 2005; Loeber, 1990; Loeber et al., 1993; Moffitt, 1993, 2018; Odgers et al., 2007). However, many who engage in conduct problem behaviour in childhood and adolescence do not become involved in later offending behaviour. Instead, their conduct problem behaviour decelerates by mid-adolescence, becoming increasingly less frequent or serious until it no longer causes harm to themselves or others (Loeber et al., 2016; Moffitt, 2018; Odgers et al., 2008). Many developmental scholars have proposed that prosocial behaviours, such as sharing, hel**, and comforting, may play an important role in this deceleration of conduct problem behaviour. The prevailing idea is that as individuals acquire more prosocial behaviours, their engagement in conduct problem behaviours decreases (Loeber et al., 2016; Masten & Tellegen, 2012; McGee et al., 2015; Moffitt, 2018, 1993). However, empirical examinations of the temporal dynamics between the increase in prosocial behaviour and the decrease in conduct problem behaviour are limited. Therefore, the direction of this relationship remains unclear. Filling this gap—the focus of this paper—can potentially improve the accuracy and timing of prevention and intervention efforts.

We undertake an investigation into the influence of the acceleration of prosocial behaviour on the deceleration of conduct problem behaviour. This examination encompasses two main aspects: (a) ascertaining the direction of the relationship between prosocial and conduct problem behaviour over time and (b) determining the long-term effects of change in one of these behaviours on the other. Through these explorations, we aim to understand whether prosocial behaviour acts as a facilitator or a measurable outcome of the deceleration in conduct problem behaviour from early childhood to middle adolescence. To achieve our aims, we employ an innovative analytical method, the General Cross-Lagged Panel Model (GCLM), using data from the Longitudinal Study of Australian Children (LSAC), a nationally representative sample. We will detail these data and the GCLM method later, but first, we explore the theoretical and empirical foundations of our research.

Albeit an underexplored topic, developmental and life-course criminology provides a lens through which the development of prosocial behaviour and its association with problem behaviour can be considered. Through this lens, the Social Development Model (SDM; Catalano et al., 2021) offers the most comprehensive and explicit theoretical framework for understanding how prosocial behaviour may develop and encourage children towards more prosocial than conduct problem behaviour. The SDM highlights the role of prosocial bonds in the development of prosocial and conduct problem behaviours across childhood and adolescence, suggesting that strong relationships within positive families, schools, and peer groups create opportunities for and reinforces prosocial, rather than conduct problem behaviour (Catalano et al., 2021). Research has consistently highlighted the significant role of positive social bonds in reinforcing prosocial behaviour and subsequently protecting against conduct problem behaviour (Loeber et al., 2016; Masten & Tellegen, 2012; McGee et al., 2015; Moffitt, 2018, 1993), providing evidence for the postulates of the SDM. However, in contrast to the SDM, this body of work traditionally focuses on prosocial behaviour as a factor that predicts the absence or presence of problem behaviour, rather than as a distinct behavioural pathway separate from pathways to problem behaviour. Contrary to this, our research proposes that prosocial and conduct problem behaviours are not merely opposites but separate behaviours with distinct pathways that influence each other over time, a postulate supported by recent studies (Speyer et al., 2023b; Zondervan-Zwijnenburg et al., 2022). Our aim is to explore, guided by the SDM framework, how changes in prosocial behaviour might influence the development of conduct problem behaviour across different stages of child and adolescent development. While drawing on the SDM’s theoretical underpinnings, our study is not a direct test of the theory itself but seeks to contribute to our understanding of the interplay between prosocial and conduct problem behaviour.

What has not been extensively examined in previous research is how reductions in conduct problem behaviour may potentially foster an increase in prosocial behaviour. Instead, previous research tends to explore the unidirectional influence of prosocial behaviour on conduct problem behaviour, typically in relation to its protective effects (Carlo et al., 2014; Padilla-Walker et al., 2015; Speyer et al., 2023a, b). However, there has been some recent interest in understanding the bidirectional relationship between prosocial and conduct problem behaviour. Specifically, Loeber et al. (2016) proposed that decreasing conduct problems might pave the way for enhanced prosocial behaviours, positing that such reductions could offer children more opportunities to develop prosocial skills—a hypothesis yet to be fully empirically explored. Recent research has provided preliminary empirical evidence in support of this hypothesis (Memmott-Elison & Toseeb, 2023; Padilla-Walker et al., 2018). However, Zondervan-Zwijnenburg et al. (2022) reported no within-person effect of past conduct problem behaviour on subsequent prosocial behaviour. This points to inconsistencies across studies.

Existing research has yet to conclusively determine whether the observed relationship between prosocial and conduct problem behaviour reflects a genuine interplay between the two constructs or if it is instead an artifact of overlap** developmental influences. While ample evidence suggests that both constructs are shaped by similar social mechanisms—such as parenting style, peer interactions, and socio-cognitive development (Ahmed et al., 2020; Bevilacqua et al., 2021; Carlo et al., 2014)—the question remains whether this relationship holds above and beyond these shared mechanisms. This highlights the need for research to explore the relationship between prosocial and conduct problem behaviour, independent of any unobserved mechanisms that are not direct drivers of this relationship, a gap that is addressed by our research.

What has also not been examined extensively in previous research are the long-term effects of changes in one behaviour on the other across critical developmental periods—early childhood, middle childhood, early adolescence, and middle adolescence. Longitudinal research tends to focus on a single developmental period, typically early adolescence (Cullen, 2011; Padilla-Walker et al., 2018; Speyer et al., 2023b). This focus may stem from the assumption that these behaviours are heterotypic (Chen & Jaffee, 2015). However, although the levels, motivations, and recipients of prosocial or conduct problem behaviours may vary with age and socio-cognitive maturity, these constructs remain relatively stable over time (Eisenberg et al., 2015; Moffitt, 2018; Padilla-Walker et al., 2015). Further, recent research confirming the invariance of measures of both prosocial and conduct problem behaviours suggests that these behaviours are likely homotypic (Murray et al., 2022; Speyer et al., 2023a). Given this stability, research should not be confined to a single critical developmental period.

Recent research, when considered collectively, offers insights into the relationship between prosocial and conduct problem behaviour over different developmental periods. Padilla-Walker et al. (2018) explored this relationship from early to late adolescence and found a reciprocal relationship from one period to the next, though they noted that there were no long-term effects of conduct problem behaviour in early adolescence on prosocial behaviour in late adolescence. Zondervan-Zwijnenburg et al. (2022) explored early to middle childhood, finding a consistent between-person reciprocal relationship, but no relationship when exploring within-person effects. Further, Speyer et al., (2023b) focused on the period from early to mid-adolescence and found that generally, high levels of prosocial behaviour in one period were negatively correlated with conduct problem behaviours (bullying and aggression) in the next period. Notably, Memmott-Elison and Tooseeb (2023) expanded the examination to ages 3 to 14, finding no relationship between behaviours from ages 3 to 5, but identifying reciprocal and long-term effects from age 5 to 14. These findings suggest that the relationship between prosocial and conduct problem behaviours may vary depending on the age range studied. Therefore, a comprehensive understanding of the reciprocal relationship between prosocial and conduct problem behaviour and whether this relationship persists over time may require exploration over extended developmental periods—a gap in current research or study aims to address.

Current Study

This study contributes to the limited research on the relationship between prosocial and conduct problem behaviour. Similar to recent studies (Speyer et al., 2023b; Zondervan-Zwijnenburg et al., 2022), we explore the temporal ordering of this relationship. We extend this research through two innovative statistical techniques: the General Cross-Lagged Panel Model (GLCM) and impulse response function (IRF). These methods, derived from econometrics, allow for the examination of the persistence and accumulation of reciprocal effects over time, making them particularly relevant for exploring key questions in developmental and life-course criminology. Additionally, we examine how prosocial and conduct problem behaviour interact across a significant developmental timeframe, from early childhood (age 4) to mid-adolescence (age 15). Only one other notable study, by Memmott-Elison and Tooseeb (2023), has examined this relationship over a similar developmental period (from age 3 to 14). Our research replicates the examination of this developmental period but differs due to the application of the novel statistical techniques described above, providing further empirical evidence for the nature of the relationship over this timeframe. Overall, we contribute a deeper understanding of the relationship between prosocial and conduct problem behaviour in the field of developmental and life-course criminology by drawing on the theoretical underpinnings of the SDM, using innovative dynamic modelling techniques, and considering the relationship over an extended developmental period.

Method

Data, Sample, and Measures

The data for this research are secondary data, sourced from the Longitudinal Study of Australian Children (LSAC), which tracks children and families across Australia to explore policy-relevant questions about their development and well-being throughout their lives. This nationally representative cohort was initially selected from 257 postcodes using the Australian Health Insurance Commission’s Medicare database, which includes records for all Australian citizens. Since its commencement in 2004, LSAC researchers have conducted biennial data collections, known as waves (Australian Institute of Family Studies, 2018). This study utilises data from the first six waves, covering an extensive developmental period from entry into formal schooling to mid-adolescence (aged 4 to 15). The main caregiver reports on prosocial and conduct problem behaviour, chosen for their consistency and minimal attrition, were used in this analysis. The subsample comprises 2780 children (1403 males and 1377 females) who had complete measures across all waves for both variables under study, representing 55% of the total initial sample size of 4983. The average age of the participants at wave 1 was 4.2 years and 1.4% (n = 38) identified as having an Australian First Nations background.

Attrition

By wave six of the LSAC study, there was an 11% longitudinal attrition rate. The LSAC data are non-monotonic (or general), in that some participants who dropped out in an earlier wave returned to the study at a later date (Norton & Monahan, 2015). In particular, there was a high rate of general attrition in wave three (when participants were aged 8 or 9) with over 13% of the informants not completing any questionnaires in this wave but returning in later waves. There was also a higher rate of attrition of First Nations children compared to other ethnicities, with a loss of over half of these participants by wave six (63%), suggesting that results should be interpreted with care in relation to First Nations participants.

To study the potential impact of attrition on the current analyses, we assessed whether there were any patterns of missing responses in the dependent variables (conduct problem and prosocial behaviour) based on key demographic variables including participant sex, informant ethnicity, or residential location. A pattern of missingness in these variables would suggest that this missingness may impact on the representativeness of the data. Visual assessment and Little’s MCAR test were used to investigate patterns of missingness through the MissMech package (Jamshidian et al., 2014) in R. Although males and females had similar rates of missingness across both prosocial and conduct problem behaviour, children with missing data across all the variables explored were more likely to be male than female. Thus, it cannot be claimed these data were missing completely at random.

To provide a method for researchers to address missingness, LSAC offers a weighting variable for each measurement period. This weighting variable is calculated from regression analysis of variables not used within this study (refer to Norton and Monahan (2015) for details on these variables). With no available data on missing participants, it is not feasible to examine if the absent individuals have different scores in the dependent variables compared with those who participated during that wave (Johnson & Young, 2011). Consequently, the weightings were excluded from this research. Instead, listwise deletion was applied, allowing for the examination of the accumulated effects of change in prosocial behaviour on changes in conduct problem behaviour. In order to counter any potential inflation of standard errors caused by missingness as well as any violations of normality and linearity, we used maximum likelihood to estimate the models and bootstrapped the standard errors. These approaches are robust to these issues when there is a large sample size and excessive violations are not present (Tabachnick & Fidell, 2001).

Measures

The Strengths and Difficulties Questionnaire: Prosocial and Conduct Problem Behaviour Subscales

This research uses the main caregiver reports of the Prosocial and Conduct Problem Behaviour subscales of the Strengths and Difficulties Questionnaire (SDQ). The SDQ is a screening instrument that targets indicators for child and adolescent mental health and developmental disorders (Goodman, 2001; Goodman et al., 2000). The reliability and validity of the SDQ is well established (Mellor, 2005). Studies have shown good concurrent validity of the SDQ across a wide range of populations and good sensitivity of the individual sub-scales in predicting strengths and difficulties (Goodman et al., 2000). The 5-item Prosocial Behaviour Sub-scale is a measure of strengths (Goodman et al., 2000). Items in this sub-scale ask the main caregiver if the child or adolescent: “often volunteers to help”, “is kind to younger children”, “is helpful if someone is hurt”, “readily shares with other children or adolescents”, and if they are, “considerate of other’s feelings.”

The 5-item Conduct Problem Behaviour Sub-scale, a measure of difficulties, asks the main caregiver if their child or adolescent: “steals”, “often lies or cheats”, “often fights or bullies other children”, “obey requests”, and if they “have a temper”. This sub-scale, along with other difficulty sub-scales (i.e., hyperactivity, peer problems, and emotional problems), is used for making clinical screening decisions about whether to refer a child for further assessments for behavioural disorders (Goodman, 2001).

The scores for both subscales are a derived variable created by LSAC where the mean of the five items is rescaled to provide a score between zero and 10. If less than three items in the scale are completed, the data are treated as missing. A score of zero represents low prosocial behaviour, and a score of 10 represents high prosocial behaviour. For conduct problem behaviour, a score of zero represents low conduct problem behaviour, and a score of 10 represents high conduct problem behaviour. The scales have acceptable internal consistency for each of the waves used in the current study (Prosocial Behaviour: a = 0.78 to 0.84; Conduct Problem Behaviour: a = 0.60 to 0.70), aligning with the internal consistency observed in previous research (e.g., Gomez-Beneyto et al., 2013). Table 1 shows the descriptive statistics for these sub-scales, including the correlations between prosocial and conduct problem behaviour, and population variance. The correlations indicate that conduct problem behaviour and prosocial behaviour are positively associated with themselves and negatively associated with each other across all waves.

Table 1 Means, standard deviations, correlations, with confidence intervals, and population variance
Full size table

Analytical Technique

In this study, we employ the General Cross-Lagged Panel Model (GCLM) and the impulse response function (IRF) to explore the developmental interplay between prosocial behaviour and conduct problems. These analytical techniques are particularly advantageous as they allow us to assess how changes in one behaviour impact the other over time, without the direct measurement of time-varying or time-invariant controls. It is important to clarify that while fixed factors such as gender (Pursell et al., 2008; Van der Graaff et al., 2018) and those that vary with age, including the social motivation to behave prosocially (Moffitt, 2018, 1993), may influence the development of both these behaviours, they are treated as unobserved in our analyses. The GCLM effectively manages these unobserved, time-varying and time-invariant (fixed) influences by statistically controlling for them in the model (Zyphur et al., 2019). This approach is chosen deliberately to focus on the developmental interactions between prosocial and conduct problem behaviour without the potential confounding effects of explicitly measured controls that could obscure the primary relationships under investigation. The absence of direct measures for such variables does not diminish the validity of our findings but rather highlights the challenge of disentangling relationships from their shared common factors. Our methodology, particularly the use of GCLM, compensates for this limitation by controlling for these unobserved effects, ensuring that our analysis remains robust. Furthermore, the IRF extends our insights by examining the long-term impacts and potential intervention points, focusing specifically on the temporal dynamics that are revealed through the GCLM.

Results

Preliminary Analyses

The General Cross-Lagged Panel Model (GCLM) used in this research is an extension of a standard longitudinal cross-lagged panel model. Therefore, the analytical strategy follows the standard staged approach for fitting and statistically selecting an appropriate cross-lagged panel model in a longitudinal structural equation modelling framework. The current research closely followed this analytic strategy using the Lavaan Package (Rosseel, 2012) in the R environment (R Core Team, 2019). The R code used in this study is from the supplementary materials provided by Zyphur and his colleagues (2019) and Shamsollahi et al. (2021). These supplementary materials provide extensive scripts and example outputs that allow for replication of the method.

Before employing this staged approach, it is important to confirm that data used in this study meet the assumptions of the GCLM. To do so, we explored the distributional properties of the data used in this study, by taking a small random sample. These analyses showed no issues in relation to multicollinearity but did show minor violations of the linearity and normality assumptions. Moreover, there was sample attrition at each wave (see the “Attrition” section for more information). However, these violations were not excessive in nature. This meant that combined with a large sample size (N = 2780) and the use of a maximum likelihood estimation procedure that is robust to non-normality and missingness (as recommended by Tabachnick & Fidell, 2001), no variable transformations were performed.

We then used a sequential approach to test whether the model benefits from the inclusion of the cross-lagged relationships between prosocial and conduct problem behaviour. These substantive checks included systematically constraining the cross-lagged paths and comparing the statistical fit of a cross-lagged panel model, with single-lagged autoregressive and cross-lagged terms of prosocial and conduct problem behaviour (Model 1), a unidirectional model with the effect of prosocial behaviour on conduct problem behaviour constrained to zero (Model 2), and a unidirectional model with the effect of conduct problem behaviour on prosocial behaviour constrained to zero (Model 3). Each model was compared using the Akaike Information Criterion (AIC) and adjusted Bayesian Information Criterion (aBIC) (as recommended by Zyphur et al., 2019). The results of these comparisons indicated that not including the relationship between prosocial and conduct problem behaviour (i.e. Models 2 and 3) resulted in a poorer fit than the full generalised cross-lagged panel model (Model 1) that included all these relationships (see Table 2). The improved fit of the full model (Model 1) over these unidirectional models (Models 2 and 3) suggests that there is a bidirectional relationship between prosocial and conduct problem behaviour.

Table 2 Substantive checks and fit comparisons for Models 1, 2, and 3
Full size table

The final step in the preliminary analysis is to statistically examine the model specification that will inform our results. To do this, we examined the fit of the model with different higher-order autoregressive and moving average terms (Zyphur et al., 2019). The purpose of this examination is to determine whether higher-order autoregressive and moving averages are required. A second order autoregressive term may control for the effect of early childhood change on middle childhood outcomes (a second-order lag from waves one to three) and middle childhood change on adolescent outcomes (a second-order lag from waves three to six). A second order moving average also allows for the temporary increase or decrease of these autoregressive effects in these periods to assess these effects. The requirement for these higher-order lags is assessed in a sequential manner, by comparing the model with single-lagged autoregressive terms and a single moving average term (Model 4; AR[1]MA[1]), with a single-lag autoregressive term and a second order moving average (Model 5; AR[1]MA[2]), and second order autoregressive terms with a moving average (Model 6; AR[2]MA[1]).

The statistics used for model specification are the Akaike Information Criterion (AIC) and the adjusted Bayesian Information Criterion (aBIC). Examination of these models revealed no difference in the AIC and aBIC between Models four and five and a reduced fit for Model six (see Table 3). Additionally, there was no substantial difference in the strength or direction of the coefficients across all three models. Therefore, Model four (AR[1]MA[1]) was chosen as the most parsimonious model that also provided the most information about the relationship between prosocial and conduct problem behaviour across the entire developmental period. The final model was refit with bootstrapped standard errors and 95% confidence intervals to provide a robust measure of the indirect effects. The fitted model is illustrated in Fig. 1 and will be described in more detail in the following sections.

Table 3 Fit comparisons of the GCLMs to determine the need for higher order lags and moving average terms and for final model selection
Full size table
Fig. 1
figure 1

Fitted generalised cross-lagged panel model of prosocial and conduct problem behaviour Note: µ(χγ)it = impulse. Coefficients are unstandardised, except for covariances. *** = significant at p < 0.001. a = Fixed effect covariances; b = Time-varying correlations; c = Autoregressive paths; d = Moving Averages; e = Sum of the cross-lagged and cross-lagged moving average (CLMA + MA). All other paths omitted for c

Full size image

The Relationship Between Prosocial and Conduct Problem Behaviour

The first aim of this study was to explore the relationship between prosocial behaviour and conduct problem behaviour. The general cross-lagged panel model (GLCM) enables this through the evaluation of both time-varying and time-invariant effects. Time-varying effects quantify the interplay of prosocial and conduct problem behaviours at each wave, indicated by correlations that reflect the proportion of shared variance between these behaviours. As shown in Table 4, these correlations range between − 0.26 and − 0.40. This suggests that between 6 and 16% of the variance in prosocial behaviour is explained by its relationship with conduct problem behaviour across the different developmental stages and vice versa. Notably, this relationship is consistently negative within each wave, indicating that higher levels of prosocial behaviour are associated with lower levels of conduct problem behaviour. However, there is some indication that this relationship is weaker in children aged 6–7 (β =  − 0.26), with results in Table 4, suggesting that only 6% of the variance is accounted for in this age group. In addition to these time-varying relationships, the covariance of the fixed effects can be examined to determine the relationship from age 4 to 15. The fixed effect covariance of β =  − 0.32, shown in Table 4, indicates a persistent negative relationship from age 4 to 15 accounting for 10% of the variance. This indicates a stable and negative relationship between prosocial and conduct problem behaviours at and across all time points in this study.

Table 4 The variance and covariance of the relationship between prosocial and conduct problem behaviour in each wave and across time
Full size table

Prosocial Behaviour Predicting Conduct Problem Behaviour

To assess whether prosocial behaviour contributes to the deceleration of conduct problem behaviour, we explored whether increases in prosocial behaviour precede decreases in conduct problem behaviour, and whether these effects accumulate over time. To address this aim, we examined combined cross-lagged and moving average (CLMA + MA) effects, which indicate the strength of influence from one wave to the subsequent one. The analysis, as detailed in Table 5, indicates a modest but statistically significant relationship between prosocial behaviour in one wave and conduct problem in the next. An increase in prosocial behaviour by one unit in any given wave corresponds to a subsequent decrease in conduct problem behaviour in the next wave, as shown by the CL + CLMA in Table 5 (CL + CLMA: B =  − 0.07(0.01), 95% CI [− 0.10, − 0.05]). Further examination of the autoregressive (AR) paths for conduct problem behaviour sheds light on the longitudinal impact of prosocial behaviour on conduct problem behaviour. Here, the results suggest a persistent influence of conduct problem behaviour from one period to the next, indicating that 89% (AR: B = 0.89 (0.09); 95% CI [0.68–0.93]) of the change in conduct problem behaviour as a result of change in prosocial behaviour is likely to carry forward to the next wave. However, the dampening effect of the moving average (MA) term is notable, indicating a 34% reduction (B =  − 0.55; 95% CI [− 0.61, − 0.48]) in this influence in the subsequent period.

Table 5 Statistics for the autoregressive and moving average paths of conduct problem behaviour and the effect of prosocial behaviour on conduct problem behaviour
Full size table

Examination of the impulse response function (IRF), illustrated in Fig. 2, allows for a further understanding of both the immediate and cumulative effects of prosocial behaviour on conduct problem behaviour. We examined whether these effects accumulate, and if so, whether they consistently drive down conduct problem behaviour. To investigate this, we visually examined the cumulative impulse response function (IRF). The results, non-cumulative illustrated in Fig. 2, indicate that there is limited impact of prosocial behaviour on conduct problem behaviour from one time to the next. However, cumulative in Fig. 2 clearly demonstrates that increases in prosocial behaviour correspond to systematic reductions in conduct problem behaviour in the following period, with these impacts accumulating to assist in the gradual deceleration of conduct problem behaviour over time.

Fig. 2
figure 2

The non-cumulative and cumulative impulse response function: The influence of prosocial behaviour on conduct problem behaviour. Note : “shocked” refers to the impulse which measures how a change at each period carries through to the next wave ( 1 ) and subsequent waves ( 2 )

Full size image

Conduct Problem Behaviour Predicting Prosocial Behaviour

To assess whether prosocial behaviour is a measurable outcome of the deceleration of conduct problem behaviour, we explored if decreases in conduct problem behaviour precede increases in prosocial behaviour and if these effects accumulate over time. This directional relationship was assessed by analysing the combined cross-lagged and lagged moving average (CL + CLMA) effects, as displayed in Table 6. The results reveal that when conduct problem behaviour decreases by one unit in a given wave, there is a corresponding 0.13 increase in prosocial behaviour in the following wave (B =  − 0.13; 95% CI [− 0.17, − 0.08]), signifying a modest yet statistically significant increase.

Table 6 Statistics for the autoregressive and moving average paths of prosocial behaviour and the effect of conduct problem behaviour on prosocial behaviour
Full size table

Further investigation into the strength of this effect was conducted by examining the autoregressive effects (AR), which showed a moderate yet significant effect of prosocial behaviour on conduct problem behaviour across the study period (B = 0.73; 95% CI [0.61, 0.92]). However, upon reviewing the moving average (MA) term (B =  − 0.41; 95% CI [− 0.54, − 0.32]), the results reveal that 32% of the impact of reduced conduct problem behaviour on prosocial behaviour would extend into the following wave, indicating a loss of effects over time.

Examination of the impulse response function (IRF), illustrated in Fig. 3, allows for a further understanding of both the immediate and cumulative effects of decreased conduct problem behaviour on prosocial behaviour. As shown in non-cumulative in Fig. 3, there is a direct association between decreases in conduct problem behaviour and subsequent increases in prosocial behaviour. In contrast, cumulative in Fig. 3 presents the cumulative analysis, which suggests that despite the initial positive trends, prosocial behaviour tends to revert to its baseline over time, regardless of the consistent negative relationship observed between conduct problem and prosocial behaviours.

Fig. 3
figure 3

The non-cumulative and cumulative impulse response function: the influence of conduct problem behaviour on prosocial behaviour.  Note : “shocked” refers to the impulse which measures how a change at each period carries through to the next wave ( 1 ) and subsequent waves ( 2 )

Full size image

Discussion: The Role of Prosocial Behaviour in the Deceleration of Conduct Problem Behaviour

This study explored the interplay between prosocial and conduct problem behaviour within a representative sample of Australian children aged 4–15. We investigated the reciprocal longitudinal relationship between these behaviours over an extended time period, spanning from early childhood to middle adolescence. Our aim was to determine whether prosocial behaviour is a facilitator for, or a measurable outcome of, the deceleration in conduct problem behaviour. Our findings indicate that prosocial behaviour is likely a facilitator, rather than a measurable outcome of the deceleration of conduct problem behaviour. This is supported by key findings that indicate that although decreases in conduct problem behaviour in one wave correspond to increases in prosocial behaviour in the next wave, these effects do not persist over time.

The Relationship Between Prosocial and Conduct Problem Behaviour

Our examination of the relationship between prosocial and conduct problem behaviours reveals a negative relationship between these behaviours. This negative relationship was observed in all waves in our research. These patterns are consistent with previous research that suggests a protective effect of prosocial behaviour against conduct problem behaviour (Carlo et al., 2014; Padilla-Walker et al., 2015; Speyer et al., 2023b). Similarly, we observed that decreases in conduct behaviour in one wave correlated to increased prosocial behaviour in the next. These findings provide initial support for Loeber et al.’s (2016) hypothesis that decreases in conduct problem behaviour may result in more exposure to prosocial behaviour and subsequently increase a child or adolescent’s repertoire of prosocial behaviour. An alternative perspective is that increases in conduct problem behaviour in one period may relate to decreases in prosocial behaviour in the next. However, as we will discuss in the next section, these decreases do not accumulate over time and, therefore, may not have long-term impacts on prosocial behaviour.

The relatively long developmental period used in the current study also allowed us to explore whether the negative relationship between prosocial and conduct problem behaviour holds up to and across multiple critical developmental periods (early childhood, middle childhood, early adolescence, and middle adolescence). Our results indicate that this relationship holds across these critical periods and that this relationship is stable over time. These findings are consistent with those of Memmott-Elison and Toseeb (2023) who found that the negative relationship between prosocial and conduct problem behaviour is consistent from one time point to the next from age 5 to 14. However, our results contrast with those of Speyer et al., (2023b) that showed variability in this relationship from early to middle adolescence. It is important to acknowledge that Speyer et al., (2023b) used a different operationalisation of problem behaviour; they distinguished between bullying and aggression. Therefore, future studies should seek to extend on these findings to understand the unique interactions between different types of prosocial and conduct problem behaviour across multiple critical developmental periods.

The Role of Prosocial Behaviour in the Deceleration of Conduct Problem Behaviour

Our examination also revealed that increased prosocial behaviour in one period not only correlated with decreased conduct problem behaviour in the next, but these effects also accumulated across the entire developmental period. These findings support Loeber et al.’s (2016) other key postulate that prosocial behaviour contributes to the deceleration of conduct problem behaviour. Previous research has demonstrated the positive effects of interventions on problem behaviour, especially when delivered to children at various developmental stages (Mesurado et al., 2019). Our findings suggest that providing children and adolescents with consistent opportunities to engage in prosocial behaviour, across the developmental period spanning early childhood to middle adolescence, may contribute to the deceleration of conduct problem behaviour. Thus, our research provides context to this previous research by indicating that these programs not only should be targeted to children across different developmental periods but should also be delivered early and often in order to prevent persistent conduct problem behaviour.

Conversely, while decreases in conduct problem behaviour during one period are correlated with increases in prosocial behaviour in the subsequent period, these increases do not accumulate. Instead, prosocial behaviour gradually reverts to its initial baseline levels. Thus, this does not support the second part of Loeber et al.’s. (2016) hypothesis that increases in prosocial behaviour are a measurable outcome of the deceleration of conduct problem behaviours. One explanation for these results is that the engagement in conduct problem behaviour may not have long-lasting impacts on a child’s or adolescent’s ability to learn or regain the skills and opportunities necessary to engage in prosocial behaviour. Indeed, Moffitt (2018) argues that those who engage in adolescence-limited antisocial behaviour are able to desist because they have acquired prosocial behaviours in childhood. Our findings indicate that the skills required to engage in prosocial behaviour develop early in life and remain stable, but whether the child or adolescent engages in these behaviours may be subject to their situational motivation to act prosocially. This is congruent with previous literature that suggests engagement in prosocial behaviour is relationally and situationally driven, with children and adolescents choosing to act more or less prosocially, especially in the company of peers (Loeber et al., 2016; Masten & Tellegen, 2012; McGee et al., 2015; Moffitt, 2018, 1993). Future studies should more closely examine these relational and situational influences on these behaviours.

Theoretical Implications

The Social Development Model (SDM) posits that prosocial and conduct problem behaviours are related yet distinct constructs (Catalano et al., 2021). Our findings reveal a consistent negative relationship between these behaviours, which might suggest they are opposites. However, by controlling for unobserved variance, we are able to clarify that this is not necessarily the case. When this unobserved variance is statistically controlled, each behaviour only moderately explains the variance in the other. This indicates that the interaction between prosocial and conduct problem behaviours is more complex, involving both shared and unique factors. Thus, rather than being opposites on a continuum, our results support the SDM’s postulate that these are separate constructs that are influenced by both overlap** and independent factors.

Importantly, our findings suggest new avenues for empirical tests of the SDM. Future research could explore how specific relational and situational mechanisms, such as the quality of social bonds or the availability of prosocial opportunities (Catalano et al., 2021), can bolster the effect that prosocial behaviour has on the deceleration of conduct problems. Additionally, designing studies that can identify what the unique factors are in our study—those unobserved factors that do not overlap between prosocial and conduct problem behaviours—could reveal more about the individual pathways through which interventions could be tailored. This would not only provide further evidence of the SDM’s postulates but also refine our understanding of how and when these mechanisms exert their influence over the relationship between prosocial and conduct problem behaviour.

Practical Implications

Our research also has practical implications, providing a starting point for the consideration of how prosocial behaviour interventions may work to reduce conduct problem behaviour. Our results suggest that interventions that aim to foster prosocial behaviour, delivered early and often, are likely to reduce conduct problem behaviour. These interventions need not specifically target children and adolescents presenting with conduct problem behaviour because increases in prosocial behaviour precede decreases in conduct problem behaviour. This offers the opportunity for develo** universal interventions that focus on develo** prosocial behaviour rather than preventing conduct problem behaviour.

There may still be the need for targeted interventions that address high levels of conduct problem in adolescence. Although the findings of our research suggest that prosocial behaviour interventions should work to reduce conduct problem behaviour, they also indicate that the success of these interventions may rely on an understanding of the patterns of conduct problem behaviour of the participants. Our findings suggest that prosocial behaviour interventions designed for adolescents and children already engaging in high conduct problem behaviour may not have a strong impact on their levels of conduct problem behaviour. This finding may explain why many interventions aimed at increasing prosocial behaviour do not always benefit children or adolescents with existing high levels of conduct problem behaviour (see Beelmann & Lösel, 2020 and Ciocanel, et al., 2017 for reviews of these types of interventions). This may be overcome by getting ahead of the conduct problem behaviour by delivering interventions early and repeating these interventions to gain the benefit of the accumulated effects of changes in prosocial behaviour on conduct problem behaviour.

Strengths, Limitations, and Future Direction

Our research presents some distinct strengths over previous investigations into the relationship between prosocial behaviour and conduct problem behaviour. Besides being one of the few studies to explore this relationship longitudinally, it takes a novel approach to understand the relationship over time. The use of the GCLM enables for the controlling of unobserved time-varying and time-invariant effects. Controlling for these effects is crucial for separating them from shared factors that influence levels of both prosocial behaviour and conduct problem behaviour (Zyphur et al., 2019). Further, the addition of the examination of the impulse response function (IRF) allows for the exploration of the long-term impacts of change in one behaviour on the other (Shamsollahi et al., 2021), something that has not been addressed in previous studies. As a result, our research offers a robust insight into how prosocial behaviour might relate to the deceleration of conduct problem behaviour. Further to this, our study uses a representative sample of Australian children and is thus can be generalised to this population of children. Finally, the time period under investigation spans a broad range of critical developmental periods, allowing for an understanding of the persistence of effects across these developmental periods.

However, our research, despite its strengths, possesses certain limitations that suggest directions for future studies. First, while our study provides a robust method of examining the role of prosocial behaviour on the deceleration of conduct problem behaviour, it does not confirm the termination of this behaviour. Second, while it is essential to control for shared time-varying and time-invariant factors to understand how changes in prosocial behaviour affect conduct problem behaviour, the GCLM does not allow for an understanding of what these shared factors are. This may obscure the heterogeneity in the relationship between prosocial and conduct problem behaviour, specifically in relation to factors such as gender (Pursell et al., 2008; Van der Graaff et al., 2018) that may drive similarities and differences in both the relationship between the behaviours and the factors that influence these behaviours—including the social motivation to behave prosocially (Moffitt, 2018, 1993). Examining these shared factors more closely could deepen our understanding of the specific factors impacting both behavioural outcomes. This could, in turn, guide more targeted intervention strategies. Third, although we used a well-validated and stable measure of prosocial and conduct problem behaviours (the SDQ), more nuanced measures may be required to fully capture age-graded manifestations of these behaviours or different outcomes related to different forms of prosocial and conduct problem behaviour. Finally, it is vital to evaluate how prosocial and conduct problem behaviours are related within the specific contexts that children and adolescents are embedded. The Social Development Model (Catalano et al., 2021) posits that encouraging prosocial behaviour may be difficult when children or adolescents lack access to environments fostering opportunities for such behaviours.

The current study’s limitations pave the way for further research into the role of prosocial behaviour in the deceleration of conduct problem behaviour. Future studies might more closely explore the desistance process, including the relationship of specific prosocial behaviours (e.g. sharing) to specific conduct problem behaviours (e.g. stealing). To understand the role of prosocial behaviour in the deceleration of conduct problem behaviour more comprehensively, researchers could expand their explorations to compare children and adolescents who either progressed or did not progress to offending behaviour. Lastly, investigating the factors influencing both prosocial and conduct problem behaviours could offer insights into whether interventions centred around prosocial behaviour can enhance outcomes for children and adolescents engaging in frequent and serious conduct problem behaviour.

Our study provides robust evidence that prosocial behaviour plays an important role in facilitating the deceleration of conduct problem behaviour across developmental stages from early childhood to middle adolescence. Our findings indicate that while increases in prosocial behaviour follow decreases in conduct problem behaviour, these positive changes are not long-lasting. This finding suggests that conduct problem behaviour may not have long-lasting implications on a child or adolescent’s ability to engage in prosocial behaviour. It also highlights that early and repeated prosocial interventions can facilitate the deceleration of conduct problem behaviour. Thus, paving the way for more focus on develo** positive behaviours, rather than focusing on deficits. Additionally, the stable yet complex relationship between these behaviours across various developmental periods calls for more nuanced research that considers different types of conduct problems and the distinct ways they may influence prosocial behaviour. Future research should aim to explore these dynamics further, exploring how different intervention strategies can be tailored to maintain and enhance prosocial behaviour effectively over time. Our study contributes to the understanding of the interplay between prosocial and conduct problem behaviour from a developmental and life-course criminology perspective and to the evidence base for intervention models that foster positive developmental trajectories in children and adolescents.