Why Monitoring Doesn’t Always Matter: The Interaction of Personal Propensity with Physical and Psychological Parental Presence in a Situational Explanation of Adolescent Offending

ABSTRACT

Parental monitoring is often shown to have a negative relationship with crime involvement. However, research often ignores both the mechanism by which these relationships occur and the conditions under which they might (and might not) be found. Building on the interactional hypotheses of Situational Action Theory (SAT) and the parental monitoring definitions provided by a new model of Goal-Directed Parental Action (GDPA), this paper assesses evidence of the role of the presence or absence of guardians, adolescent-perceived parental knowledge and personal crime propensity in explaining why a crime does or doesn’t occur in a particular situation. To test these new hypotheses, this paper uses specialist data (including innovative Space-Time Budget data) from the Peterborough Adolescent and Young Adult Development Study (PADS+) and employs multiple methods to assess both statistical interaction (dependence) and situational interaction (convergence).

Findings provide evidence for interaction between key personal characteristics (specifically, generalized law relevant morality and ability to exercise self control) and features of environments (in this study, aspects of parental monitoring) in the perception-choice process that is proposed by SAT to explain action, including acts of crime. They suggest that despite high correlations with crime in the literature, monitoring doesn’t always matter because many crime averse young people do not offend anyway. Furthermore, in an entirely new finding, generalized perceived parental knowledge of the circumstances of activity (psychological presence) results in a lower rate of crime among crime prone adolescents when they are unsupervised.

Acknowledgments

Thanks to P-O Wikström for approval to publish this analysis of the PADS+ data, and to Jenni Barton-Crosby for comments on a draft of this paper.

Disclosure statement

No financial interest or benefit has arisen from the direct applications of this research.

Notes

¹ This study uses the term ‘parent’ to refer to biological and birth parents as well as those appointed to or adopting the parental role, e.g., foster or adoptive parents and other legal guardians. Therefore, the term parental monitoring includes the monitoring of wards and minors by parents or guardians appointed to or adopting the parental role.

² Such an analytical approach aims to advance, through theory development and empirical testing, a general, dynamic and mechanism-based explanation of crime and its causes to inform effective crime prevention policies and practices (see further, Pauwels, Ponsaers, and Svensson 2009; Wikström 2006, 2017; Wikström and Treiber 2017).

³ Laird, Marrero, and Sentse (2010) acknowledge that parental monitoring might not always be crucial to offending, and only relevant under certain circumstances; but fail to specify developmental and situational roles for parental monitoring and do not distinguish between contexts of development and contexts of action. As a result, this study cannot truly address the mechanisms of person-environment interaction.

⁴ At best, studies rigorously examine multiple factors that mediate relationships between parental monitoring and crime (e.g., Janssen 2015); however, it is important to emphasize that mediating factors are not mechanisms (Elster 2007; Hedström and Ylikoski 2010), and that testing the relative contribution of factors (representing for example, individual and environmental influences) assesses only additive and not interactive processes (Hardie forthcoming,c; Wikström and Treiber 2017).

⁵ Some studies hold initial promise for the study of the role of parental monitoring in interactive situational processes (e.g., Ishoy 2017; Janssen, Weerman, and Eichelsheim 2017; Mann et al. 2015); however, they suffer from one or more problems, ultimately due to the lack of a suitable guiding framework. Levels of explanation are conflated, processes are mis-specified, definitions of key variables are unspecific and therefore concepts are poorly operationalized (Hardie forthcoming,b).

⁶ By highlighting action and motivation, the model of GDPA also distinguishes active parenting behaviors from other more passive sources of parental knowledge (Hardie forthcoming,a). The model of GDPA is therefore consistent with the principles behind the seminal redefinition of parental monitoring by Stattin and Kerr (Kerr and Stattin 2000; Stattin and Kerr 2000); however, Hardie’s definition of parental monitoring is more specific, and the model of GDPA specifies additional related concepts. Other very recent attempts to clarify parental monitoring-related concepts fail to adequately delineate such elements (e.g., Flanagan, Auty, and Farrington 2019).

⁷ For SAT, a ‘situation’ arises from a particular person in a particular context, so studying situational processes inherently involves attention to both individuals and environments, and their interaction (Hardie forthcoming,c; Wikström and Treiber 2016).

⁸ A setting is defined as the features of the environment (objects, people, events) that an individual can access with their senses, which are the only features capable of influencing a persons’ actions (Wikström 2006).

⁹ Warr (1993, 2002) extends Hirschi’s (1969) psychological presence concept and aims to explain how adolescents are still influenced to conform by their parents even in the face of strong peer pressure to offend, but ultimately also fails in this regard (Hardie 2017:170).

¹⁰ Parents often trust some nominally, formally or informally appointed adults as their proxy, to bridge the gap between parental presence and being unsupervised (Hardie forthcoming,a), therefore the presence of guardians other than parents is also studied in this paper. The effect of parental presence is hypothesized to be stronger than that of other guardians because parental investment in the child’s pro-social outcomes is higher, and stronger bonds with parents increase the impact of their presence on the adolescent’s perception of the moral context.

¹¹ Only 7 participants were not retained for the full four years. Data for these participants is excluded from analyses at the individual level.

¹² Annual scale alphas were .89, .90, .88 and .89.

¹³ Annual scale alphas were .78, .76, .78 and .77.

¹⁴ Annual scale alphas were .77, .83, .82 and .84.

¹⁵ For a review of problems and the benefits and drawbacks of solutions, see, Hardie (Hardie forthcoming,c), see also, e.g, McClelland and Judd (1993); Svensson and Oberwittler (2010).

¹⁶ Models were estimated using the robust linear models (RLM) macro for SPSS (coded by Darlington and Hayes 2017). Results presented use the HC3 method but results for all five methods were substantively the same.

¹⁷ When categorized independent variables are intersected, the resultant groups of participants vary in size. This is due to the relationships between these variables, which are likely driven by processes of selection and emergence that are not studied here (see further, Hardie 2017:266–67). Any evidence of the situational process must remain consistent irrespective of group size because SAT states that the same cause (situational process of perception and choice) influences action regardless of the causes of the cause (i.e., social processes such as selection and emergence).

¹⁸ This descriptive comparison method does not allow for statistical control of covariates, however, as discussed above, this is not necessarily problematic.

¹⁹ Ratios were not calculated for rate comparisons to zero. This is a particular difficulty with rare phenomena such as crime, but is not easily solve because the convention of adding 0.5 to each cell for calculating risk ratios involving zeros (e.g., Deeks and Higgins 2010; Pagano and Gauvreau 2000) is not well suited to this application (Hardie forthcoming,c).

²⁰ More attention should be paid to the effect size (RR) than the significance (z-score and associated p value) because the application of significance testing is somewhat arbitrary and in this specific case arguably inappropriate (Hardie forthcoming,c).

²¹ An initial model predicting unlogged crime frequency that included the three predictors and their quadratic terms (also estimated using heteroscedasticity-consistent robust standard errors) showed that only the quadratic term of the crime propensity variable was significant (model not shown).

²² It is unlikely to be necessary to include quadratic terms in models predicting a logged outcome, and in fact, it may even result in over-controlling. However, the quadratic term for propensity was actually slightly significant (although small) when predicting logged crime frequency. For this reason, models 1 and 2 in Table 1 were also run including the quadratic term for propensity. Results were almost identical to those without the quadratic term (models not shown).

²³ For a detailed discussion, see, Hardie (Hardie forthcoming,c). Not only do results become difficult to interpret, but notoriously ‘shy’ interaction effects are often further weakened by the removal of vital variation via dependent variable transformations (Hannon and Knapp 2003; Svensson and Oberwittler 2010).

²⁴ Some argue that it is possible to remove non-significant two-way interaction terms from the model (Aiken and West 1991). For models 2 and 4 in Table 1, the three-way interaction term retained the same significance and had very slightly weaker effect when the non-significant two-way interaction terms were removed from the model, and the total explained variance was almost unchanged (models not shown).

²⁵ PPPKxPROP r = .65; UNSUPxPPPK r_s = .45; UNSUPxPROP – r_s = .49. All correlations p = .000.

²⁶ The same rate difference is also not significant for medium propensity hours (0.5 and 0.7; rr = 1.23; z = 0.39; p = 0.70).

²⁷ For example, if a person were never unsupervised, their perceived level of parental knowledge of their unsupervised activity would be meaningless to an explanation of the role of perceived parental knowledge in the criminogeneity of a setting (though it may retain meaning as a marker for factors potentially relevant to social, developmental processes).

²⁸ The criminogenic convergence of crime propensity and the absence of guardians at the hour (event) level affords only conclusions at the hour level, so it is not strictly accurate to conclude about individuals from situation-level data like this. However, this situational finding can be further interpreted in the light of the individual level findings. The analysis of situational data crucially increases the specificity of conclusions compared to that of non-situational, individual-level data; but the most similar finding at the individual level affords the conclusion that ‘the amount of supervision a person experiences is relevant to the amount of crime they commit, but only for certain kinds of people’. Thus, it is not too far a stretch to conclude about situations and individuals from these combined findings (see further, Hardie forthcoming,c).

²⁹ Osgood et al. (1996) actually “do not assume that everyone is equally receptive to the temptations of situations conducive to deviance” (p. 639), but “neither do we [they] assume that exposure to them is relevant only to a small group of “motivated offenders”” (p. 639). Augustyn and McGloin (2013) observe that despite this original statement by Osgood et al, “how and why the degree of susceptibility to temptations offered by informal socializing varies has received minimal attention.” (p. 118).

³⁰ Bernasco et al. (2013), also assess the relative risk of individuals offending when they are under particular setting and environmental conditions including the absence of guardians; however they do not test individual differences, thus ignoring existing specific empirical evidence that such probabilities vary dramatically according to features of the individuals exposed to those settings.

³¹ This may involve the development of new methods, or adaptation of the STB to include hour-by-hour measures of parental monitoring features (for examples, see, Hardie 2017:319).

³² SAT argues that a generalized measure of individual crime propensity does not vary across situations and will be generally related to the probability that the individual will perceive and choose to carry out an act crime in response to a specific criminogenic setting (Wikström et al. 2012:132). Whilst generalized crime propensity might be expected to be relatively stable across situations, it does change gradually from year to year (Wikström, Treiber, and Roman forthcoming). For this reason, this study applies an annual individual-level measure of generalized propensity to each hour only within that year.

³³ Hirschi also suggested that monitoring is only conditionally relevant: “The child supervises himself … Monitoring works for those low on self control. It is unnecessary for those high on self control” (Travis Hirschi, speaking to John Laub; Laub 2011). However, Hirschi over-simplifies the mechanism that links psychological presence with the behavioral outcome by conflating virtual supervision (i.e., psychological presence) and self control. In addition, he also ignores the role of personal morality and moral rules of contexts in guiding behavior, and that controls are only conditionally relevant (see above). Laird, Marrero, and Sentse (2010) also suggest that monitoring may not always be needed for pro-social behavior outcomes, but they posit a different moderating factor and explicitly reject discussion of a causal mechanism.

³⁴ Various smartphone applications allow parents to track their children using real time GPS, render adolescents unable to ignore their parents messages, or both; for example ‘Respond ASAP’ (Herbert 2017), ‘Footprints’ (Sollico Software 2017), and ‘Family Tracker’ (Logsat Software 2019).

Additional information

Funding

The author of this work is grateful to the University of Cambridge Domestic Research Scholarship and Churchill College, Cambridge, for their financial support; and to the UK Economic and Social Research Council which funded PADS+ during the period of data collection relevant for this article (Grant number L330253002).

Notes on contributors

Beth Hardie

Beth Hardie is a Research Associate and member of the Center for Analytic Criminology at the Institute of Criminology, University of Cambridge, where she is also Research Manager of the Peterborough Adolescent and Young Adult Development Study (PADS+). She holds a BA in Social and Political Sciences and a PhD in Criminology from the University of Cambridge.

Her research is grounded in a theory-driven analytical approach that integrates individually and environmentally focused explanations of human behavior, including crime. She is interested in innovative ecological research methods such as the Space-Time Budget, analytical techniques for assessing situational interaction, developing and testing Situational Action Theory (including cross-national replication), and parental monitoring.

Empirical publications include the book “Breaking Rules: the social and situational dynamics of young people’s urban crime” (with Per-Olof Wikström, Dietrich Oberwittler and Kyle Treiber, published by Oxford University Press), “Activity fields and the dynamics of crime: advancing knowledge about the role of the environment in crime causation” (with Per-Olof Wikström, Vania Ceccato and Kyle Treiber, published in the Journal of Quantitative Criminology), “Young people’s differential vulnerability to criminogenic exposure” (with Per-Olof Wikström and Richard Mann, published in the European Journal of Criminology), and “Studying person-environment interaction: Clarifying appropriate analysis of situational interaction to explain behaviour” a forthcoming SpringerBrief volume.