Comparing the characteristics of firesetting and non-firesetting adults in a New Zealand un-apprehended community sample

Abstract

This study examines the prevalence of deliberate firesetting in an un-apprehended New Zealand sample and factors that differentiate firesetting and non-firesetting individuals. An anonymous online survey assessing demographic and background characteristics, history of firesetting and psychological characteristics was advertised on Facebook community groups and pinboards across New Zealand. A total of 626 adults completed the survey. One-hundred and three participants (16.45%) self-reported having deliberately set at least one fire over the age of 14 years. When controlling for all other variables in a logistic regression model, having a high-school qualification as the highest level of educational attainment (odds ratio, OR = 2.24), history of a mental health diagnosis (OR = 1.91), fearful attachment style (OR = 1.16) and fire interest (OR = 1.05) showed the strongest unique association with un-apprehended firesetting. Further research is needed to examine the full range of behavioural and psychological characteristics associated with un-apprehended firesetting.

Deliberate firesetting is a pervasive and harmful behaviour, which has significant impacts on the environment, economy and human life (Tyler et al., 2019). With changing climate conditions, there has been an increase in extreme weather patterns across the world, including high temperatures, low rainfall and extended dry periods, creating environmental conditions that promote increased fire occurrence and severity (Jolly et al., 2015; Jones et al., 2020; Stanley et al., 2020). Such conditions elevate the risks associated with and potential impacts of deliberately started fires in both urban and rural settings, making primary prevention of deliberate firesetting an increasing priority.

Estimating the number of individuals responsible for deliberate fires is incredibly difficult as, relative to other crimes, arson has one of the poorest detection and prosecution rates (Smith et al., 2013). UK statistics suggest that only 5.7% of all recorded deliberate fires result in a prosecution (Arson Prevention Forum, 2017). In the Australian State of Victoria in 2016, 62.6% of arson cases remained unsolved, with only 19.2% resulting in arrest and 10.6% resulting in a summons to court (Crime Statistics Agency, 2021). As a result, most perpetrators remain unidentified and un-apprehended within the community.

Given deliberate firesetting is a relatively common offence that has the potential to cause great harm, a comprehensive understanding of this behaviour is vital to effectively manage and reduce its impacts. However, despite this, deliberate firesetting is one of the most under-researched and poorly understood criminal behaviours (Davis & Lauber, 1999). Moreover, most of our knowledge about deliberate firesetting comes from research with apprehended populations, such as adults in psychiatric care (e.g. Coid et al., 2001), individuals with a recorded arson offence (e.g. Ducat, McEwan, et al., 2013), or youth referred to specialist intervention programmes (e.g. Lambie et al., 2013). However, there has been very little focus on those who set fires but remain un-apprehended, despite most perpetrators evading detection.

Prevalence of un-apprehended firesetting

To date, there are only a handful of studies that have examined the prevalence of firesetting in un-apprehended (i.e. community) samples. Wave 1 of the National Epidemiological Survey of Alcohol and Related Conditions (NESARC; Blanco et al., 2010; Hoertel et al., 2011; Vaughn et al., 2010), a nationally representative face-to-face survey of around 43,000 adults residing in the United States, included a single question in the antisocial personality disorder module, which asked participants if they had ever intentionally set a fire to destroy another’s property or to see something burn. Secondary analysis of the NESARC found that between 1.0% and 1.3% of adults in the US general population reported lifetime engagement in firesetting (Blanco et al., 2010; Hoertel et al., 2011; Vaughn et al., 2010). However, the NESARC methodology may under-estimate the true nature of un-apprehended firesetting due to reluctance to disclose illegal behaviour through in-person questioning for fear of consequences (Dickens & Sugarman, 2012; Gannon & Barrowcliffe, 2012). The use of a single item also inhibited collection of information on the motivations, frequency and age at which the firesetting occurred (outside of above or below 15 years; Gannon & Barrowcliffe, 2012): factors that are known to influence the prevalence of self-reported firesetting (Mackay et al., 2012).

To improve on the NESARC studies, Gannon and Barrowcliffe (2012) developed a refined methodology for assessing firesetting in the general population, improving the specificity and operationalisation of firesetting and enabling anonymous responding. Individuals who indicated a history of deliberate firesetting were also asked to elaborate on the motive, frequency and ignition method used, to obtain a more detailed account of the behaviour. Using this improved methodology, the pooled prevalence of deliberate firesetting across three UK community studies (Barrowcliffe & Gannon, 2015, 2016; Gannon & Barrowcliffe, 2012) suggests that approximately 14% of UK adults self-report having set at least one deliberate fire since the age of 10 years (Gannon et al., 2022). These statistics indicate that deliberate firesetting may be more prevalent in the general population than previously thought. A more detailed understanding of un-apprehended firesetting is therefore needed to inform the development of effective primary prevention strategies, which can reduce the risk of deliberate firesetting at a general population level.

Characteristics of un-apprehended firesetting individuals

There have only been a handful of studies that have examined the characteristics of un-apprehended adults who set fires, relative to non-firesetting individuals, and no studies have directly compared apprehended and un-apprehended firesetting individuals. Although findings vary across studies, relative to non-firesetting community participants, un-apprehended firesetting individuals are typically reported to have a more extensive history of antisocial behaviour (Barrowcliffe & Gannon, 2016; Barrowcliffe et al., 2022; Blanco et al., 2010; Vaughn et al., 2010), are more likely to have experimented with fire before the age of 10 years (Barrowcliffe & Gannon, 2016; Barrowcliffe et al., 2022), and are more likely to have a family history of deliberate firesetting (Barrowcliffe & Gannon, 2015, 2016; Barrowcliffe et al., 2022) and/or antisocial behaviour (Blanco et al., 2010; Vaughn et al., 2010). In terms of psychopathology, compared to non-firesetting individuals, firesetting individuals are significantly more likely to have a diagnosis of a behavioural or mental health disorder (Barrowcliffe & Gannon, 2015, 2016; Blanco et al., 2010) and have a history of self-harm or suicide attempts (Barrowcliffe & Gannon, 2015, 2016; Blanco et al., 2010). Furthermore, un-apprehended firesetting individuals self-report significantly higher scores on measures of attitudes and interests towards fire, propensity to set fires, impulsivity, boredom proneness, anger provocation, psychopathy, physical sadism and antisocial attitudes (Barrowcliffe & Gannon, 2015, 2016; Barrowcliffe et al., 2019, 2022; Wehner et al., 2022).

While the emerging body of research on un-apprehended firesetting represents a significant advancement in our understanding of this behaviour, most studies are restricted by small samples of firesetting participants (≤60) and have been conducted exclusively in the UK and USA, limiting our understanding of un-apprehended firesetting across different socio-cultural contexts. Further, only three studies have examined psychological characteristics associated with un-apprehended firesetting, other than fire-related factors, using psychometric measures (e.g. Barrowcliffe & Gannon, 2016; Barrowcliffe et al., 2022; Wehner et al., 2022). More research is therefore needed to advance knowledge about this understudied group.

Theoretical explanations of un-apprehended firesetting

Theoretical explanations are critical for guiding empirical investigation of firesetting as well as for identifying potential risk factors for this behaviour and the development of effective prevention initiatives. However, existing theoretical explanations for firesetting are based entirely on research with apprehended individuals (Gannon et al., 2022). Therefore, it is currently unclear whether existing theory can adequately account for firesetting in un-apprehended individuals.

The most contemporary theoretical explanation of deliberate firesetting is the Multi-Trajectory Theory of Adult Firesetting (M-TTAF; Gannon et al., 2012, 2022). The M-TTAF is a two-tiered theory and provides an aetiological framework (Tier 1) that aims to explain the onset, maintenance and desistance of deliberate firesetting, as well as five theoretically nested subtypes or trajectories of adults who set fires (Tier 2). Tier 1 of the M-TTAF (Gannon et al., 2012, 2022) suggests that there are four clusters of psychological vulnerabilities that, when primed by proximal events/triggers, act as risk factors for firesetting: inappropriate fire interest and scripts, offence-supportive cognition, self/emotional regulation issues and relationship/communication problems. These psychological vulnerabilities develop through biological, developmental and environmental experiences. Mental health and self-esteem are proposed to moderate the relationship between psychological vulnerabilities and proximal factors, either buffering against or exacerbating vulnerabilities. Firesetting is hypothesised to be maintained through positive and negative reinforcers (internal and external), through the experience of post-firesetting affect and cognition. Desistance from firesetting is suggested to occur through cognitive transformations (e.g. increased self-regulation, problem solving and pro-social attitudes) as a product of intervention and/or identity change.

Research with apprehended samples suggests that the four domains of psychological vulnerability and moderating factors within the M-TTAF distinguish adults who have set fires from non-firesetting adults (Gannon et al., 2013; Wyatt et al., 2019). However, there has only been one study to date that has explicitly examined whether these domains of psychological vulnerability discriminate between un-apprehended firesetting and non-firesetting adults (Barrowcliffe & Gannon, 2016). While the broad pattern of findings from the research on un-apprehended firesetting suggests that this group may have psychological needs similar to apprehended firesetting individuals, Barrowcliffe and Gannon (2016) found that only three background factors (suspension from school, experimentation with fire before age 10 and having a family member who has set fires) predicted firesetting status in an un-apprehended sample, suggesting that there may be subtle differences between these groups (Gannon et al., 2022). Further research is therefore needed to understand whether the psychological domains within the M-TTAF distinguish un-apprehended firesetting individuals from non-firesetting individuals in the same way as they do in apprehended samples.

The current study

While recent research has started to develop our understanding of the prevalence of un-apprehended firesetting, and the demographic, background and offence behaviours of individuals who engage in this behaviour, the psychological characteristics of those who engage in un-apprehended firesetting remain poorly understood, particularly across different socio-cultural contexts. This study aims to extend understanding of firesetting behaviour by presenting the first examination of un-apprehended firesetting in New Zealand. Given the paucity of previous research and equivocal findings, the present study is exploratory in nature. We aimed to examine: (a) the prevalence of un-apprehended firesetting in New Zealand, (b) whether there are demographic, background and psychological characteristics (as outlined within the M-TTAF) that can differentiate un-apprehended firesetting individuals from non-firesetting individuals, and, (c) which combination of factors distinguishes firesetting from non-firesetting individuals.

Method

Participants

A community sample was recruited through advertisements on Facebook community groups across New Zealand as well as by posting on community and university pin boards in two major cities (Wellington and Christchurch). A semi-stratified approach to recruitment was adopted; study adverts were posted in Facebook community groups across both the North and South Island of New Zealand as well as in rural and urban areas. Regions where larger proportions of the population reside were targeted more heavily by posting in multiple community groups.

Individuals were eligible to participate in the study if they were aged 18 years or older and currently residing in New Zealand. Seven hundred and ninety-eight individuals accessed the online survey. Following data cleaning (i.e. removing longstring responses, spam, abnormal response times, surveys where three or more questionnaires were not completed), 626 people had completed the survey in full (resulting in a completion rate of 78.4%). The majority of participants identified as female (n = 538, 85.9%) and New Zealand European/Pākehā (n = 518, 82.7%). The average age of participants was 37 years (range 18–76 years). Most participants reported holding some form of formal education (n = 591, 94.4%), with a university degree (n = 204, 32.6%) or National Certificate of Educational Achievement (NCEA) Level 3¹ (n = 129, 20.6%) the most common levels of educational attainment. See Table 1 for a detailed overview. Our sample has a similar median age to that of the New Zealand general population (36 years vs. 37.4 years); however, it has a higher proportion of individuals who identify as female (85.9% vs. 50.4%) and New Zealand European/Pākehā (82.7% vs. 70.2%) and a higher proportion of individuals who have completed formal education (94.4% vs. 81.18%; Statistics New Zealand, 2018).

Table 1.

Demographic characteristics of total sample.

Demographic characteristics	n	%
Gender
Male	77	12.3
Female	538	85.9
Other	8	1.3
Prefer not to say	3	0.5
Age
18–24 years old	142	22.7
25–34 years old	144	23.0
35–44 years old	133	21.3
45–54 years old	123	19.7
55–64 years old	40	6.4
65 years or older	27	4.3
Not specified	17	2.7
Ethnicity
NZ European/Pākehā	518	82.7
Māori	81	12.9
Pasifika	17	2.7
European	54	8.6
African	4	0.6
Latin American	5	0.8
Asian	21	3.4
Middle Eastern	1	0.2
Other	18	2.9
Education
No formal education	35	5.6
NCEA Level 1	21	3.4
NCEA Level 2	42	6.7
NCEA Level 3	129	20.6
Vocational training	80	12.8
University degree	204	32.6
University degree with honours	52	8.3
Masters	53	8.5
PhD/doctorate	6	1
Not specified	4	0.6
Occupationa
Unemployed	157	25.1
Managers	19	3.0
Professionals	85	13.6
Technicians & trade workers	20	3.2
Community & personal service workers	27	4.3
Clerical & administrative workers	26	4.2
Sales workers	21	3.4
Machinery operators and drivers	4	0.6
Labourers	11	1.8
Not covered by labour force definition	9	1.4
Insufficient information	54	8.6
No answer but is employed	193	30.8
Offence history
No	584	93.3
Not specified	2	0.3
Yes	40	6.4
Conviction
Vandalism/antisocial behaviour	3	7.5
Violent offence	7	17.5
Drug or alcohol offence	14	35
Arson	0	0
Acquisitive offence	9	22.5
Property damage	6	15
Driving offence	18	45
Other	3	7.5
History of mental health diagnosis
Yes	262	41.9
No	364	58.1

Note: Total percentage for ethnicity adds up to more than 100% as some participants identified as more than one ethnicity. Percentages for conviction type are for those who have an offence history. Missing data are categorised as ‘not specified’. NCEA = National Certificate of Educational Achievement.

^a Specific occupations identified in the free response format were collapsed into the above categories based on the Australian and New Zealand Standard Classification of Occupations, First Edition, 2006 (Australian Bureau of Statistics, 2006).

Measures

Data were collected via an online questionnaire hosted in Qualtrics. The online questionnaire comprised three main sections: demographic and background characteristics, history of firesetting behaviour and measures assessing the M-TTAF domains of psychological vulnerability.

Demographic and background characteristics captured gender, ethnicity, age, highest level of education, occupational status, offence history and whether participants had ever been diagnosed with a mental health condition (yes/no).

History of firesetting behaviour asked participants to indicate whether they had ever deliberately set a fire by responding to a firesetting screening question adapted from Gannon and Barrowcliffe (2012). Participants were asked whether they had ever set a fire(s) to annoy other people, relieve boredom, to create excitement, for revenge, for insurance purposes, due to peer pressure or to destroy evidence. Participants were instructed to exclude fires set accidentally, set for organised or social events (e.g. bonfires or hāngī²), set with a fire permit or set before the age of 14 years. The original question was adapted to make it more relevant for use within the New Zealand context (e.g. inclusion of examples of culturally sanctioned fire use such as hāngī and fires set with a fire permit, and exclusion of ‘hog roast’). Participants were asked to indicate the number of fires they had set that met the above definition on a scale from 0 to 5+. Participants who indicated they had set one or more fires were then asked to provide information on these incidents through a series of closed questions including: age at first and most recent firesetting incident, motivations for their firesetting, if they had ever been caught setting a fire by someone in a position of authority or a whānau³/family member, and if they had ever received intervention for their firesetting.

The M-TTAF domains section comprised eight measures that tapped into each of the psychological vulnerabilities and moderating factors outlined in the M-TTAF (Gannon et al., 2012, 2022). Details of each measure are provided in Table 2. Internal reliability is reported according to George and Mallery’s (2003) criteria: $\ge .$90 excellent, $\ge$.80 good, $\ge$.70 acceptable and $\ge$.60 questionable. A measure of social desirability (the impression management subscale of the Balanced Inventory of Desirable Responding (BIDR); Paulhus, 1988) was originally intended to be included in the questionnaire battery; however, due to a fault with the Qualtrics software, the full range of response options was not presented. As a result, this measure was excluded from analysis.

Table 2.

M-TTAF domain measures.

Measure	Construct measured	M-TTAF Domain	N items	Item format and example	Reliability	Reliability in current study
FSS	Antisocial behaviour and fire interest	Inappropriate fire interest/scripts	20	Participants respond to items (e.g. ‘I am fascinated by fire’ or ‘I am a rule breaker’ on a 7-point Likert scale from 1= not at all like me to 7 = very strongly like me. Only the Fire Interest subscale was used in the current study (10 items.)	α = .85 (Gannon & Barrowcliffe, 2012)	(α = .92)
Previous exposure to fire	Previous exposure to fire	Inappropriate fire interest/scripts (learning- developmental context)	3	Participants respond to items (e.g. ‘Between 10 and 18 years of age, how often were you in the presence of fire(s)? E.g. log fires, bonfires, campfires’) on a 6-point Likert scale from 1 = never to 6 = daily or almost daily.	n/a	n/a
BIS	Impulsivity	Self/emotional regulation Issues	8	Participants respond to items (e.g. ‘I am a careful thinker’) on a 4-point Likert scale from 1 = rarely/never to 4 = almost always/always.	α = .73–.83 (Steinberg et al., 2013)	α = .81
RSES	Self-esteem	Self-esteem (Moderator)	10	Participants respond to items (e.g. ‘At times I think I am no good at all’) on a 4-point Likert scale from 1 = strongly disagree to 4 = strongly agree.	α = .91 (Robins et al., 2001)	α = .90
RQ	Attachment styles	Communication/relationships	4	Participants are required to read four paragraphs outlining four attachment styles (secure, preoccupied, fearful, dismissing). Participants indicate which style they believe is most applicable to them, and then provide a continuous rating of each style on a 7-point Likert scale (1 = it does not describe me at all, 7 = it very much describes me).	n/a	n/a
MSPSS	Social support	Communication/relationships	12	Participants respond to each item (e.g. ‘I can talk about my problems with my family’ on a 7-point Likert scale from 1 = very strongly disagree to 7 = very strongly agree.	α = .88 (Zimet et al., 1988)	α = .94
Criminal associates	Criminal attitudes and associates	Offense-supportive attitudes	1	Participants are requested to either agree (1) or disagree (0) with the item (‘I have friends who are criminals’).	n/a	n/a
Anger	Anger	Self/emotional regulation issues	3	Participants respond to each item (e.g. ‘I often get mad’) on a 3-point Likert scale, where 1 = never, 2 = sometimes and 3 = always.	α = .85 (Barrowcliffe et al., 2022)	α = .75

Note: M-TTAF = Multi-Trajectory Theory of Adult Firesetting; FSS = The Fire Setting Scale (Gannon & Barrowcliffe, 2012); Previous exposure to fire (adapted from Murray et al., 2015); BIS = Barratt Impulsiveness Scale–Brief (Steinberg et al., 2013); RSES = Rosenberg Self-Esteem Scale (Rosenberg, 1965); RQ = Relationship Questionnaire (Bartholomew & Horowitz, 1991); MSPSS = Multidimensional Scale of Perceived Social Support (Zimet et al., 1988); Criminal associates (Barrowcliffe et al., 2022); Anger (Barrowcliffe et al., 2022).

Procedure

Ethical approval was obtained from the Victoria University of Wellington Human Ethics Committee (Ref: 0000028938). Data were collected between 26 April and 11 May 2021. Participants accessed the online survey through a weblink or QR code provided on the advertisements. Advertisements explicitly stated that the research aimed to ‘explore similarities and differences between people who choose to set fires, and people who choose not to’. Participants were first presented with an information sheet and consent form that emphasised their anonymity and were required to tick a box stating they agreed to participate before they could proceed. Participants were specifically requested not to report any identifiable information about themselves or any fires they had deliberately set. The survey took on average 20 minutes to complete. Participants first completed the section on background and demographics, followed by the fire behaviour section, and finally the M-TTAF domain questionnaires. The M-TTAF domain questionnaires were presented in a randomised order to reduce the effect of ordering bias. Following study completion, individuals were thanked for their participation and were offered the opportunity to be entered into a prize draw to win one of five $100 supermarket vouchers. Details for the prize draw were collected by accessing a link to a separate survey to ensure participants’ survey answers remained anonymous. Participants were then provided with a debrief sheet including details for support services.

Statistical analysis

Power analyses using the pwr package in R statistics (Champely et al., 2020) indicated that at least 102 participants were needed for each independent t test and 88 participants for each chi-quare test to detect a medium effect size (0.50 t test, 0.30 χ²) at 80% power and α = .05. In addition, a minimum of 10 participants were needed for each independent variable per outcome event for a binomial logistic regression, based on Vittinghoff and McCulloch’s (2007) guidelines.

Missing item responses on M-TTAF questionnaires were addressed by computing alternative total scores using pro-rating, since it has been demonstrated as a reliable method for dealing with small numbers of missing data points (Schretlen & Ivnik, 1996). Total scores were pro-rated using average scores on completed items. In the present study, there was a total of 50 pro-rated responses. Participants who reported setting at least one deliberate fire were classified as firesetting individuals, and those who reported setting zero fires were classified as non-firesetting individuals.

Analyses were conducted in IBM SPSS Statistics Version 28. Frequencies and descriptive statistics from each group were used to determine firesetting prevalence and features. Between-group comparisons were conducted using independent-sample t tests and chi-square tests of association. Prior to commencing analysis, general and test-specific assumptions were examined. If data were not normally distributed, Mann–Whitney U tests were computed instead of independent-sample t tests. When assumptions of equal variances were violated, the Levene’s test for equality of variances was reported. If expected cell counts were less than 5, Fisher’s exact probability test was reported instead of chi square.

Odds ratios (ORs) were reported as a measure of effect size for chi-square tests. Effect sizes for t tests were interpreted using Cohen’s (1988) criteria, where Cohen’s d = 0.2, 0.5 and 0.8 are considered small, medium and large effect sizes, respectively. Correlation coefficients were calculated as an effect-size measure for Mann–Whitney U tests using the following formula, r = Z/√n, where Z is the standardised U value, and were interpreted using Cohen’s guidelines for Pearson’s r, whereby r = .1, .3 and .5 indicate small, medium and large effect sizes, respectively (Fritz et al., 2012).

After applying Bonferroni adjustments, variables that reached statistical significance and those trending towards significance were entered into a binary forced entry logistic regression to identify which variables distinguish firesetting and non-firesetting participants, while controlling for other factors. The logistic regression was limited to those participants who had no missing data on any of the selected variables (n = 606, 96.8%).

Results

Firesetting prevalence and features

One-hundred and three participants (16.45% of the total sample) reported having deliberately set a fire over the age of 14 years for antisocial or fire interest purposes. The majority of firesetting participants were female (n = 86, 83.5%) and identified as New Zealand European/Pākehā (n = 83, 80.6%), reflecting the base rate of identifying as female and New Zealand European/Pākehā in the total sample. Nearly all firesetting participants reported achieving some formal education (n = 98, 95.1%), with 45.7% (n = 47) having achieved high-school qualifications as their highest level of formal education, 8.7% (n = 9) having achieved vocational training, 28.2% undergraduate qualifications (n = 29) and 12.6% (n = 13) postgraduate qualifications. Sixty-eight firesetting participants were employed (66%), and 60 (58.3%) reported having been previously diagnosed with a mental health disorder. Ten of the firesetting participants (9.7%) reported previous criminal convictions; however, none reported previous convictions for arson. (See Table 3 for overview.)

Table 3.

Characteristics of deliberate firesetting individuals.

Offence characteristics	M (SD)	% yes (n)
Age at firesetting incident
First deliberate fire	15.08 (6.22)
Last deliberate fire	21.25 (8.44)
Number of deliberate fires ignited
One		31.1 (32)
Two		20.4 (21)
Three		17.5 (18)
Four		4.9 (5)
Five or more		26.2 (27)
Motivations a
Experimenting/curious and understood the dangers		55.3 (57)
Bored		50.5 (52)
Loves fire		33.0 (34)
To create excitement		31.1 (32)
Experimenting/curious with lack of fire knowledge		16.5 (17)
Stressed/frustrated		15.5 (16)
Issues at home		11.7 (12)
Issues at school		9.7 (10)
Sanctioned		8.7 (9)
Act of vandalism		8.7 (9)
Wanted attention		6.8 (7)
Reaction to stressful life event		6.8 (7)
Angry		5.8 (6)
Dared to/pranked		4.9 (5)
Revenge		2.9 (3)
Protecting self		1.9 (2)
Covering crime/destroying evidence		1.9 (2)
Performance		1.0 (1)
Taking drugs		1.0 (1)
Peer pressure		1.0 (1)
Issues at work		0
Insurance/financial gain		0
Firesetting behaviour detected
By someone in a position of authority		2.9 (3)
By a parent or member of the whānau		23.3 (24)
Received intervention for deliberate firesetting
Yes		1.0 (1)
No		99.0 (102)

^aParticipants were able to select multiple motives; therefore, these do not add up to 100%.

On average, firesetting participants reported igniting their first deliberate fire during adolescence (M = 15 years, SD = 6.22); however, age of first fireset ranged between 4 years and 60 years. Individuals’ last fireset occurred, on average, during early adulthood (M = 21 years, SD = 8.44), with age of last fireset ranging between 14 years and 61 years. The average length of time since an individual’s last set fire was 8.72 years (SD = 10.20); however, it ranged from 0 years to 53 years. Over two thirds of firesetting individuals reported igniting multiple fires (69%, n = 71), with 27 of these (38%) reporting having set five or more fires.

Firesetting participants reported a variety of motives for their firesetting. The most common motivations included experimenting/curious and understood the dangers (n = 57, 55.3%), boredom (n = 52, 50.5%), loves fire (n = 34, 33.0%) and to create excitement (n = 32, 31.1%). Seventy-nine firesetting participants (76.7%) reported their fires had never been detected by a parent or member of their whānau/family, and 100 (97.1%) reported having never been detected by someone in a position of authority (e.g. police, teacher). Only one participant had previously received an intervention for their firesetting.

Comparison of firesetting and non-firesetting individuals

Demographic and background characteristics

After applying Bonferroni adjustments (p ≤ .001), firesetting and non-firesetting individuals could only be differentiated on a small number of demographic and background characteristics (see Table 4). Compared to non-firesetting participants, firesetting participants were more likely to report high-school qualifications (e.g. NCEA Level 1, 2 or 3) as their highest level of educational attainment, χ²(1, n = 622) = 13.23, p < .001, OR = 2.21, and were more likely to have previously received a mental health diagnosis χ²(1, n = 626) = 13.62, p < .001, OR = 2.22.

Table 4.

Demographic and background characteristics of firesetting and non-firesetting individuals.

Variable	Firesetting individuals (n = 103)		Non-firesetting individuals (n = 523)
	M (SD)	% (n)	M (SD)	% (n)
Current age (years)	30.18 (11.40)		38.49 (13.95)
Male		14.6 (15)		11.9 (62)
Female		83.5 (86)		86.4 (452)
Other gender		1.9 (2)a		1.1 (6)
NZ European/Pākehā		80.6 (83)		83.2 (435)
Māori		19.4 (20)		11.7 (61)
European		9.7 (10)		8.4 (44)
Pasifika		4.9 (5)a		2.3 (12)
Other ethnicity		5.8 (6)		8.3 (43)
Formal education		95.1 (98)		93.5 (489)
High-school qualifications		45.7 (47)*		27.7 (145)
Vocational training		8.7 (9)		13.6 (71)
Undergraduate qualifications		28.2 (29)		33.5 (175)
Postgraduate qualifications		12.6 (13)		18.8 (98)
No formal education		3.9 (4)a		5.9 (31)
No formal education reported		1.0 (1)a		0.6 (3)
Employed		66.0 (68)		76.7 (401)
Criminal convictions		9.7 (10)		5.7 (30)
Mental health diagnosis		58.3 (60)*		38.6 (202)

^aAssumption of expected cell count violated: Fisher’s exact test used.

*Statistically significant result: Bonferroni correction p ≤ .001.

M-TTAF domains

After applying Bonferroni adjustments (p ≤ .004), firesetting individuals scored significantly higher than non-firesetting individuals on levels of fire interest, t(132.87) = 6.49, p < .001, d = 0.78, levels of impulsivity, U = 33,489.500, p < .001, r = .16 and fearful attachment styles, U = 32,368.000, p < .001, r = .15. Firesetting participants also scored significantly higher than their non-firesetting counterparts on levels of anger t(625) = 3.51, p < .001, d = 0.38, and were significantly more likely to report having antisocial associates than non-firesetting individuals, χ²(1, n = 620) = 14.41, p < .001, OR = 2.39. There were no significant differences between the groups on levels of fire exposure, social support, self-esteem or the remaining three attachment styles (secure, pre-occupied and dismissive; see Table 5).

Table 5.

M-TTAF domain scores for firesetting and non-firesetting individuals.

Measure	Firesetting individuals (n = 103)		Non-firesetting individuals (n = 523)		t/U	p	95% CI	d/r
	M (SD)	N	M (SD)	N
FSS
Fire interest items	38.46 (14.51)	103	28.54 (12.39)	523	6.49	<.001*	[6.90, 12.94]	0.78
Exposure to fire (before age 10)	3.23 (1.31)	103	3.30 (1.449)	522	0.46	.647	[0.22, 0.35]	0.05
Exposure to fire (between ages 10–18)	3.16 (1.14)	103	3.33 (1.31)	522	1.385	.168	[0.07, 0.42]	0.14
Exposure to fire (since age 18)	3.62 (1.24)	103	3.84 (1.18)	522	1.74	.082	[0.03, 0.48]	0.19
BIS	17.68 (4.23)	101	15.86 (4.11)	511	33,489.500a	<.001*	[1.00, 3.00]	.16
RQ
Secure	4.03 (1.62)	101	4.30 (1.85)	511	23,659.000a	.085	[0.00, 1.00]	.07
Fearful	4.41 (1.94)	101	3.65 (1.94)	511	32,368.000a	<.001*	[0.00, 1.00]	.15
Preoccupied	3.56 (1.91)	101	3.08 (1.77)	511	30,112.500a	.017	[0.00, 1.00]	.10
Dismissing	4.79 (1.76)	101	4.79 (1.61)	511	26,583.500a	.792	[0.00, 0.00]	.01
RSES	31.01 (6.14)	103	32.73 (4.87)	522	2.70	.008	[0.46, 3.00]	0.34
Anger	5.39 (1.37)	103	4.90 (1.26)	522	3.51	<.001*	[0.21, 0.75]	0.38
MSPSS	5.00 (1.29)	101	5.39 (1.21)	511	22,251.500a	.005	[0.08, 0.58]	.11
	% yes	n	% yes	n	χ2	p	OR
Antisocial associates	36.6	37	19.5	101	14.41	<.001*	2.39

Note: M-TTAF = Multi-Trajectory Theory of Adult Firesetting; FSS = The Fire Setting Scale; RQ = Relationship Questionnaire; RSES = Rosenberg Self-Esteem Scale; MSPSS = Multidimensional Scale of Perceived Social Support; CI = confidence interval; OR = odds ratio.

^aAssumption of normality violated: Mann–Whitney U test used – test statistic U, effect size r.

*Statistically significant result: Bonferroni correction p ≤ .004.

Factors that distinguish firesetting and non-firesetting individuals

All variables that significantly differentiated firesetting and non-firesetting individuals in the univariate analyses and those trending towards significance were entered into a single forced entry binary logistic regression model (high-school qualifications, mental health diagnosis, fire interest, fearful attachment style, impulsivity, self-esteem, anger, antisocial associates and social support). No variables within the model were correlated more than .67. The overall model significantly improved upon chance prediction, χ²(9, n = 606) = 82.39, p < .001 (see Table 6), explaining between 12.7% (Cox and Snell R²) and 21.6% (Nagelkerke R²) of the variance in firesetting status. Four of the variables in the model (high-school qualification, mental health history, fearful attachment style and fire interest) made statistically significant contributions. Firesetting participants had twice the odds of reporting high school as their highest level of formal education (OR = 2.24) and just under twice the odds of reporting a history of a mental health diagnosis (OR = 1.91). In addition, firesetting participants had slightly higher odds than non-firesetting participants of exhibiting fearful attachment styles (OR = 1.16) and self-reported fire interest (OR = 1.05).

Table 6.

Logistic regression model: firesetting versus non-firesetting individuals.

Variables	β	Wald χ2	p	OR	95% CI for OR
High-school qualifications	.81	10.27	.001*	2.24	[1.37, 3.67]
Mental health diagnosis	.65	5.92	.015*	1.91	[1.13, 3.21]
Fire interest	.05	34.61	<.001*	1.05	[1.04, 1.07]
Fearful attachment style	.15	3.90	.048*	1.16	[1.00, 1.35]
Impulsivity	.02	0.58	.446	1.02	[0.96, 1.09]
Self-esteem	.05	2.12	.146	1.05	[0.99, 1.11]
Anger	.14	2.03	154	1.15	[0.95, 1.40]
Antisocial associates	.26	0.91	.339	1.30	[0.76, 2.22]
Social support	−.09	0.75	.386	0.91	[0.74, 1.12]

Note: n = 606. OR = odds ratio. Hosmer Lemeshow Test (HL) goodness of fit: χ²(8) = 10.38, p = .24.

*p < .05.

Discussion

This study represents the first examination of the prevalence and characteristics associated with un-apprehended firesetting in a New Zealand community sample. One hundred and three participants (16.45%) self-reported having set at least one deliberate fire since the age of 14, with over two thirds of this group disclosing that they had set multiple fires. The self-reported prevalence of firesetting in this study is broadly in line with the pooled prevalence of Barrowcliffe et al.’s three earliest studies (14%; Barrowcliffe & Gannon, 2015, 2016; Gannon & Barrowcliffe, 2012; Gannon et al., 2022), yet notably higher than those reported in the NESARC studies (1.0%–1.3%; Blanco et al., 2010; Hoertel et al., 2011; Vaughn et al., 2010). However, this likely reflects methodological similarities between the current study and Barrowcliffe et al.’s studies and differences in methodology with the NESARC studies.

In terms of key features, the majority of firesetting participants identified as female, New Zealand European/Pākehā, and having achieved some level of formal education. While these findings reflect the demographic composition of our wider sample as opposed to highlighting characteristics that distinguish firesetting from non-firesetting participants, the ratio of males to females in un-apprehended firesetting samples is typically reported to be more equal than in apprehended samples, suggesting females may be represented at a higher rate in un-apprehended populations (Gannon et al., 2022). Over two thirds of firesetting participants reported having set multiple fires, mirroring rates reported in UK studies with un-apprehended samples (Barrowcliffe & Gannon, 2015, 2016; Barrowcliffe et al., 2022). The elevated rates of multiple firesetting in un-apprehended samples may reflect increased opportunity and reinforcement of firesetting as a consequence of not being detected or apprehended for this behaviour, with only a minority of participants reporting being detected for firesetting and only one reporting having received intervention. The most frequently reported motives for firesetting included experimentation, excitement, boredom and fire interest. These findings are similar to motivations reported in UK un-apprehended studies and potentially suggest that firesetting in this group is driven by an underlying need for immediate gratification in un-apprehended individuals (Gannon et al., 2022).

On average, participants’ self-reported firesetting occurred between adolescence (i.e. average age of first fireset = 15 years) and early adulthood (i.e. average age of last fireset = 21 years). This pattern of findings suggests that un-apprehended firesetting may follow an adolescent-limited developmental trajectory, where firesetting is a transient behaviour that dissipates through social maturation by early adulthood (Moffitt, 1993). Indeed, previous research suggests that firesetting behaviour typically peaks in early adolescence (i.e. age 12–14 years) and declines thereafter (Del Bove et al., 2008; Jacobson, 1985; Klein et al., 2008). However, given the current study only examined time at first and last fire, further longitudinal multi-wave research is needed to examine developmental pathways for un-apprehended firesetting including onset and desistance.

Firesetting participants were significantly more likely than non-firesetting participants to hold a high-school qualification (e.g. NCEA Level 1, 2 or 3) as their highest level of educational attainment and have a history of a mental health diagnosis. In addition, firesetting individuals scored significantly higher than non-firesetting individuals on measures of impulsivity, fearful attachment style (i.e. high anxiety, high avoidance), anger, fire interest and antisocial associates. In the regression model, having a high-school qualification as highest level of education attainment, history of a mental health diagnosis, fearful attachment style and fire interest were the variables most strongly associated with firesetting status, indicating that people with this combination of factors may be more likely to set fires than those who do not have these factors.

Having a high-school qualification as the highest level of educational attainment (i.e. NCEA Level 1, 2 or 3) was the strongest predictor of firesetting status. This finding is consistent with previous un-apprehended firesetting research (Barrowcliffe & Gannon, 2015, 2016; Gannon & Barrowcliffe, 2012), yet contrasts with findings from the apprehended literature, which reports that apprehended firesetting adults have low levels of educational achievement (Anwar et al., 2011; Räsänen et al., 1995), low levels of intelligence (Bradford, 1982; Lewis & Yarnell, 1951) and poor problem-solving skills (Jackson et al., 1987; Tyler et al., 2014). It has previously been suggested that the ability to effectively evade detection for firesetting may be determined, in part, by un-apprehended individuals’ educational competency (Barrowcliffe & Gannon, 2015). However, it is important to note that the level of educational attainment reported by firesetting individuals was typically lower than those reported by non-firesetting individuals in our study, with a higher proportion of non-firesetting participants reporting post-secondary education achievement.

Having a diagnosis of a mental health condition was the next strongest predictor of firesetting status. These findings are in line with previous apprehended research, where mental health issues have been found to be associated with firesetting in both identified adolescents (Martin et al., 2004; Repo & Virkkunen, 1997; Roe-Sepowitz & Hickle, 2011) and apprehended adults (Ducat, Ogloff, et al., 2013; Ó Ciardha et al., 2015). However, previous research with un-apprehended samples has yielded mixed findings with regards to psychopathology, with some studies reporting differences in self-reported suicide, self-harm and mental health diagnoses (e.g. Barrowcliffe & Gannon, 2016; Blanco et al., 2010; Vaughn et al., 2010) and others finding no difference (e.g. Barrowcliffe & Gannon, 2015; Barrowcliffe et al., 2022). Research with apprehended samples has indicated that certain diagnoses are particularly prevalent amongst adults who set fires including mood disorders, personality disorders, substance use disorders and psychotic disorders (Ducat, Ogloff, et al., 2013; Nanayakkara et al., 2015; Tyler & Gannon, 2012), and that mental health difficulties may occur proximally to firesetting (Barnoux et al., 2015; Tyler et al., 2014). Future research would benefit from examining specific aspects of un-apprehended individuals’ diagnoses (e.g. diagnosis type, age at diagnosis, previous contact with mental health services) to determine whether there are any diagnostic considerations that may explain these equivocal findings.

Only one study to date has examined attachment styles in un-apprehended firesetting populations (Barrowcliffe et al., 2022), which, in contrast to the present study, found that firesetting and non-firesetting individuals could not be distinguished on attachment style. Although there has been no specific research examining attachment styles in the apprehended firesetting literature, the wider literature on attachment and offending suggests that individuals with fearful attachment styles often lack intimacy, struggle with distress and experience self and emotional regulation issues (Adshead & Moore, 2021). Further, firesetting adults are reported to experience problems with forming and maintaining relationships, often lack social support, and experience feelings of loneliness and isolation (Inciardi, 1970; Rice & Harris, 1991; Ritchie & Huff, 1999). The current findings suggest that individuals who engage in un-apprehended firesetting may experience relationship issues due to a fearful attachment style; however, further research is needed to understand the relationship between attachment, firesetting and other risk factors for this behaviour.

Finally, firesetting participants reported significantly higher levels of fire interest than non-firesetting participants. The relationship between increased fire interest and firesetting is consistent with previous research on apprehended and un-apprehended firesetting in both adults and adolescents (Barrowcliffe & Gannon, 2015, 2016; Gannon et al., 2013; Mackay et al., 2006; Rice & Harris, 1991; Tyler et al., 2015). However, interestingly, the mean level of fire interest reported by both our firesetting and non-firesetting participants is considerably higher than the levels reported in previous research with un-apprehended UK firesetting adults (Barrowcliffe & Gannon, 2015, 2016). One potential explanation is that there may be variation in normative levels of fire interest as a result of differences in socio-cultural norms and values around fire and types of fire exposure (Fessler, 2006; Gannon et al., 2012). Although no relationship was found between level of fire exposure and firesetting status in the current study, there has been no empirical research to date that has studied type of fire exposure or attitudes, interests and associations with fire across different socio-cultural contexts. These findings emphasise the need for cross-cultural research on firesetting and fire-related constructs to understand variations in normative levels so that assessment and prevention approaches can be tailored accordingly.

Implications

Univariate analyses indicated differences between un-apprehended firesetting and non-firesetting participants across several psychological vulnerabilities and moderators within the M-TTAF; including interests, attitudes and associations with fire, self and emotional regulation difficulties, and mental health difficulties, consistent with the apprehended literature (e.g. Gannon et al., 2013). However, when controlling for all other variables in a regression model, only fire interest, fearful attachment and mental health diagnosis differentiated firesetting and non-firesetting participants. This suggests that these characteristics are likely important areas of treatment need for firesetting in both apprehended and un-apprehended samples; however, other psychological vulnerabilities within the M-TTAF (e.g. emotion regulation, self-esteem) may be less important in the aetiology of un-apprehended firesetting.

Further research is needed to understand the full range of risk and protective factors for un-apprehended firesetting. For example, while there was no difference in the number of criminal convictions between firesetting and non-firesetting individuals, firesetting individuals were more likely to report having antisocial associates. Future research would benefit from examining how general antisociality and factors known to contribute to antisocial behaviour more broadly may help to explain firesetting in un-apprehended samples. This would allow for the examination of factors that are not fully accounted for by existing theories of firesetting (e.g. situational variables), to enable refinement of aetiological frameworks (e.g. the M-TTAF) to account for this population.

Given empirical research on un-apprehended firesetting is still in its infancy, using qualitative methods to develop a micro-theory of un-apprehended firesetting may provide a helpful starting point. Micro-theories have proved useful for informing understanding of how the offence process unfolds for apprehended firesetting (e.g. Barnoux et al., 2015; Collins et al., 2021; Tyler & Gannon, 2017; Tyler et al., 2014) as well as other types of offending behaviour (e.g. Chambers et al., 2009; Gannon et al., 2008; Ward et al., 1995), including un-apprehended samples (e.g. Ford et al., 2021). Identifying the sequalae of cognitive, affective, situational and contextual factors surrounding incidents of un-apprehended firesetting can improve understanding of the aetiology of behaviour, highlight points for intervention and generate new hypotheses for empirical inquiry.

Our results indicate that very few un-apprehended firesetting individuals have ever been detected for firesetting (23.3%), and even fewer have received treatment for their behaviour (1%). While specialist treatment has been demonstrated as a useful approach for reducing psychological vulnerabilities associated with firesetting in apprehended samples (e.g. Gannon et al., 2015; Tyler et al., 2018), there is a lack of research on effective primary prevention strategies for firesetting. School and community-based fire safety education is one existing approach that may prove useful for increasing fire safety awareness and reducing fire interest. However, evaluations on the effectiveness of this approach in preventing onset of firesetting are still needed (Gannon et al., 2022). Given that un-apprehended firesetting individuals report mental health and attachment issues, it is possible they have contact with other support services (e.g. GP, mental health services or social services), providing opportunities for early identification and screening of concerning interests and behaviours with fire, and providing onward referral to existing youth fire safety education interventions (Gannon et al., 2022). This is particularly important given our findings about the relationship between increased fire interest and probability of firesetting.

Strengths and limitations

This research represents the first empirical study of un-apprehended firesetting in New Zealand. It is also the largest empirical study to date to explicitly examine the characteristics associated with deliberate firesetting in an adult community sample. However, there are several methodological limitations within which the current findings should be considered. First, although this study used a semi-stratified recruitment process, its generalisability is limited by the overrepresentation of certain demographics in the sample (e.g. females and educated individuals). It is therefore unclear whether the same results would be observed in a sample that more closely represented the demographic composition of the New Zealand population. Research with apprehended samples suggests that there are differences between males and females in terms of their presentation, antisocial behaviour, psychopathology and motivations for deliberate fire use (Hoertel et al., 2011; Nanayakkara et al., 2020). Future research would therefore benefit from recruiting a more representative community sample, to determine whether gender differences are also evident in un-apprehended populations. Second, the results are limited by the study’s use of self-report measures. While participants were assured that their responses were anonymous and were specifically requested not to report any personal information that could lead to identification, individuals may still have been reluctant to participate or provide honest answers regarding their firesetting behaviour. In addition, similar to other un-apprehended firesetting research, the majority of our participants were retrospectively reporting on their firesetting behaviour, with the average age of first fireset reported as during mid-adolescence (i.e. 15 years) and the average age of last fireset in early adulthood (i.e. 21 years). Therefore, memory recollection failures are a possibility. Un-apprehended firesetting research has started to focus on younger groups of participants to address these potential effects (see Barrowcliffe et al., 2022); however, longitudinal multi-wave research where participants have to report firesetting over a shorter period of time may also reduce these impacts, as well as further our understanding of the aetiology of un-apprehended firesetting over time. Finally, the firesetting disclosure question (Gannon & Barrowcliffe, 2012) used to screen for firesetting behaviour focuses on identifying firesetting for antisocial and fire interest purposes, meaning it may have failed to capture fires set for other reasons (e.g. as a cry-for-help, in response to psychotic symptoms). However, this may be less of an issue given that antisocial and fire interest purposes capture the most commonly reported reasons for deliberately setting fires (Barrowcliffe et al., 2022). Future research would benefit from further refining the firesetting screening question to capture other functions for problematic fire use to broader our understanding of firesetting behaviour in the general population.

Conclusions and future directions for research

Based on the current findings, deliberate firesetting appears to be a relatively common behaviour in the New Zealand general population, with rates comparable or slightly higher than those reported in the UK and USA. Furthermore, there appear to be notable differences between community individuals who engage in deliberate firesetting behaviour and those who do not, particularly in terms of lower levels of education attainment (e.g. high-school qualification as highest level of educational achievement), presence of mental health diagnoses, fearful attachment style and interest in fire. While the current study examined demographic, background and psychological factors associated with firesetting status, it did not explore wider contextual, social or offence-related characteristics nor factors that may protect against un-apprehended firesetting. Given the utility of these factors for prevention and intervention initiatives, future research would benefit from examining contextual, social, offence-related and protective factors for firesetting in un-apprehended samples. Further, given the young average age at first and last fireset, incorporating psychological characteristics associated with general antisocial behaviour in adolescence in future research may also prove to be a fruitful avenue for enquiry. Finally, discriminant studies with both apprehended and un-apprehended firesetting individuals are needed to advance our understanding of the similarities and differences between these groups. By developing our knowledge of un-apprehended firesetting, we can develop more effective and tailored primary prevention strategies to reduce the incidence and harmful consequences of this pervasive behaviour.

Ethical standards

Declaration of conflicts of interest

Kendal Johnston has declared no conflicts of interest

Nichola Tyler has declared no conflicts of interest

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the Victoria University of Wellington Human Ethics Committee (0000028938) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed consent

Informed consent was obtained from all individual participants included in the study.