Download - Adolescent risk-taking and the five-factor model of personality

Transcript

Journal of Adolescence 2000, 23, 393±407doi:10.1006/jado.2000.0327, available online at http://www.idealibrary.com on

Adolescent risk-taking and the five-factor modelof personality

ELEONORA GULLONE AND SUSAN MOORE

The aim of this study was to investigate the links between adolescent risk-taking andpersonality in a comprehensive manner, as opposed to focusing on any one particularrisk or personality factor as has occurred in much previous research. We conceptualizedrisk-taking broadly through use of the Adolescent Risk Questionnaire (ARQ) whichyields information relating to risk judgements and behaviours in four areas (i.e. thrill-seeking risk behaviours, reckless risks, rebellious risks and antisocial risks). Personalitywas conceptualized using the Five-factor Model of personality, a comprehensive traitmodel of personality. The ARQ and the NEO Five-factor Inventory were administeredto 459 school-based adolescents aged 11 to 18 years. Consistent with past research,younger adolescents and girls generally reported engaging in risk behaviours lessfrequently than older adolescents and boys. Also, younger adolescents and girlsgenerally rated the ARQ behaviours as more risky than their older male counterparts.This was in line with the significant negative correlations found between riskjudgements and risk behaviours of all types. Also consistent with past research, few agedifferences were found for the personality traits. Sex differences were evident, withfemale adolescents scoring higher on neuroticism, agreeableness and conscientiousnessthan male adolescents. Together, risk judgements, personality factors, age and sex werefound to be significant predictors of risk behaviours; however, the personality factor ofsignificance was found to differ depending upon the risk type. This was also true for ageand sex which differed in importance depending upon the risk type. Overall, thesefactors were most successful in predicting rebellious risk-taking and least successful inpredicting thrill-seeking.

# 2000 The Association for Professionals in Services for Adolescents

Introduction

Much recent research on adolescent risk-taking behaviour has focused on factors underlyingadolescents' decisions to participate in risky behaviour and characteristics which describerisk-taking adolescents. However, most of this research has been limited by typically studyingonly one type of risk behaviour (such as smoking) and/or only one or two personalitycharacteristics (such as sensation-seeking). Also, the conceptualization of risky behaviourhas typically encompassed only maladaptive risks, although there is some evidence that traitsunderlying risk-taking in its negative sense may also underlie socially approved riskybehaviours such as creative endeavours (Silbereisen and Noack, 1988; Chassin et al., 1989;Moore and Gullone, 1996).

Our definition of risk-taking is the participation in behaviour which involves potentialnegative consequences (or loss) balanced in some way by perceived positive consequences(or gain). Defining risk-taking in this general way allows the possibility of testing whether

Reprint request and correspondence should be addressed to Dr E. Gullone, Dept of Psychology, MonashUniversity, Clayton, Victoria 3168, Australia. (E-mail: [email protected])

0140-1971/00/040393+15 $35�00/0 # 2000 The Association for Professionals in Services for Adolescents

394 E. Gullone and S. Moore

seemingly diverse behaviours such as substance abuse and sky-diving might be predicted bysimilar underlying factors such as specific personality traits. In fact, there is some evidencethat this may be the case, with respect to both maladaptive (or deviant) risk-taking (e.g.Gorsuch and Butler, 1976; Jessor and Jessor, 1977; Kandel, 1980; Block et al., 1988) andsocially approved risk-taking. An example is Chassin et al.'s (1989) study, which showed thatsome substance-abusing youth were also more likely to be unconventional, creative,independent and assertive than their non-substance abusing peers.

Another advantage of defining risk-taking in a general way is that such a definition invitesthe opportunity to explore empirical groupings of different types of risk behaviours, such ashas been done by a small number of researchers (e.g. Alexander et al., 1990; Lavery andSiegel, 1993). Through factor analysis, our own research has confirmed four broad groupingsof risk-taking behaviour, including thrill-seeking, rebellious, reckless and antisocial riskbehaviours (Gullone et al., 2000).

Thrill-seeking risks refer to behaviours which are challenging but (relatively) sociallyacceptable, for example, engaging in dangerous sports and experimenting with relationshipsand sexuality. Reckless behaviours include those which, though thrill-seeking for the mostpart, have a higher chance of negative social or health-related outcomes than is usuallyconsidered acceptable in the adult population. Examples include drinking and driving,having unprotected sex, sharing needles and speeding. Rebellious behaviours include thosewhich are often experimental rites of passage for young people as they seek independence.Their potential outcomes may be quite negative but not to the extent likely for recklessbehaviours. In addition, these are behaviours which are often engaged in by the adultgeneration but disapproved of for adolescents. Examples include smoking, drinking alcohol,swearing and staying out late. Finally, antisocial behaviours are those which are sociallyfrowned upon for adults and adolescents alike and not usually considered rites of passage.Examples are cheating, overeating and teasing others.

In several of our previous studies (Moore and Gullone, 1996; Moore et al., 1997; Gulloneet al., 2000) we have proposed that a workable model of risk behaviour can only be developedfollowing a conceptualization of risk-taking based on the views of young people themselves.One serious conceptual problem within the study of adolescent risk-taking has been that thedecision regarding what constitutes risky behaviour has invariably been made by theresearcher. Adult conceptions of risky activity may or may not correspond with those ofyoung people (Alexander et al., 1990). In response to this concern we propose that, in orderto better understand adolescent risk-taking, we must apply adolescent conceptualizations ofrisk.

In one of the few studies to incorporate adolescent nominated risks, Moore and Gullone(1996) collected data from 570 adolescents aged between 12 and 17 years in which theadolescents nominated behaviours they believed to be risky and evaluated these in terms ofpositive and negative outcomes. They also rated each behaviour according to how often theyengaged in it. A major finding was that whether or not the adolescents engaged in theirnominated behaviours, which included such things as drug-taking, dangerous driving, sex-related behaviours and sport-related activities, was significantly associated with adolescents'judgements about the riskiness of the behaviour. Clearly, however, more work of this nature isrequired before firm conclusions can be drawn.

An additional limitation of past research is that when individual characteristics ofrisk-taking adolescents have been investigated, typically only single personality traits havebeen examined in any one study. These have included, for example, sensation seeking,

Adolescent risk-taking and the five-factor model of personality 395

temperament, impulsivity (Zuckerman, 1979; Moore and Rosenthal, 1993), locus of control(Werner, 1986) and self-esteem (Garmezy, 1983). Protective factors against risk include ahigh level of self-esteem and an orientation towards an internal locus of control. In contrast,adolescents characterized by high levels of sensation-seeking and impulsivity appear to bemore vulnerable to engaging in risky behaviours. However, the manner in which personalityfactors combine to influence risk is unclear. In order to overcome this limitation, acomprehensive examination of individual difference variables is required.

Recently, there has been a proliferation of research into the Five-factor Model (FFM) ofpersonality. According to recent proposals within trait theory, the FFM represents the mostcomprehensive trait model to date (McCrae and Costa, 1987). Moreover, consistent supportfor the model has been found using a variety of data sets, including data from differentcultures (e.g. Digman and Inouye, 1986; Goldberg, 1990, 1992; Benet and Waller, 1995;McCrae et al., 1996).

As described by Goldberg (1993) the five broad domains incorporate hundreds of traits.The first factor of neuroticism, for example, incorporates anxiety, anger/hostility,impulsiveness and vulnerability. The extraversion factor is characterized by excitement-seeking, gregariousness and assertiveness. The third factor (i.e. openness to experience)incorporates traits such as fantasy, aesthetics, openness to feelings and values, andintellectual curiosity. The fourth factor (i.e. agreeableness) incorporates trust, straightfor-wardness and compliance, and the fifth factor (i.e. conscientiousness) incorporates traitssuch as self-discipline, deliberation, competence and orderliness.

In an attempt to develop a more comprehensive understanding of adolescent risk-taking,in this study we propose to overcome several of the shortcomings of past research whichhave been addressed above. Thus, we will assess adolescent risk-taking using a measure(i.e. the Adolescent Risk Questionnaire, ARQ) comprising the risks nominated byadolescents themselves (see Gullone et al., 2000 for details regarding developmentof this measure). We propose to examine the relationships between the four differenttypes of risk assessed using this measure and the five factors of the FFM. Furthermore, giventhe finding by Moore and Gullone (1996) that adolescent participation in risky behaviourcan be predicted through knowledge of adolescents' judgements of the riskiness of behaviour,we will examine the relationship of judgements to risk behaviour. We hypothesize thatpersonality traits will be significantly related to risk judgements (cf. Costa and McCrae,1980). Also, on the basis of past risk research, we propose that risk judgements will besignificantly and negatively associated with risk behaviours. Finally, we will investigate therelative importance that personality traits and risk judgements have for predicting riskbehaviours.

Method

SubjectsA sample of 459 (239 males, 220 females) adolescents, from four secondary governmentschools in Melbourne, Australia, completed both risk and personality questionnaires. Theschools were selected in order to approximate a representative sample, with regard tolocation, of Melbourne adolescents attending regular schools. There were 197 youngeradolescents aged 11 to 14 years (108 males, 89 females) and 262 older adolescents aged 15 to18 years (131 males, 131 females).

396 E. Gullone and S. Moore

The instruments

The Adolescent Risk-taking Questionnaire (ARQ; Gullone et al., 2000).Risk-taking behaviours and judgements were assessed using the two-part risk questionnairedeveloped on the basis of adolescent-nominated risk behaviours (Moore and Gullone, 1996).The 73-item questionnaire involves the rating of each item twice, first in relation tojudgement of riskiness (on a 5-point Likert scale: 0=not at all risky, 1=not very risky,2=risky, 3=very risky and 4=extremely risky) and a second time in relation to frequency ofparticipation in the behaviour (also on a 5-point scale: 0=never done, 1=hardly ever done,2=done sometimes, 3=done often, to 4=done very often). A total risk judgement score iscalculated by adding ratings on all items, with a high score indicating a stronger overalljudgement of riskiness for the behaviours depicted in the questionnaire. Similarly, a totalbehaviour score is calculated by summing the frequency rating of all items, with a high scoreindicating a higher overall level of participation in risky activity.

For an abbreviated ARQ, Gullone et al. (2000) reported a four factor structure for each ofthe ARQ judgement and behaviour scales explaining 50?1 and 53 percent of the variance,respectively. The factors are: ``Thrill-seeking risks'' (items include: Roller-blading,Parachuting, Snow skiing), ``Rebellious risks'' (items include: Smoking, Getting drunk,Staying out late), ``Reckless risks'' (items include: Drinking and driving, Having unprotectedsex, Speeding) and `Antisocial risks'' (items include: Cheating, Overeating, Teasing others).The thrill-seeking factor was found to explain the majority of the variance for riskjudgements (i.e. 17?6%) followed by the rebellious risks factor (i.e. 13?9%). Conversely, therebellious risks factor was found to explain the majority of the variance for risk behaviour (i.e.14?4%) followed by the thrill-seeking factor (i.e. 13?7%). Support for the invariance of thefactor structure was provided through confirmatory factor analysis.

The 73-item ARQ has also been demonstrated to have adequate internal consistency andgood test±retest reliability over a one-week period. Cronbach's alpha coefficients of 0?97 and0?99 have been reported for the overall risk judgement and behaviour scales. For the riskjudgement subscales, alpha coefficients ranged between 0?86 and 0?96. Alpha coefficientswere also adequate for the risk behaviour subscales and ranged from 0?87 to 0?96. One-weektest±retest reliability was reported to be 0?79 for risk judgements and 0?78 for riskbehaviours.

Finally, support for the validity of the ARQ was demonstrated by Gullone et al. (2000)by the age and sex differences yielded. As would be expected on the basis of past research,younger adolescents and girls scored higher on the ARQ judgements scale but lower on thebehaviour scale.

NEO Five Factor Inventory (NEO-FFI; Costa and McCrae, 1992). This is a 60-item questionnaire comprising 12 items representing each of the facets encompassed by thefive-factor model of personality (neuroticism, extraversion, openness to experience,agreeableness and conscientiousness). Respondents are required to endorse each item on a5-point scale (i.e. 5=strongly agree, 4=agree, 3=not sure or neutral, 2=disagree and1=strongly disagree) depending on what is most true for him or her. Items include, forexample, ``I keep my belongings clean and neat'', ``I work hard to accomplish my goals''(conscientiousness), ``I like to have a lot of people around me'', ``I often feel as if I'm burstingwith energy'' (extraversion), ``I would rather co-operate with others than compete withthem'' and ``Most people I know like me'' (agreeableness). High scores on the subscales of theNEO-FFI represent higher levels of the named trait. The psychometric properties reported

Adolescent risk-taking and the five-factor model of personality 397

for the NEO-FFI, while less strongly supported and researched than for the longer version ofthe questionnaire (i.e. the NEO Personality Inventory-Revised), remain sound. In the scalemanual, Costa and McCrae (1992) report intercorrelations between the five factors asmeasured with adjective self-reports and the NEO-FFI. They also report intercorrelationsbetween self-reports, spouse ratings and peer ratings. These correlations provide goodsupport for the convergent and divergent validity of the NEO-FFI scales. Stability has alsobeen demonstrated to be good through three-month test±retest reliability which has beenreported to range between 0?75 and 0?83 for the five scales.

Procedure. The risk and personality questionnaires were administered to the adolescents,in counterbalanced order, on a group basis. The two parts of the ARQ (judgements andbehaviours) were also presented to the students in counterbalanced order. Students weregrouped according to their regular school classroom groups. Data were collected by tworegistered psychologists with some assistance from the classroom teachers. The adolescentswere required to complete the questionnaires according to the specified directions. It wasparticularly emphasized to the adolescents that they were required to endorse each item withthe response that was most true for him or her. It was also emphasized that there were noright or wrong answers and that in responding to each item they should not be concernedwith other students' responses. Any questions which arose were clarified by the psychologist.The questionnaires took approximately 50 minutes for the students to complete but thisvaried slightly depending upon the age of the respondent, with the older adolescentsrequiring slightly less time for completion. The response rate was approximately 90 percent,with the majority of adolescents and parents who were approached to take part in the studyagreeing to participate.

Results

Below we present descriptive statistics for each of the risk variables (i.e. total and subscalescores) for the overall sample and each of the subsamples (i.e. 11±14-year-old, 15±18-year-old, boys and girls). Importantly, given that the risk subscales comprise different numbers ofitems, in order to enable comparison across subscales, all subscale mean scores which arereported have first been divided by the number of items in that particular subscale. However,the total risk scores (i.e. total judgement and behaviour scores) are presented as the totalsum of the 73 item ratings. Also presented are descriptive statistics for each of the fivepersonality factors (for the overall sample and subsamples). Age-group and sex differenceshave been examined for each of these variables using multiple analysis of variance(MANOVA) and ANOVA analyses. Finally, we examined the relationships between riskjudgements, behaviours and personality using Pearson's correlations and standard multipleregression analyses.

Age and sex differences for risk-taking judgements and behavioursTwo 2 (age-group)62 (sex) ANOVA analyses were carried out with each of the total riskjudgement and risk behaviour scores as dependent variables. These were followed by two 2(age-group)62 (sex) MANOVA analyses. The first MANOVA examined age and sexdifferences in the four risk judgement factors and the second examined age and sexdifferences in the four risk behaviour factors. In each MANOVA, significant multivariate age

398 E. Gullone and S. Moore

and sex effects were found. The details for these are provided in the notes of Tables 1 and 2,with each table relating to risk judgements and risk behaviours, respectively. These tablesalso report the significant univariate F values for the risk judgement and behaviour analyses.

As shown in Table 1, males and females as well as adolescents of varying age differedsignificantly with regard to their judgements of the riskiness of the behaviours depicted in the

Table 1 Total and subscale risk judgement scores by age and sex (n=459)

Risk scale Age Sex Total sample

11±14 years 15±18 years Males Females

M S.D. M S.D. M S.D. M S.D. M S.D.

Total risk 229?72 36?83 208?43 38?00 217?58 36?83 223?18 28?57 217?58 36?83(F(1,450)=44?05, p50?001) (F(1,450)=13?20, p50?001)

Thrill-seeking 3?24 0?38 3?09 0?48 3?10 0?51 3?21 0?35 3?15 0?44(F(1,450)=14?02, p50?001) (F(1,450)=7?02, p50?01)

Antisocial 2?86 0?63 2?52 0?64 2?55 0?72 2?78 0?55 2?66 0?66(F(1,450)=34?73, p50?001) (F(1,450)=15?56, p50?001)

Rebellious 3?34 0?56 2?91 0?69 3?04 0?77 3?15 0?55 3?09 0?67(F(1,450)=55?33, p50?001) (F(1,450)=4?63, p50?05)

Reckless 3?61 0?42 3?40 0?60 3?36 0?66 3?63 0?30 3?49 0?54(F(1,450)=22?93, p50?001) (F(1,450) = 29?74, p 5 0?001)

Notes: 1F value provided only for means which are significantly different,2 Multivariate age effect: (Wilks Lambda=0?87, F(4,447)=16?70, p50?001),Multivariate sex effect: (Wilks Lambda=0?88, F(4,447)=15?37, p50?001).

Table 2 Total and sub scale risk behaviour scores by age and sex (n=459)

Risk scale Age Sex Total sample

11±14 years 15±18 years Males Females

M S.D. M S.D. M S.D. M S.D. M S.D.

Total risk 126?76 46?10 144?38 40?94 142?48 48?26 130?76 38?15 136?85 44?04(F(1,448)=19?79, p50?001) (F(1,450)=9?53, p50?01)

Thrill-seeking 2?28 0?70 2?46 0?57 2?43 0?68 2?34 0?58 2?39 0?63(F(1,448)=9?84, p50?01) (F(1,450)=7?02, p50?01)

Antisocial 1?68 0?70 1?97 0?70 1?93 0?78 1?77 0?63 1?85 0?72(F(1,448)=20?22, p50?001) (F(1,448)=6?96, p50?01)

Rebellious 1?72 0?84 2?10 0?75 1?98 0?88 1?89 0?73 1?94 0?81(F(1,448)=26?98, p50?001)

Reckless 0?92 0?74 1?05 0?73 1?19 0?83 0?78 0?55 0?99 0?73[F(1,448)=4?76, p50?05) (F(1,448)=37?96, p50?001)

Notes: 1F value provided only for means which are significantly different.2Multivariate age effect: (Wilks Lambda=0.93, F(4,445)=8.85, p 5 0?001),Multivariate sex effect: (Wilks Lambda=0.89, F(4,445)=14.29, p 5 0?001).

Adolescent risk-taking and the five-factor model of personality 399

ARQ. In fact, a significant difference was found for the overall scale as well as each of its fourfactors. As indicated by the means, the differences resulted from lower judgements of riskbeing made, in each case, by girls and younger adolescents. It is also evident from Table 1that reckless risks were judged as most risky and antisocial risks as least risky.

In addition to the significant main effects reported in Table 1, the total judgement scoreANOVA revealed a significant interaction effect between age and sex resulting from therebeing a more pronounced sex difference for the older age-group than the younger (i.e. meandifference between 11±14-year-old male and female adolescents was 4?38, and for the 15±18-year-olds the difference was 17?65).

As shown in Table 2, significant age and sex differences were found on each of the totalrisk behaviour and the four risk behaviour factors, with the exception of a sex difference forthe rebellious risks factor. The differences were due to older and male adolescents reportinghigher risk behaviour scores than younger and female adolescents, respectively. Nosignificant interaction effects between age and sex were found. Also, it is noteworthy thatadolescents of varying age and sex reported engaging least often in reckless risks and mostoften in thrill-seeking risks.

Age and sex differences for the five personality factorsA 2 (age-group)62 (sex) MANOVA was carried out for the five personality factors of theNEO-FFI. In addition to significant multivariate age and sex main effects (see notes in Table3 for details), a significant multivariate age by sex interaction effect was found (WilksLambda = 0?96, F(5,432)=3?88, p50?01). Univariate examination of the interactionindicated that it was only significant for the conscientiousness factor (F(1,436)=5?29,p50?05). Further, the means showed that it was the result of older female adolescentsscoring higher on conscientiousness than older male adolescents whereas younger femaleadolescents scored lower than younger males.

Significant univariate sex effects were found only for the neuroticism, agreeableness andconscientiousness factors (See Table 3 for details of these results). Females scored higher on

Table 3 NEO-FFI personality score by age-group and sex (n=459)

PersonalityFactor

Age Sex Total sample

11±14 years 15±18 years Males Females

M S.D. M S.D. M S.D. M S.D. M S.D.

Neuroticism 35?93 6?47 35?91 5?73 34?92 5?91 36?94 6?02 35?92 6?04(F(1,436)=11?87, p50?01)

Extraversion 40?49 5?27 40?83 5?18 40?42 5?84 40?97 4?47 40?69 5?21

Openness to Ex. 36?32 5?06 36?83 4?67 36?75 5?35 36?48 4?25 36?62 4?83

Agreeableness 37?27 5?42 38?59 5?25 37?53 5?68 38?56 4?96 38?04 5?35(F(1,436)=6?92, p50?01) (F(1,436)=5?68, p50?05)

Conscient'ness 37?96 5?89 39?18 4?71 38?67 5?26 38?06 5?63 39?29 4?79(F(1,436)=5?91, p50?05) (F(1,436) = 4?13, p 5 0?05)

Notes: 1F value provided only for means which are significantly different.2Multivariate age effect: (Wilks Lambda=0?97, F(5,432)=2?85, p50?05),Multivariate sex effect: (Wilks Lambda=0?95, F(5,432)=4?18, p50?01).

400 E. Gullone and S. Moore

neuroticism and agreeableness, whereas males scored higher on conscientiousness.Significant univariate age effects were only found for agreeableness and conscientiousness,with younger adolescents scoring lower on both.

Intercorrelations between risk factors and personality factorsTable 4 shows the correlations between each of the four risk judgement factors and the fivepersonality factors, each of the four risk behaviour factors and the five personality factors aswell as the correlations between each of the risk judgement factors and the correspondingrisk behaviour factors. What is most evident from the correlation analyses is thatextraversion and agreeableness are each negatively and significantly (although weakly insome cases) related with all four of the risk judgement factors. That is, higher scores onextraversion and agreeableness relate with lower risk judgements for all types of risk.Conversely, for three of the four risk factors (thrill-seeking being the exception) high scoreson conscientiousness relate to high risk judgements (i.e. there are positive, albeit weak,correlations).

Risk behaviours correlate less consistently with personality, compared with riskjudgements. The most prominently associated personality factor appears to be agreeableness,which is positively (although weakly) associated with three risk factors (reckless risk-takingbeing the exception). Extraversion is positively associated with thrill-seeking risk behaviour.Openness to experience and conscientiousness are both negatively associated with rebelliousand reckless risk-taking. Finally, risk behaviours are significantly and negatively associatedwith risk judgements for all four risk factors. The correlations are all moderately sized withthe exception of those for the thrill-seeking factor which is somewhat smaller.

Multiple regression analysesFive standard multiple regression analyses (i.e. enter method) were calculated in order todetermine the predictive value of risk judgements, personality factors, age and sex (coded as0=male, 1=female) for risk behaviours. The total risk behaviour score and the four riskbehaviour factors each constituted a dependent variable per analysis. Also, the riskjudgement variable corresponding to the dependent risk behaviour variable was entered intoeach regression analysis as a predictor. For example, for the regression analysis examiningtotal risk behaviour as the dependent variable, the total risk judgement score was entered asa predictor. Likewise, for the thrill-seeking behaviour regression analysis, the thrill-seekingjudgement score was entered as the predictor. The outcomes of these analyses are reported inTable 5. As can be seen from this table, the analyses were all significant, with the predictorvariables explaining between 8 percent (for thrill-seeking behaviours) and 34 percent (forrebellious behaviours) of the variance. In all analyses, with the exception of that for thrill-seeking behaviours, the corresponding risk judgement score was a significant and alsorelatively more important predictor compared with other variables in the analyses. The age ofthe adolescent also featured prominently and was a significant predictor in all cases, with theexception of reckless risk-taking behaviour for which the adolescent's sex was significant.With regard to personality, a different personality factor appeared as important for eachdifferent type of risk-taking behaviour. Extraversion was found to be important for thrill-seeking, neuroticism for antisocial behaviour, conscientiousness and agreeableness forrebellious behaviour, and conscientiousness for reckless behaviour.

Tab

le4

Pear

son'

sP

rodu

ct±M

omen

tin

terc

orre

latio

nsbe

twee

nth

eri

skju

dgem

ent

and

beha

viou

rfa

ctor

san

dth

efiv

epe

rson

ality

fact

ors

Jud

gem

ents

Beh

avio

urs

Pers

onal

ity

Th

rill

-se

ekin

gA

nti

soci

alR

ebel

liou

sR

eck

less

NE

OA

C

Th

rill

-see

kin

g7

0?1

9***

70?

19

70?

22***

70?

01

70?

19***

0?04

An

tiso

cial

70?

43***

70?

11**

70?

17***

0?0

57

0?25***

0?17**

Reb

elli

ous

70?

55***

70?

06

70?

16**

0?0

77

0?21***

0?17***

Rec

kle

ss7

0?4

6***

70?

04

70?

12**

0?0

57

0?11*

0?12**

Neu

roti

cism

0?0

37

0?0

60?

02

70?

08

Ext

rave

rsio

n0?2

0***

0?0

80?

08

70?

05

Ope

nn

ess

0?0

37

0?0

47

0?1

1*

70?

15**

Agr

eeab

len

ess

0?1

7***

0?1

2**

0?1

8***

0?0

1C

onsc

ien

t'n

ess

0?0

37

0?0

57

0?1

9***

70?

22***

*p5

0?0

5,**p5

0?0

1,***p5

0?0

01

Adolescent risk-taking and the five-factor model of personality 401

Table 5 Five regression analyses for total risk behaviour and each of the risk factors with thecorresponding risk judgement variable, each of the five personality factors, age and sex as predictorvariables

Dep. variable Adj. r2 F df Significant predictors Beta t

Overall risk beh'r 0?23 17?46*** 8,423 Overall judgements 70?37 77?31***Age 0?11 2?29*

Thrill-seeking 0?08 5?86*** 8,423 Extraversion 0?23 3?56***Age 0?13 2?77**

Antisocial 0?20 14?77*** 8,423 Antisocial judgements 70?40 77?81***Neuroticism 70?15 72?94**Age 0?09 1?97*

Rebellious 0?34 27?33*** 8,425 Rebellious judgements 70?44 79?25***Conscientiousness 70?19 73?36***Agreeableness 0?14 2?72**Age 0?09 2?29*

Reckless 0?27 20?78*** 8,425 Reckless judgements 70?39 78?53***Sex 70?17 73?93***Conscientiousness 70?14 72?51*

*p50?05, **p50?01, ***p50?001, ****p50?0001

402 E. Gullone and S. Moore

Discussion

Prior to investigating the relationships between risk behaviour, risk judgements andpersonality traits, we examined age and sex differences in all the variables. As with pastresearch, younger adolescents reported higher levels of risk for the ARQ behaviours than didolder adolescents (cf. Gullone et al., 2000). Also, the behaviours of the adolescents wereconsistent with their judgements, in that older adolescents not only perceived the negativebehaviours to be less risky, they also reported participating in them more frequently than theyounger group. This is supportive of our hypothesis that risk judgements would be related torisk behaviour (cf. Moore and Gullone, 1996).

With regard to sex differences, female adolescents were found to score higher than maleadolescents on overall risk judgements as well as the four different types of risk. Consistentwith the risk judgement ratings, female adolescents reported engaging less frequently inthe ARQ behaviours overall, as well as the different risk behaviour types, with the exceptionof rebellious risks for which there was no sex difference. These findings, to an extent,consistently reflect gender stereotypes which characterize males as less empathic (cf.antisocial behaviours) but more courageous (cf. reckless risks and thrill-seeking), perhapsindicating that the adolescents provided responses on the questionnaires which they believedto be socially appropriate. Support for this proposal comes from the interaction effect onoverall risk judgements between age and sex, which indicated that the disparity in riskjudgements between males and females was markedly bigger in older adolescents, as would beexpected from increased gender role socialization at older ages.

On the other hand, these findings may be a reflection of actual sex differences in risk-taking behaviour (as opposed to a social desirability response bias). The fact that it is olderadolescents and male adolescents who are involved in a higher frequency of accidents lendssome support to the validity of these self-report data. The differences are consistent with

Adolescent risk-taking and the five-factor model of personality 403

those reported by Gullone et al. (2000) in their study, which also incorporated the ARQ witha different sample.

In contrast to the risk variables, age and sex differences were not as pervasive for the fivepersonality traits. That few age differences were found is not surprising since, according totrait theory, personality traits have a sizeable heritability component which predicts stability(e.g. Dworkin, et al., 1976; Conley, 1985; Eysenck, 1990). Nevertheless, there were agedifferences found for agreeableness and conscientiousness wherein the 11±14-year-oldadolescents scored lower than the 15±19-year-olds. These traits have also been reported tobe less heritable than, for example, neuroticism and extraversion. Similarly, Costa andMcCrae (1992) reported that older adults score higher on agreeableness and conscientious-ness but slightly lower on neuroticism, extraversion and openness to experience. Thus, itappears that individuals become more agreeable and conscientious as they get older andmore socialized to adult norms.

Sex differences were also found in relation to neuroticism, agreeableness andconscientiousness, such that females scored higher than males on the first two whereasmales scored higher on the latter trait. Consistent with these findings, Costa andMcCrae (1992) reported that adult females score higher on neuroticism and agreeablenessthan adult males. In contrast to the present findings, Costa and McCrae (1992) reportedno sex difference for the conscientiousness factor. The conscientiousness factor hasbeen proposed to bear a strong similarity with the construct referred to as `Achieve-ment Motivation'' (Liebert and Spiegler, 1994). Given that ``achievement'' has traditionallybeen a highly masculine sex-stereotyped activity (McClelland, 1965), it is not surprisingto find that, overall, males scored higher on conscientiousness than females. However, it isinteresting to note that an interaction effect was found between age and sex on theconscientiousness factor, such that older female adolescents scored higher on conscientious-ness than older male adolescents, whereas younger female adolescents scored lower thanyounger males. This suggests a possible lessening or changing of sex-role stereotyping withrespect to this variable, which is influencing young women in the later years of secondaryschooling.

Consistent with our hypothesis that personality traits would be significantly correlatedwith risk judgements, our results showed that high scores on agreeableness and extraversionwere associated with low scores on risk judgements, such that those adolescents who weremore extraverted and more agreeable were also less likely to judge the behaviours depicted inthe ARQ as risky. It was also found that adolescents who scored higher on conscientiousnesswere likely to judge all types of risk, with the exception of thrill-seeking, as riskier than theirless conscientious peers.

Also supporting our hypothesis, judgements of risk relating to the different risk types weresignificantly and negatively associated with reported risk behaviours. This was particularlythe case for antisocial, rebellious and reckless risks for which correlations were stronger thanfor thrill-seeking risks.

In order to bring these two sets of findings together, we investigated the relativepredictability of risk judgements and personality traits for risk behaviours. These factors weremost successful in predicting rebellious risks, since one third of the variance in rebelliousbehaviours was explained by judgements of the riskiness of the behaviours in conjunctionwith conscientiousness, agreeableness and age. Specifically, judgements that the behaviourswere highly risky, high levels of conscientiousness, low levels of agreeableness and youngerage predicted lower frequencies of rebellious risk-taking.

404 E. Gullone and S. Moore

Between one quarter and one fifth of the variance in reckless and antisocial riskbehaviours was explained by risk judgements and personality traits. However, whereas agewas important for antisocial risks, sex was important for reckless risks. Moreover, thepersonality trait concerned differed for the two types of risks. Specifically, those who reportedengaging more frequently in antisocial behaviours were most likely to judge those behavioursas less risky. They were also likely to score lower on neuroticism and to be older than thosewho reported lower frequencies of antisocial behaviours. Similarly, adolescents reporting highfrequencies of reckless risk engagement were likely to perceive the behaviours characteristicof this risk type as less risky. They were also likely to be male and to be less conscientious.

In contrast to the other risk types for which risk judgements were the most importantfactor in explaining variance in behaviour, extraversion was a significant predictor of thrill-seeking behaviours. Interestingly, for this risk type, risk judgements were not found to besignificant at all. However, extraversion and age together explained less than 10 percent ofthe variance. Therefore, although being extraverted and older were found to play some rolein thrill-seeking behaviours, other more important factors remain unaccounted for.

Collectively, the present findings have indicated that age and sex are important factors forrisk judgements, risk behaviours and personality traits, such that there are independent ageand sex trends for each. On the whole, boys and older adolescents were more likely to ratedifferent risk types as less risky than girls or younger adolescents. Also, the more extravertedand agreeable but less conscientious adolescents were less likely to rate the differentbehaviour types depicted in the ARQ as risky. Despite the fact that personality traits werefound to significantly predict risk behaviour, risk judgements appear to be more important, inthat their relationship with risk behaviours was generally stronger and more pervasive. This isconsistent with the findings of past research (e.g. Benthin et al., 1993; Lavery and Siegel,1993; Moore and Gullone, 1996). For example, Lavery and Siegel (1993) found that risk-taking involvement and risk perception were negatively related.

The findings regarding personality are consistent with what would be expected on the basisof the FFM trait descriptions. For example, extraverted adolescents are more likely to seekexcitement (i.e. high in sensation seekingÐcf. Arnett, 1992), hence engage in thrill-seekingbehaviours more often. Adolescents high in agreeableness are more trusting and perhaps moresusceptible to peer pressure. Indeed, at an extreme level, high agreeableness has beendescribed as manifesting in self-effacing and dependent behaviour (Costa and McCrae, 1988).Likewise, given the nature of the neuroticism disposition (i.e. worrisome, anxious; Parkes,1986; Jorm, 1987), it is not inconsistent that adolescents who scored higher on this trait werefound to be less likely to engage in antisocial risks. Finally, conscientiousness, which describesa hardworking, ambitious and energetic individual who tends to run with society rather thanagainst it (Digman, 1989; Dollinger and Orf, 1991), also predicted less engagement inrebellious and reckless risks. The only personality factor which did not significantly predictrisk behaviour was the openness to experience factor. This is somewhat surprising given thatindividuals who score high on openness are described as daring and tolerant of uncertainty,among other things (McCrae, 1990). The limitations of the study may help to explain whyopenness was not found to be significant in adolescent risk-taking.

Limitations of the present investigation that require mention include the use of the NEO-FFI to assess personality. This is a measure which has been extensively used with adults butless so with adolescents. As such, it is possible that certain NEO-FFI items may not havebeen as relevant or reliable with adolescents. This may explain why openness to experiencewas not found to be a significant predictor of adolescent risk-taking behaviour, despite its

Adolescent risk-taking and the five-factor model of personality 405

intuitive link. Future research should attempt to replicate the present findings through theuse of an adolescent personality measure. However, given the absence of as comprehensive ameasure for adolescents as the NEO-FFI, this may prove difficult.

Secondly, this research was carried out entirely through self-report. A question thatremains to be answered is whether the consistency between self-reported risk judgementsand the self-reported frequency of engaging in different risk behaviours is merely an artefactof our research methodology. That is, are adolescents merely responding to the differentrating scales in consistent ways? In essence, are the self-reported risk behaviours a reflectionof actual behaviours or are they influenced by social desirability responding? The fact that inpast research (i.e. Gullone et al., 2000) we have found good consistency between test±retestadministrations of the ARQ provides some support for the validity of the adolescent reportsas indicators of actual behaviour.

Conclusions

Three important conclusions can be drawn from these results. Firstly, adolescents' riskjudgements have been shown to be significantly predictive of risk behaviours. This was foundto be the case for risk behaviour generally, as well as different risk types, with the exception ofthrill-seeking. Secondly, risk judgements were found to generally be better predictors of riskbehaviour than personality traits, particularly in relation to reckless and rebellious risks. Ofcourse, it is not possible to determine whether the strong relationship between judgementsand behaviour is a function of judgements influencing behaviour, or the converse. Asproposed by Benthin et al. (1993), adolescents' behaviour may, in fact, drive theirperceptions, rather than the reverse. According to these authors, cognitive dissonance theoryis a probable explanation for the high relationship between judgements and behaviour. Inorder to elucidate possible underlying mechanisms, future research should examine whetherjudgement changes brought about through structured intervention predict correspondingchanges in behaviour.

A final important conclusion that can be made is that the factors underlying participationin positive (i.e. thrill-seeking) versus negative (e.g. reckless) risks appear to be different, suchthat adolescents most likely to engage in positive risk-taking behaviours were extraverted,and these behaviours were more strongly predicted by this than by the judgements of riskthey made relating to these behaviours. Nonetheless, neither judgements nor personalitytraits were found to be particularly good predictors of thrill-seeking behaviour. In contrast,other types of risk-taking behaviours were highly related to the judgements that adolescentsheld in relation to them.

Notwithstanding its limitations, this study is the first to apply a comprehensive model ofindividual differences for the prediction of adolescent risk-taking judgements and behaviour.We have shown that the five factor model of personality, in conjunction with riskjudgements, is useful for making predictions about adolescent risk behaviours.

Acknowledgements

We would like to thank the Australian Rotary Health Research Fund by whom this researchwas funded. We would also like to express our thanks to the Directorate of School Education

406 E. Gullone and S. Moore

for its approval of the research and to all the students, teachers and schools involved in theproject. This investigation would not have been possible without their cooperation andassistance.

References

Alexander, C. S., Kim, Y. J., Ensminger, M., Johnson, K. E., Smith, B. J. and Dolan, L. J. (1990). Ameasure of risk-taking for young adolescents: reliability and validity assessments. Journal of Youthand Adolescence, 19, 559±569.

Arnett, J. (1992). Review: Reckless behaviour in adolescence: a developmental perspective.Developmental Review, 12, 339±373.

Benet, V. and Waller, N. G. (1995). The big seven factor model of personality description: evidence forits cross-cultural generality in a Spanish sample. Journal of Personality and Social Psychology, 69,701±718.

Benthin, A., Slovic, P. and Severson, H. (1993). A psychometric study of adolescent risk perception.Journal of Adolescence, 16, 153±168.

Block, J., Block, J. H. and Keyes, S. (1988). Longitudinally foretelling drug usage in adolescence: earlychildhood personality and environmental precursors. Child Development, 59, 336±366.

Chassin, L., Presson, C. C. and Sherman, S. (1989). Constructive vs destructive deviance in adolescenthealth-related behaviours. Journal of Youth and Adolescence, 18, 245±262.

Conley, J. J. (1985). Longitudinal stability of personality traits: a multitrait-multimethod-multioccasionanalysis. Journal of Personality and Social Psychology, 49, 1266±1282.

Costa, P. T. Jr and McCrae, R. R. (1980). Influence of extraversion and neuroticism on subjective well-being: happy and unhappy people. Journal of Personality and Social Psychology, 38, 668±678.

Costa, P. T. Jr and McCrae, R. R. (1988). From catalog to classification: Murray's needs and the five-factor model. Journal of Personality and Social Psychology, 55, 258±265.

Costa, P. T. Jr and McCrae, R. R. (1992). Revised NEO Personality Inventory (NEO PI-R) and NEO Five-Factor Inventory (NEO-FFI) professional manual. Odessa, FL: Psychological Assessment Resources,Inc.

Digman, J. M. (1989). Five robust trait dimensions: development, stability, and utility. Journal ofPersonality, 57, 195±214.

Digman, J. M. and Inouye, J. (1986). Further specification of the five robust factors of personality.Journal of Personality and Social Psychology, 50, 116±123.

Dollinger, S. J. and Orf, L. A. (1991). Personality and performance in ``personality'': conscientiousnessand openness. Journal of Research in Personality, 25, 276±284.

Dworkin, R. H., Burke, B. W. and Maher, B. A. (1976). A longitudinal study of the genetics ofpersonality. Journal of Personality and Social Psychology, 34, 510±518.

Eysenck, H. J. (1990). Genetic and environmental contributions to individual differences: the threemajor dimensions of personality. Journal of Personality, 58, 245±261.

Garmezy, N. (1983). Stressors of childhood. In Garmezy, N. and Rutter, M. (Eds), Stress, coping, anddevelopment in children. New York: McGraw-Hill.

Goldberg, L. R. (1990). An alternative ``Description of personality'': the Big-Five factor structure.Journal of Personality and Social Psychology, 59, 1216±1229.

Goldberg, L. R. (1992). The development of markers of the Big-Five factor structure. PsychologicalAssessment, 4, 26±42.

Goldberg, L. R. (1993). The structure of phenotypic personality traits. American Psychologist, 48, 26±34.Gorsuch, R. L. and Butler, M. C. (1976). Initial drug abuse: a review of psychological factors.

Psychological Bulletin, 83, 120±137.Gullone, E., Moore, S. M., Moss, S. and Boyd, C. P. (2000). The Adolescent Risk-Taking Questionnaire

(ARQ): development and psychometric evaluation. Journal of Adolescent Research, 15, 231±250.Jessor, S. L. and Jessor, R. (1977). Problem behavior and psychosocial development: a longitudinal study of

youth. New York: Academic Press.Jorm, A. F. (1987). Sex differences in neuroticism: a quantitative synthesis of published research.

Australian and New Zealand Journal of Psychiatry, 21, 501±506.

Adolescent risk-taking and the five-factor model of personality 407

Kandel, D. B. (1980). Drug and drinking behavior among youth. Annual Review of Sociology, 6,235±285.

Lavery, B. and Siegel, A. W. (1993). Adolescent risk-taking: an analysis of problem behaviors inproblem children. Journal of Experimental Child Psychology, 55, 277±294.

Liebert, R. M. and Spiegler, M. D. (1994). Personality: strategies and issues (7th ed.). Pacific Grove, CA:Brooks Cole.

McClelland, D. C. (1965). Toward a theory of motive acquisition. American Psychologist, 20, 321±333.McCrae, R. R. (1990). Traits and trait names: how well is openness represented in natural languages?

European Journal of Personality, 4, 119±129.McCrae, R. R. and Costa, P. T. Jr (1987). Validation of the five-factor model of personality across

instruments and observers. Journal of Personality and Social Psychology, 52, 81±90.McCrae, R. R., Costa, P. T. and Yik, M. S. M. (1996). Universal aspects of Chinese personality

structure. In Bond, M. H. (Ed.), The handbook of Chinese psychology. Hong Kong: OxfordUniversity Press.

Moore, S. and Gullone, E. (1996). Predicting adolescent risk behavior using a personalized cost-benefitanalysis. Journal of Youth and Adolescence, 25, 343±359.

Moore, S. M., Gullone, E. and Kostanski, M. (1997). An examination of adolescent risk-taking using astory completion task. Journal of Adolescence, 20, 369±379.

Moore, S. and Rosenthal, D. (1993).Venturesomeness, impulsiveness, and risky behaviour among olderadolescents. Perceptual and Motor Skills, 76, 98.

Parkes, K. R. (1986). Coping in stressful episodes: the role of individual differences, environmentalfactors, and situational characteristics. Journal of Personality and Social Psychology, 51, 1277±1292.

Silbereisen, R. K. and Noack, P. (1988). On the constructive role of problem behaviour in adolescence/childhood. In Bolger, N., Caspi, A., Downey, G. and Moorehouse, M. (Eds), Person and context:developmental process. Cambridge: Cambridge University Press.

Werner, E. E. (1986). Resilient offspring of alcoholics: a longitudinal study from birth to age 18. Journalof Studies in Alcoholism, 47, 34±40.

Zuckerman, M. (1979). Sensation seeking: beyond the optimal level of arousal. Hillsdale, NJ: Erlbaum.