OR I G I N A L ART I C L E

A worthy self is a caring self Examining the developmentalrelations between self-esteem and self-compassion in adolescents

James N Donald | Joseph Ciarrochi | Philip D Parker |

Baljinder K Sahdra | Sarah L Marshall | Jiesi Guo

Institute for Positive Psychology andEducation Australian Catholic University

CorrespondenceJames N Donald Institute for PositivePsychology and Education AustralianCatholic University 25A Barker RoadStrathfield NSW 2135 AustraliaEmail jamesndonaldgmailcom

Funding informationGrant DP140103874 from the AustralianResearch Council

Abstract

Objective Self-compassion has been framed as a healthy alternative to self-esteemas it is nonevaluative However rather than being alternatives it may be that the twoconstructs develop in a mutually reinforcing way The present study tested this possi-bility among adolescents

Method A large adolescent sample (N5 2809 498 female) reported levels oftrait self-esteem and self-compassion annually for 4 years Autoregressive cross-lagged structural equation models were used to estimate the reciprocal longitudinalrelations between the two constructs

Results Self-esteem consistently predicted changes in self-compassion across the 4years of the study but not vice versa

Conclusions Self-esteem appears to be an important antecedent of the developmentof self-compassion perhaps because the capacity to extend compassion toward theself depends on onersquos appraisals of worthiness These findings add important insightsto our theoretical understanding of the development of self-compassion

KEYWORD S

adolescent development longitudinal self-compassion self-esteem structural equation modeling

1 | INTRODUCTION

The worst loneliness is to not be comfortable with yourselfmdashMark Twain

Research on self-esteem over the past five decades high-lights the benefits as well as the risks of self-evaluativebehavior (Baumeister Campbell Krueger amp Vohs 2003)For example self-esteem has been found to be an importantantecedent of the development of social networks (MarshallParker Ciarrochi amp Heaven 2013) but seeking to boostself-esteem noncontingently may lead to narcissism and anti-social behavior (Baumeister Heatherton amp Tice 1993 Bau-meister et al 2000 Crocker amp Park 2004) More recentlyself-compassion has been proposed as an alternative to self-esteem pursuit as it emphasizes self-acceptance rather than

self-evaluation and thereby avoids some of the risks ofattachment to positive self-concept (Neff amp Vonk 2009)

However rather than being alternatives it may be thatself-esteem and self-compassion are mutually reinforcing intheir development For example it may be that self-compassion helps people not link failures and mistakes totheir self-worth and therefore is an antecedent to the develop-ment of self-esteem (the antecedent model) Or perhaps oneneeds to evaluate oneself as worthy (ie have relatively highself-esteem) in order to extend compassion toward the self(the consequence model) It is also possible that self-esteemand self-compassion influence each other over time (the recip-rocal model) To our knowledge there are no published longi-tudinal studies that either explicitly propose or systematicallytest these alternate models of the development of self-esteem

Journal of Personality 20171ndash12 wileyonlinelibrarycomjournaljopy VC 2017 Wiley Periodicals Inc | 1

Received 9 March 2017 | Revised 6 June 2017 | Accepted 13 August 2017

DOI 101111jopy12340

and self-compassion This omission seems surprising giventhe research interest in the link between self-esteem and self-compassion in the last several years (eg Barry Loflin ampDoucette 2015 Leary Tate Adams Allen amp Hancock2007 Muris Meesters Pierik amp De Kock 2016 Neff ampMcGehee 2010 Neff amp Vonk 2009) The present study aimsto fill this gap in the literature by utilizing a large four-wavestudy of adolescent development to test the three models ofthe link between self-esteem and self-compassion

11 | The costs and benefits of self-esteempursuit

Self-esteem has been defined as positive or negative self-directed evaluations (Rosenberg Schooler Schoenbach ampRosenberg 1995) Self-esteem describes an individualrsquosassessment of his or her global self-worth and as such differsfrom self-concept which tends to be more domain specific(Marsh amp Craven 2006) Global self-esteem has been shownto be an important predictor of social and emotional well-being For example self-esteem has been found to predictgreater happiness (Cheng amp Furnham 2004) positive affect(Orth Robins amp Widaman 2012) and social support (Mar-shall et al 2013) and less depression and anxiety (Sowisloamp Orth 2013) Conversely having low global self-esteem asan adolescent has been linked with antisocial behavior anddelinquency (Donnellan Trzesniewski Robins Moffitt ampCaspi 2005) as well as poor physical health and antisocialbehavior in adulthood (Trzesniewski et al 2006)

Despite the value of having a high and relatively stableglobal self-esteem a number of scholars have highlighted thepotential drawbacks and vulnerabilities associated with self-esteem pursuit (Baumeister Smart amp Boden 1996 Baumeisteret al 2003 Crocker amp Park 2004 Neff amp Vonk 2009 Ryanamp Brown 2003) A key concern is that because self-esteem isself-evaluative it brings with it a contingent sense of self that isinherently vulnerable to threat (Baumeister et al 2003 Neff ampVonk 2009) Assessments of oneself as being a worthy anddeserving individual can easily be undermined by contraryfeedback from others at times precipitating defensive or narcis-sistic responses in an effort to reinforce the self (Baumeisteret al 1993 2003 Crocker amp Park 2004) Among young peo-ple this translates to the tendency to pass up valuable learningopportunities following failure in the service of self-esteemmaintenance and the tendency to unrealistically inflate andeven defend positive qualities in an effort to restore lost self-esteem (Baumeister et al 2003 Nussbaum amp Dweck 2008)

12 | Self-compassion as an alternativeto self-esteem pursuit

Self-compassion has been defined as comprising three corebehaviors mindfully noticing onersquos suffering practicing

self-kindness in the presence of suffering and recognizingthat suffering is common to all humanity (Neff 2003) Self-compassion has been shown to be more weakly correlatedwith traits such as narcissism anger and public self-consciousness than self-esteem (Neff amp Vonk 2009) More-over self-compassion has been shown to predict unique var-iance in outcomes such as anger public self-consciousnessand rumination beyond the effects of self-esteem (Neff ampVonk 2009) As such self-compassion has been framed asan alternative to self-esteem pursuit (Neff 2003)

Despite considerable research interest in self-compassionover the past decade and the very large literature on self-esteem there has been no developmental research to ourknowledge examining the link between self-esteem and self-compassion Several cross-sectional studies have found self-compassion and self-esteem to be moderately to highly cor-related including among young people perhaps as both con-structs involve extending positive emotions toward oneself(Barry et al 2015 Leary et al 2007 Muris et al 2016Neff amp Vonk 2009) Indeed these findings suggest that thedevelopment of self-esteem and self-compassion may bemore closely linked than has been previously thought

13 | Mechanism of influence

Rather than being opposed to each other it may be that self-compassion and self-esteem influence each other across thedevelopmental process The self-compassion as antecedentmodel suggests that self-compassionate people forgive them-selves for mistakes and failures and such forgiveness isexpected to lead to more positive evaluations of self-worth(Neff amp Vonk 2009 Reilly Rochlen amp Awad 2014) Theself-compassion as consequence model suggests that believ-ing you are a person of worth (ie self-esteem) is a precursorto feeling worthy of self-compassion and to the developmentof self-compassion Finally the reciprocal influence modelsuggests that both self-esteem and self-compassion are self-reinforcing with an increase in one leading to an increase inthe other creating a positive spiral of development

Over the past decade several studies have examined therelations between self-esteem and self-compassion cross-sec-tionally Studies of adults have found correlations betweenglobal self-esteem and self-compassion from 50 to 70 (egNeff amp Vonk 2009 Reilly et al 2014) whereas amongadolescent samples the constructs have been found to becorrelated in the 40 to 50 range (eg Barry et al 2015Muris et al 2016) In addition the two constructs have beenshown to be correlated with a range of outcomes in relativelysimilar ways For example Neff and Vonk (2009) found thatglobal self-esteem and self-compassion were both correlatedwith a number of self- and affect-related outcomes (egsocial comparison orientation public self-consciousness

2 | DONALD ET AL

self-rumination narcissism need for cognitive closure andanger) in the same direction and to a similar degree Morerecent studies have found consistently similar-sized correla-tions between these two constructs and other traits includingshame (Reilly et al 2014) aggression and depression (Barryet al 2015) and anxiety depression and self-efficacy(Muris et al 2016)

Together this evidence suggests that self-esteem andself-compassion predict similar kinds of outcomes in similarways However to our knowledge the relations betweenthese two constructs have not been tested in a longitudinalmanner making it impossible to identify the temporal order-ing between the two In the following sections we explorethe theoretical rationale for these underexplored developmen-tal paths

14 | The antecedent consequence andreciprocal influence models

Concerning self-compassion as an antecedent to the develop-ment of self-esteem we expect that self-compassion willlead to more positive (and fewer negative) self-evaluationsSelf-compassion involves cultivating kind self-directedthoughts and emotions in the face of onersquos own suffering(Neff Kirkpatrick amp Rude 2007) Among adolescentsgreater self-compassion has been associated with less depres-sion and anxiety (Neff amp McGehee 2010) both of whichare associated with negative self-directed evaluations (Sowi-slo amp Orth 2013) In addition self-compassion has beenfound to be negatively associated with other negative self-evaluations such as self-rumination and social comparisonorientation (Neff amp Vonk 2009) and shame (Reilly et al2014) Self-compassionate adolescents are more forgiving oftheir faults and failures and are less harsh on themselves inresponse to setbacks and stress (Neff Hsieh amp Dejitterat2005) Our first prediction was therefore that self-compassion will facilitate the development of self-esteemamong adolescents

Alternatively self-compassion might be a consequenceof self-esteem There is evidence that at times people feelinsecure or fearful about extending compassion to them-selves (Gilbert McEwan Matos amp Rivis 2011 KellyVimalakanthan amp Carter 2014 Pauley amp McPherson2010) That is people sometimes feel as if they do notdeserve self-compassion and that extending self-compassionto themselves will result in their ldquoflawsrdquo being exposed (Gil-bert et al 2011) Consistent with this notion research onother-oriented compassion has highlighted the role ofappraisals of the other person as being worthy of compassionas a key element of the act of extending compassion toanother (Atkins amp Parker 2012) and this same process mayapply to acts of compassion toward oneself Research

therefore suggests that believing oneself to be worthwhileand acceptable are important antecedents of the capacity toextend compassion to oneself (Gilbert et al 2011) This inturn suggests that individuals who experience low self-esteem are less likely to extend compassion to themselveswhich leads to our second prediction that among adolescentsself-esteem will precede increases in self-compassion overtime

Finally it may be that self-esteem and self-compassionare mutually reinforcing in their development Self-esteemhas been shown to be a cause as well as a consequence ofmental health outcomes such as depression and anxietysymptoms (Sowislo amp Orth 2013) and it may be that a simi-lar process of mutual reinforcement applies to the relationsbetween self-esteem and self-compassion This explanationis consistent with theories of emotional development such asthe broaden-and-build hypothesis of positive emotions inwhich positive emotions are said to mutually reinforce oneanother in an upward spiral (Fredrickson 2001) This leadsto our third and final hypothesis that self-esteem and self-compassion each reciprocally influence the otherrsquosdevelopment

In addition to the hypothesized links between self-esteemand self-compassion we also expected that a number ofdemographic variables would be associated with both self-esteem and self-compassion and may therefore be confoundsin testing the above hypotheses We identified three suchdemographic variables namely gender age and parentsrsquomarital status Regarding gender there is consistent evidencethat males have higher trait self-esteem than females withthe largest differences occurring in late adolescence (KlingHyde Showers amp Buswell 1999) and that males are moreself-compassionate than females (Neff 2003 Neff amp Vonk2009) Further there is evidence that individualsrsquo self-esteemincreases during adolescence (Twenge amp Campbell 2001)and we expect a similar pattern for self-compassion as indi-viduals tend to become more resilient and tolerant of set-backs with age (Goyen amp Anshel 1998) Lastly there ismeta-analytic evidence that self-esteem is lower among chil-dren of separated or divorced parents (Amato amp Keith1991) and that self-compassion is lower among adolescentswho believe their families are not functioning well (Neff ampMcGehee 2010)

2 | METHOD

21 | Participants and procedure

This study was a part of the Australian Character Study amultiyear program of research among high school studentsin Australia that collected a range of information relating toadolescent behaviors relationships beliefs aspirations and

DONALD ET AL | 3

self-evaluations Participants in the current study attended 17Catholic high schools in two Australian states Catholicschools in Australia account for 2052 of secondaryschools (Australian Bureau of Statistics 2012) The schoolsparticipating in this study were concentrated in the cities ofWollongong (New South Wales) and Cairns (Queensland)but the study also included schools within regional and ruralareas thereby ensuring the socioeconomic and cultural diver-sity of participants The Australian Governmentrsquos socioeco-nomic index for schools sets the Australian average at 1000(httpbitly1mJK7KC) The schools in the present studyhad a socioeconomic ranking almost identical to the Austra-lian average (1025 SD5 43) meaning this sample wasbroadly reflective of the socioeconomic status of schoolsacross Australia

Participants completed measures for this study in thethird term of a four-term year in each of the 4 years of thestudy from Year 9 to Year 12 The total sample consisted of2809 participants (1395 or 497 male 1399 or 498female 15 unknown) Of the total sample 1683 participantsin Year 9 completed the measures for the present study1745 participants did so in Year 10 1530 did so in Year11 and 1525 did so in Year 12 A total of 647 subjects com-pleted the measures for this study across all 4 years of thestudy Participantsrsquo mean age was 147 years (SD5 45) inYear 9 In Year 9 73 of participantsrsquo parents were married10 were separated 13 were divorced and 4 of the sam-ple did not specify This ratio was consistent across the fourwaves of the study In Year 12 72 of participantsrsquo parentswere married 8 were separated 14 were divorced and6 of the sample did not specify Ethics approval wasgranted by the university and informed consent wasobtained from study participants

22 | Measures

221 | Self-esteem

Global trait self-esteem was measured using the 10-itemRosenberg Self-Esteem scale (RSE Rosenberg 1979) Par-ticipants were asked to indicate their agreement with state-ments such as ldquoGenerally I feel satisfied with myselfrdquo and ldquoIthink that I am a failurerdquo using a binary response scale (yesor no) The binary response scale used in this version of theRSE measure has been validated in previous research andhas been found to have internal consistency as good as orstronger than the 4-point version of the measure (HeavenCiarrochi amp Hurrell 2010 Marshall et al 2013) Cron-bachrsquos alphas for this scale across the 4 study years were 86(Year 9) 88 (Year 10) 88 (Year 11) and 86 (Year 12)

222 | Self-compassion

Self-compassion was measured using the 12-item short formof the Self-Compassion Scale (Raes Pommier Neff amp VanGucht 2011) utilizing a 5-point Likert style (15 almostnever 55 almost always) The scale includes positive andnegative items A sample positive item is ldquoI try to be under-standing and patient towards those aspects of my personalityI donrsquot likerdquo and a sample negative item is ldquoWhen I fail atsomething thatrsquos important to me I tend to feel alone in myfailurerdquo Cronbachrsquos alphas for this scale across the 4 studyyears were 78 (Year 9) 81 (Year 10) 84 (Year 11) and81 (Year 12)

223 | Demographic covariates

The three demographic covariates examined in the presentstudy were participant gender participant age and parentsrsquomarital status Gender was coded as male female or notspecified age was age in years at the time of data collectionand parental marital status was a categorical variable speci-fied as married separated divorced and other

23 | Statistical Analyses

231 | Autoregressive cross-lagged models

We employed autoregressive cross-lagged (ACL) models toexamine the relations between self-esteem and self-compassion across the 4 years of this study An ACLapproach enables one to identify the likely temporal orderingof changes in phenomena across time (Ciarrochi et al2016) In the present study this approach enabled us to testwhether (a) self-esteem precedes the development of self-compassion (ie an antecedent model) (b) self-compassionprecedes the development of self-esteem (ie a consequencemodel) or (c) the development of both constructs is mutuallyreinforcing (ie a reciprocal influence model)

To test these alternative possibilities across the 4 years ofthe study we ran a series of structural equation models(SEMs) in the R program (R Core Team 2017) using thelavaan package (Rosseel 2012) All analyses were con-ducted using latent variables for self-esteem and self-compassion following similar approaches elsewhere (egCiarrochi Parker Kashdan Heaven amp Barkus 2015 Mar-shall et al 2013) A key advantage of a latent SEMapproach is that it enables one to control for measurementerror (Weston amp Gore 2006)

The data for this study had a nested structure with the2809 students nested within 17 schools As our predictionsrelated to individual-level effects we needed to control forclustering effects To do this we used a no-poolingapproach in which each of the 17 schools was included in

4 | DONALD ET AL

self-rumination narcissism need for cognitive closure andanger) in the same direction and to a similar degree Morerecent studies have found consistently similar-sized correla-tions between these two constructs and other traits includingshame (Reilly et al 2014) aggression and depression (Barryet al 2015) and anxiety depression and self-efficacy(Muris et al 2016)

Together this evidence suggests that self-esteem andself-compassion predict similar kinds of outcomes in similarways However to our knowledge the relations betweenthese two constructs have not been tested in a longitudinalmanner making it impossible to identify the temporal order-ing between the two In the following sections we explorethe theoretical rationale for these underexplored developmen-tal paths

14 | The antecedent consequence andreciprocal influence models

Concerning self-compassion as an antecedent to the develop-ment of self-esteem we expect that self-compassion willlead to more positive (and fewer negative) self-evaluationsSelf-compassion involves cultivating kind self-directedthoughts and emotions in the face of onersquos own suffering(Neff Kirkpatrick amp Rude 2007) Among adolescentsgreater self-compassion has been associated with less depres-sion and anxiety (Neff amp McGehee 2010) both of whichare associated with negative self-directed evaluations (Sowi-slo amp Orth 2013) In addition self-compassion has beenfound to be negatively associated with other negative self-evaluations such as self-rumination and social comparisonorientation (Neff amp Vonk 2009) and shame (Reilly et al2014) Self-compassionate adolescents are more forgiving oftheir faults and failures and are less harsh on themselves inresponse to setbacks and stress (Neff Hsieh amp Dejitterat2005) Our first prediction was therefore that self-compassion will facilitate the development of self-esteemamong adolescents

Alternatively self-compassion might be a consequenceof self-esteem There is evidence that at times people feelinsecure or fearful about extending compassion to them-selves (Gilbert McEwan Matos amp Rivis 2011 KellyVimalakanthan amp Carter 2014 Pauley amp McPherson2010) That is people sometimes feel as if they do notdeserve self-compassion and that extending self-compassionto themselves will result in their ldquoflawsrdquo being exposed (Gil-bert et al 2011) Consistent with this notion research onother-oriented compassion has highlighted the role ofappraisals of the other person as being worthy of compassionas a key element of the act of extending compassion toanother (Atkins amp Parker 2012) and this same process mayapply to acts of compassion toward oneself Research

therefore suggests that believing oneself to be worthwhileand acceptable are important antecedents of the capacity toextend compassion to oneself (Gilbert et al 2011) This inturn suggests that individuals who experience low self-esteem are less likely to extend compassion to themselveswhich leads to our second prediction that among adolescentsself-esteem will precede increases in self-compassion overtime

Finally it may be that self-esteem and self-compassionare mutually reinforcing in their development Self-esteemhas been shown to be a cause as well as a consequence ofmental health outcomes such as depression and anxietysymptoms (Sowislo amp Orth 2013) and it may be that a simi-lar process of mutual reinforcement applies to the relationsbetween self-esteem and self-compassion This explanationis consistent with theories of emotional development such asthe broaden-and-build hypothesis of positive emotions inwhich positive emotions are said to mutually reinforce oneanother in an upward spiral (Fredrickson 2001) This leadsto our third and final hypothesis that self-esteem and self-compassion each reciprocally influence the otherrsquosdevelopment

In addition to the hypothesized links between self-esteemand self-compassion we also expected that a number ofdemographic variables would be associated with both self-esteem and self-compassion and may therefore be confoundsin testing the above hypotheses We identified three suchdemographic variables namely gender age and parentsrsquomarital status Regarding gender there is consistent evidencethat males have higher trait self-esteem than females withthe largest differences occurring in late adolescence (KlingHyde Showers amp Buswell 1999) and that males are moreself-compassionate than females (Neff 2003 Neff amp Vonk2009) Further there is evidence that individualsrsquo self-esteemincreases during adolescence (Twenge amp Campbell 2001)and we expect a similar pattern for self-compassion as indi-viduals tend to become more resilient and tolerant of set-backs with age (Goyen amp Anshel 1998) Lastly there ismeta-analytic evidence that self-esteem is lower among chil-dren of separated or divorced parents (Amato amp Keith1991) and that self-compassion is lower among adolescentswho believe their families are not functioning well (Neff ampMcGehee 2010)

2 | METHOD

21 | Participants and procedure

This study was a part of the Australian Character Study amultiyear program of research among high school studentsin Australia that collected a range of information relating toadolescent behaviors relationships beliefs aspirations and

DONALD ET AL | 3

self-evaluations Participants in the current study attended 17Catholic high schools in two Australian states Catholicschools in Australia account for 2052 of secondaryschools (Australian Bureau of Statistics 2012) The schoolsparticipating in this study were concentrated in the cities ofWollongong (New South Wales) and Cairns (Queensland)but the study also included schools within regional and ruralareas thereby ensuring the socioeconomic and cultural diver-sity of participants The Australian Governmentrsquos socioeco-nomic index for schools sets the Australian average at 1000(httpbitly1mJK7KC) The schools in the present studyhad a socioeconomic ranking almost identical to the Austra-lian average (1025 SD5 43) meaning this sample wasbroadly reflective of the socioeconomic status of schoolsacross Australia

Participants completed measures for this study in thethird term of a four-term year in each of the 4 years of thestudy from Year 9 to Year 12 The total sample consisted of2809 participants (1395 or 497 male 1399 or 498female 15 unknown) Of the total sample 1683 participantsin Year 9 completed the measures for the present study1745 participants did so in Year 10 1530 did so in Year11 and 1525 did so in Year 12 A total of 647 subjects com-pleted the measures for this study across all 4 years of thestudy Participantsrsquo mean age was 147 years (SD5 45) inYear 9 In Year 9 73 of participantsrsquo parents were married10 were separated 13 were divorced and 4 of the sam-ple did not specify This ratio was consistent across the fourwaves of the study In Year 12 72 of participantsrsquo parentswere married 8 were separated 14 were divorced and6 of the sample did not specify Ethics approval wasgranted by the university and informed consent wasobtained from study participants

22 | Measures

221 | Self-esteem

Global trait self-esteem was measured using the 10-itemRosenberg Self-Esteem scale (RSE Rosenberg 1979) Par-ticipants were asked to indicate their agreement with state-ments such as ldquoGenerally I feel satisfied with myselfrdquo and ldquoIthink that I am a failurerdquo using a binary response scale (yesor no) The binary response scale used in this version of theRSE measure has been validated in previous research andhas been found to have internal consistency as good as orstronger than the 4-point version of the measure (HeavenCiarrochi amp Hurrell 2010 Marshall et al 2013) Cron-bachrsquos alphas for this scale across the 4 study years were 86(Year 9) 88 (Year 10) 88 (Year 11) and 86 (Year 12)

222 | Self-compassion

Self-compassion was measured using the 12-item short formof the Self-Compassion Scale (Raes Pommier Neff amp VanGucht 2011) utilizing a 5-point Likert style (15 almostnever 55 almost always) The scale includes positive andnegative items A sample positive item is ldquoI try to be under-standing and patient towards those aspects of my personalityI donrsquot likerdquo and a sample negative item is ldquoWhen I fail atsomething thatrsquos important to me I tend to feel alone in myfailurerdquo Cronbachrsquos alphas for this scale across the 4 studyyears were 78 (Year 9) 81 (Year 10) 84 (Year 11) and81 (Year 12)

223 | Demographic covariates

The three demographic covariates examined in the presentstudy were participant gender participant age and parentsrsquomarital status Gender was coded as male female or notspecified age was age in years at the time of data collectionand parental marital status was a categorical variable speci-fied as married separated divorced and other

23 | Statistical Analyses

231 | Autoregressive cross-lagged models

We employed autoregressive cross-lagged (ACL) models toexamine the relations between self-esteem and self-compassion across the 4 years of this study An ACLapproach enables one to identify the likely temporal orderingof changes in phenomena across time (Ciarrochi et al2016) In the present study this approach enabled us to testwhether (a) self-esteem precedes the development of self-compassion (ie an antecedent model) (b) self-compassionprecedes the development of self-esteem (ie a consequencemodel) or (c) the development of both constructs is mutuallyreinforcing (ie a reciprocal influence model)

To test these alternative possibilities across the 4 years ofthe study we ran a series of structural equation models(SEMs) in the R program (R Core Team 2017) using thelavaan package (Rosseel 2012) All analyses were con-ducted using latent variables for self-esteem and self-compassion following similar approaches elsewhere (egCiarrochi Parker Kashdan Heaven amp Barkus 2015 Mar-shall et al 2013) A key advantage of a latent SEMapproach is that it enables one to control for measurementerror (Weston amp Gore 2006)

The data for this study had a nested structure with the2809 students nested within 17 schools As our predictionsrelated to individual-level effects we needed to control forclustering effects To do this we used a no-poolingapproach in which each of the 17 schools was included in

4 | DONALD ET AL

self-evaluations Participants in the current study attended 17Catholic high schools in two Australian states Catholicschools in Australia account for 2052 of secondaryschools (Australian Bureau of Statistics 2012) The schoolsparticipating in this study were concentrated in the cities ofWollongong (New South Wales) and Cairns (Queensland)but the study also included schools within regional and ruralareas thereby ensuring the socioeconomic and cultural diver-sity of participants The Australian Governmentrsquos socioeco-nomic index for schools sets the Australian average at 1000(httpbitly1mJK7KC) The schools in the present studyhad a socioeconomic ranking almost identical to the Austra-lian average (1025 SD5 43) meaning this sample wasbroadly reflective of the socioeconomic status of schoolsacross Australia

Participants completed measures for this study in thethird term of a four-term year in each of the 4 years of thestudy from Year 9 to Year 12 The total sample consisted of2809 participants (1395 or 497 male 1399 or 498female 15 unknown) Of the total sample 1683 participantsin Year 9 completed the measures for the present study1745 participants did so in Year 10 1530 did so in Year11 and 1525 did so in Year 12 A total of 647 subjects com-pleted the measures for this study across all 4 years of thestudy Participantsrsquo mean age was 147 years (SD5 45) inYear 9 In Year 9 73 of participantsrsquo parents were married10 were separated 13 were divorced and 4 of the sam-ple did not specify This ratio was consistent across the fourwaves of the study In Year 12 72 of participantsrsquo parentswere married 8 were separated 14 were divorced and6 of the sample did not specify Ethics approval wasgranted by the university and informed consent wasobtained from study participants

22 | Measures

221 | Self-esteem

Global trait self-esteem was measured using the 10-itemRosenberg Self-Esteem scale (RSE Rosenberg 1979) Par-ticipants were asked to indicate their agreement with state-ments such as ldquoGenerally I feel satisfied with myselfrdquo and ldquoIthink that I am a failurerdquo using a binary response scale (yesor no) The binary response scale used in this version of theRSE measure has been validated in previous research andhas been found to have internal consistency as good as orstronger than the 4-point version of the measure (HeavenCiarrochi amp Hurrell 2010 Marshall et al 2013) Cron-bachrsquos alphas for this scale across the 4 study years were 86(Year 9) 88 (Year 10) 88 (Year 11) and 86 (Year 12)

222 | Self-compassion

Self-compassion was measured using the 12-item short formof the Self-Compassion Scale (Raes Pommier Neff amp VanGucht 2011) utilizing a 5-point Likert style (15 almostnever 55 almost always) The scale includes positive andnegative items A sample positive item is ldquoI try to be under-standing and patient towards those aspects of my personalityI donrsquot likerdquo and a sample negative item is ldquoWhen I fail atsomething thatrsquos important to me I tend to feel alone in myfailurerdquo Cronbachrsquos alphas for this scale across the 4 studyyears were 78 (Year 9) 81 (Year 10) 84 (Year 11) and81 (Year 12)

223 | Demographic covariates

The three demographic covariates examined in the presentstudy were participant gender participant age and parentsrsquomarital status Gender was coded as male female or notspecified age was age in years at the time of data collectionand parental marital status was a categorical variable speci-fied as married separated divorced and other

23 | Statistical Analyses

231 | Autoregressive cross-lagged models

We employed autoregressive cross-lagged (ACL) models toexamine the relations between self-esteem and self-compassion across the 4 years of this study An ACLapproach enables one to identify the likely temporal orderingof changes in phenomena across time (Ciarrochi et al2016) In the present study this approach enabled us to testwhether (a) self-esteem precedes the development of self-compassion (ie an antecedent model) (b) self-compassionprecedes the development of self-esteem (ie a consequencemodel) or (c) the development of both constructs is mutuallyreinforcing (ie a reciprocal influence model)

To test these alternative possibilities across the 4 years ofthe study we ran a series of structural equation models(SEMs) in the R program (R Core Team 2017) using thelavaan package (Rosseel 2012) All analyses were con-ducted using latent variables for self-esteem and self-compassion following similar approaches elsewhere (egCiarrochi Parker Kashdan Heaven amp Barkus 2015 Mar-shall et al 2013) A key advantage of a latent SEMapproach is that it enables one to control for measurementerror (Weston amp Gore 2006)

The data for this study had a nested structure with the2809 students nested within 17 schools As our predictionsrelated to individual-level effects we needed to control forclustering effects To do this we used a no-poolingapproach in which each of the 17 schools was included in

4 | DONALD ET AL

all models as a set of dummy variables (Gelman amp Hill2007) This approach is more conservative than a classicmultilevel modeling (partial pooling) approach as it does notforce random effects to be normally distributed and therebyallows for greater heterogeneity in school-level effects (Gel-man amp Hill 2007)

The measures of both self-compassion and self-esteemincluded several negatively worded items To address well-documented problems of response bias resulting from theuse of negatively worded items in self-report measures weestimated the covariance between negative items for eachconstruct in all SEMs in addition to the main factors esti-mated for each construct at each time point (Di Stefano ampMotl 2006)

232 | Missing data

Given that this was a longitudinal study with high school stu-dents who received no financial incentive to participate par-ticipant attrition was a potential problem Participant attritioncan result in data that are not missing completely at randomleading to biased parameter estimates when methods such aspair- or list-wise deletion of missing data are used (Enders2010) To examine the influence of participant attrition wecompared individuals who participated in all 4 years of thestudy (completers) with those who participated in 3 or feweryears (non-completers) testing whether there were differen-ces between these groups on the two primary study variablesWe found no significant differences between completers andnon-completers for either self-esteem or self-compassionacross all 4 years of data suggesting that there were no sys-tematic effects associated with participant attrition (allCohenrsquos ds werelt 020) Nonetheless given the potentialbiases associated with using traditional approaches to han-dling missing data (eg list-wise or pair-wise deletion) weused the full information maximum likelihood (FIML)approach to missing data (Baraldi amp Enders 2010) Modernapproaches to missing data such as FIML are preferable totraditional approaches as they use all the available informa-tion for parameter estimation (Enders 2010)

233 | Fit statistics

Models were considered to fit the data well if parameter esti-mates were consistent with the theory proposed the solutionwas well defined and the fit indices were acceptable (McDo-nald amp Marsh 1990) In addition to the chi-square statisticwe used three other fit indices the TuckerndashLewis index(TLI) the comparative fit index (CFI) and the root meansquare error of approximation (RMSEA) Generally acceptedminimum thresholds for the former two indices are 90whereas 08 is generally considered an acceptable maximum

threshold for RMSEA (Chen 2007 Cheung amp Rensvold2002) These three fit indices have the advantage of beingrelatively insensitive to sample size making them appropri-ate for the present study with a relatively large sample(Cheung amp Rensvold 2002 Hu amp Bentler 1999)

234 | Modeling approach

In structural equation modeling two kinds of models arespecified a confirmatory factor analytic (CFA) model to testhow well the underlying measurement model fits the dataand to test the degree to which the proposed model is invari-ant (eg across time or groups) and a series of structuralmodels to test the substantive hypothesized relations amongconstructs (Bollen 1989) In testing the proposed measure-ment model in the present study we ran tests of measure-ment invariance across time as longitudinal relations amongconstructs was the focus of the research Structural modelsmay also include tests of structural invariance where pathsbetween parameters are specified to be equal across time orgroups This enables one to test for developmental equilib-rium in a hypothesized relationship across time (Bollen1989 Little 2013)

In the present study we tested a total of five latent mod-els two measurement models (CFA) and three structuralmodels (SEM) As a first step a configural measurementmodel was estimated in which all model parameters wereallowed to vary across time (CFA 1) If the hypothesis ofconfigural invariance is not rejected stronger forms of mea-surement invariance may be used (Bollen 1989) Followingthe configural model we estimated a second model (CFA 2)in which we tested for measurement invariance across timeTo achieve this the loadings of each factor onto its respec-tive items were constrained to be equal across time Supportfor this model indicates that the construct being measuredhas the same meaning at each time point and is an assump-tion of covariance-based models such as the ACL modelsestimated here (Ciarrochi et al 2016) In this situation con-straining factor loadings to be equal across time points is theonly precondition for establishing time invariance (Millsap2011)

Following tests of measurement invariance a series ofthree structural models was estimated to test whether therelationships between variables across the four waves of datawere stable across timemdashthat is whether they were charac-terized by a developmental equilibrium Under this approachevidence of invariance comes from comparing a well-fittingbaseline model with a series of alternate nested models (Bol-len 1989) To make such comparisons we used the criteriaby Cheung and Rensvold (2002) who suggest that invari-ance exists between nested models if CFI islt 01 (we usedthe same criteria for the TLI) and the criteria described by

DONALD ET AL | 5

Chen (2007) who suggests invariance between nested mod-els exists if RMSEA 15

The first of the structural models was a fully forwardmodel in which estimates for all paths (both autoregressiveand cross-lagged) were estimated including all lags greaterthan 1 across multiple time points (SEM 1) Next all lagsgreater than 1 were removed (SEM 2) Finally we con-strained estimates across single-year lags to be equal therebytesting for developmental equilibrium (SEM 3) Diagram-matic illustrations of SEM 2 and SEM 3 across 3 rather than4 years (for the sake of simplicity) are shown in Figure 1

3 | RESULTS

31 | Preliminary analyses

Latent means and standard deviations for both self-compassion and self-esteem are shown in Table 1 As can beseen these were relatively consistent across time

As shown in Table 2 bivariate correlations among studyvariables across time points were in the 05ndash07 range forself-esteem and 03ndash06 range for self-compassion Intra-variable correlations were strongest at proximal time pointsand weakest at distal time points for both variables

We next tested the links between our two main variablesof interest and the three demographic variables discussedabove participant gender participant age and parentsrsquo mari-tal status Neither age nor parentsrsquo marital status significantly

predicted either self-compassion or self-esteem at any of thefour time points However gender consistently predictedboth self-esteem and self-compassion with females reportinglower self-esteem and self-compassion than males in each ofthe 4 years of the study For self-esteem standardized esti-mates of the difference between females and males rangedfrom 49 to 58 whereas for self-compassion standardizedestimates ranged from 31 to 43 with all pslt 0001

32 | Primary analyses

We next report the results from the ACL models As dis-cussed above we tested a series of increasingly restrictivemodels In all models we controlled for gender (but not par-ticipant age or parental marital status) given the evidencereported above for gender differences for both self-esteemand self-compassion across time Table 2 indicates that thefit indices for CFA 1 (the configural measurement model)and CFA 2 (the measurement model with factor loadingsconstrained to be equal across time) were acceptable withthe latter indicating measurement invariance for this modelof the data The fit indices for SEMs 1ndash3 are also displayedin Table 3 and the changes in these indices (ie SEM 1through to SEM 3) are within the thresholds outlined in theprevious section

As shown in Figure 2 self-esteem consistently predictedgreater self-compassion across the 4 years of the studywhereas the reciprocal relationship of self-compassion pre-dicting changes in self-esteem across time was not foundThis finding provides support for the self-compassion as con-sequence model

We next ran SEM 3 the developmental equilibriummodel in which single-year estimates were constrained to beequal Estimates for this model are displayed in Figure 31

We tested whether the cross-lagged estimates were signifi-cantly different from one another and found support for this(z5 837 plt 001) The correlation between disturbances atTime 2 (b5 39) suggests a medium-sized relationshipbetween changes in self-esteem and self-compassion acrosstime

a

c

bb

a

c

dd

(a) (b)

a

b

c

d

g

h

f

e

SE SE SE SE SE SE

SC SC SC SC SC SC

FIGURE 1 A conceptual diagram of structural model SEM 2 (Figure 1a) and SEM 3 (Figure 1b) SE5 self-esteem SC5 self-compassion SEM 2Path coefficients are freely estimated for each path indicated by arrows SEM 3 Autoregressive and cross-lagged path coefficients are constrained to beequal across time intervals

TABLE 1 Descriptive statistics for self-esteem and self-compassion at each time point

Self-compassion Self-esteem

n Mean SD n Mean SD

Year 9 1769 310 094 1888 070 033

Year 10 1772 303 090 1836 069 033

Year 11 1546 296 090 1576 066 034

Year 12 1506 305 089 1496 068 034

Note SD5 standard deviation Means and standard deviations are latent

6 | DONALD ET AL

To explore the extent to which specific components ofself-compassion relate longitudinally to self-esteem we con-ducted supplementary analyses Based on Neffrsquos (2003) con-ceptualization of self-compassion as comprising threepositive (mindfulness self-kindness and common humanity)and three negative (overidentification self-judgment andisolation) components each of these six self-compassioncomponents was separately modeled to explore their longitu-dinal relations with self-esteem The results of these supple-mentary analyses are consistent with our main findings withself-esteem consistently predicting self-compassion domainsacross the four waves of the study but inconsistentlymdashwithvery small effect sizesmdashfrom self-compassion to self-esteem(see these results in the Supplemental Material) It should beflagged however that the Self-Compassion ScalendashShortForm (Raes et al 2011) used in the present study has beenshown to perform best as a single-factor rather than a multi-dimensional instrument and this is consistent with theapproach we have taken in conducting the primary analysesabove

4 | DISCUSSION

This is the first study we are aware of to examine the longitu-dinal relations between self-esteem and self-compassion Wefound support for the self-compassion as consequence modelacross 4 years of data among a large adolescent sample Ourdevelopmental equilibrium model (SEM 3) suggests a stableprocess across time with self-esteem consistently explainingyear-on-year changes in self-compassion Notably a recipro-cal effect of self-compassion predicting changes in self-esteem was not found

These findings are consistent with our expectation thatpositive self-evaluations give rise to expressions of self-compassion when an individual is faced with difficulty Theself-compassion literature suggests that when individuals feelvulnerable and possess negative self-evaluations they areless likely to be self-compassionate as efforts to engage inself-compassion are perceived as threatening and undeserved(Gilbert et al 2011 Kelly et al 2014 Pauley amp McPher-son 2010) Our findings are consistent with these

TABLE 2 Bivariate correlations between study variables

1 2 3 4 5 6 7

1 Self-esteem (Year 9) mdash

2 Self-esteem (Year 10) 0622 mdash

3 Self-esteem (Year 11) 0608 0693 mdash

4 Self-esteem (Year 12) 0525 0560 0674 mdash

5 Self-compassion (Year 9) 0542 0371 0424 0358 mdash

6 Self-compassion (Year 10) 0432 0616 0515 0453 0412 mdash

7 Self-compassion (Year 11) 0447 0520 0633 0498 0496 0593 mdash

8 Self-compassion (Year 12) 0368 0387 0468 0595 0354 0472 0566

Note plt 001

TABLE 3 Fit indices for progressively more restrictive models

Model Description v2 df RMSEA CFI TLI

CFA 1 Configural CFA 1137896 5684 019 920 912

CFA 2 CFA with loading invariance 1156491 5748 019 918 911

SEM 1 SEM with all longitudinal paths estimated 1132116 5620 019 920 911

SEM 2 SEM with only single-year lags estimated 1145675 5634 019 918 910

SEM 3 SEM with developmental equilibrium 1148025 5640 019 918 909

Note RMSEA5 root mean square error of approximation CFI5 comparative fit index TLI5TuckerndashLewis index CFA5 confirmatory factor analytic modelSEM5 structural equation model

DONALD ET AL | 7

explanations showing that evaluations of worthiness of theself are important antecedents of the capacity to extend com-passion to oneself

Research on the antecedents of other-oriented compassionmay also help to explain our findings While responding incompassionate ways to others appraisals of deservingnesshave been identified as a core elementmdashalong with noticinganotherrsquos suffering feeling empathy for him or her and actingcompassionately (Atkins amp Parker 2012) Further there isevidence that when people appraise others as being responsi-ble for their own suffering they are less likely to extend com-passion to them (Rudolph Roesch Greitemeyer amp Weiner2004) Similarly when a person appraises another to be colduncooperative or nasty he or she is less likely to extend com-passion to him or her (Batson Eklund Chermok Hoyt ampOrtiz 2007 Fiske Cuddy Glick amp Xu 2002) Such proc-esses may occur in relation to the self as well When individu-als appraise themselves as responsible for their own sufferingor more broadly in a negative way (eg as being unworthy)they may be less likely to develop compassion toward them-selves Conversely having positive self-oriented evaluationsof deservingness and worth appears to be an important ante-cedent of the development of the capacity to extend compas-sion toward oneself when faced with difficulty

One other explanation for our findings is that self-compassion may function as a coping mechanism that indi-viduals with high self-esteem use to deal effectively with

rejection and failures Adolescents with high levels of self-esteem tend to engage in more adaptive and less avoidantcoping responses following setbacks (Mullis amp Chapman2000) Self-compassion involves engaging with and accept-ing onersquos own suffering rather than avoiding it so in thatsense it can be viewed as an approach form of coping (Car-ver amp Connor-Smith 2010 Neff et al 2005)

Further our findings have potentially important implica-tions for the understanding of how self-compassion is devel-oped among adolescents and perhaps more broadly Theoriesof self-compassion emphasize the behavior of noticing suffer-ing extending kindness toward oneself and recognizing suf-fering as common to humanity (Neff 2003 Neff et al2005) Our research suggests that understanding the kinds ofself-evaluations that young people make may be critical tohelping develop self-compassion In recent years researchexamining the efficacy of compassion-based interventionshas been growing (see Galante Galante Bekkers amp Gal-lacher 2014 for a review) including among young people(Bluth Gaylord Campo Mullarkey amp Hobbs 2016 Reddyet al 2013) Our findings suggest that for enhancing the effi-cacy of such interventions it may be important to understandand influence self-oriented evaluations in particular by ensur-ing that young people feel worthy of respect and kindness

Lastly the lack of support for the self-compassion asantecedent model is noteworthy We anticipated that becauseself-compassion involves cultivating positive self-orientedemotions and forgiving oneself for failures and mistakesthis would lead to positive self-oriented evaluations (ieenhanced self-esteem) However self-compassion research-ers have distinguished between self-compassion and self-esteem on the basis that the former does not involve cultivat-ing positive self-evaluations whereas the latter does (Neff2003 Neff amp Vonk 2009) Recent evidence has shown thatrather than reinforcing onersquos sense of self and identity self-compassion is associated with efforts to transcend it (Lindsayamp Creswell 2014 Neff amp Beretvas 2012 Neff amp Pommier2013 Welp amp Brown 2016) Further support for this comesfrom studies showing that other-oriented compassion

Self-esteem

Self-esteem

Self-esteem

Self-esteem

Self- compassion

Self-compassion

Self-compassion

Self- compassion

05

16

71

72

Grade 9 Grade 10 Grade 11 Grade 12

76

17

60

03

75

12

51

03 32 43 43 17

FIGURE 2 Path estimates from SEM 2with only single-year lags estimatedNote p lt 001

Self-esteem

Self-esteem

Self-compassion

Self- compassion

14 05

73

60

39 33

FIGURE 3 Path estimates from SEM 3 the developmental equilib-riummodel with mean standardized path estimates shownNote p lt 001

8 | DONALD ET AL

interventions lead to more pro-social behaviur and implicateregions in the brain associated with self-transcendence(McCall Steinbeis Ricard amp Singer 2014 Weng et al2013) This evidence suggests that cultivating self-compassion does not lead to self-evaluations of worthinessor deservingness perhaps because it leads to a more tran-scendent sense of self

5 | LIMITATIONS AND FUTUREDIRECTIONS

This study has several limitations First although this studyprovides valuable information regarding the temporal order-ing of the relations between self-compassion and self-esteemand although we tested for links between the study variablesand demographic covariates it was not possible to com-pletely rule out the possibility that additional exogenous vari-ables accounted for these results (Morgan amp Winship 2007)Experimental research is needed to test these relationships ina way that more robustly controls for potential confoundsFor example future self-compassion research could examinedifferential effects of compassion interventions where self-evaluations are explored and made explicit prior to self-compassion training and self-compassion interventionswhere this does not occur Another limitation is that we werenot able to test mediators of the link between self-esteem andself-compassion in the self-compassion as consequencemodel As discussed we expect that appraisals of deserving-ness may be an important mediator of this link Futureresearch could explicitly test this possibility by measuringfor example self-criticism and self-reassurance (GilbertClarke Hempel Miles amp Irons 2004) as potential mediatingvariables

Future research could also examine other possible con-textual variables that explain the relations between self-esteem and self-compassion over time For example socialsupport has been developmentally linked with self-esteem(Marshall et al 2013) and it may be that among individualswith relatively high social support the role of self-esteem indeveloping self-compassion is less important than for moresocially isolated adolescents Demographic factors such aseducation level or socioeconomic status may also moderaterelations between self-esteem and self-compassion across thedevelopmental spectrum Lastly it may be that differentforms of self-esteem influence the development of self-compassion in differential ways For example having vari-able or contingent self-esteem which have both been associ-ated with defensive and maladaptive behavior (KernisLakey amp Heppner 2008) may in fact undermine the devel-opment of self-compassion whereas relatively stable or non-contingent self-esteem enhances it Future research couldexplore these and other potential moderators

Additionally the present study used a convenience sam-ple of 17 Catholic schools across two Australian statesAlthough there is evidence that these schools were broadlyrepresentative of Australiarsquos socioeconomic standingincluded a mix of urban regional and rural schools and hadan almost even gender balance future research could test thelinks between self-esteem and self-compassion amongnationally representative samples Australia is an affluentsociety and future research could explore the extent to whichour findings can be replicated in lower socioeconomic sam-ples where the effects of self-esteem on self-compassion maybe different

Lastly the present study focused on adolescents (Years9ndash12) and our conclusions are therefore constrained to thisage group Future research is needed to understand whetherthe longitudinal relations between self-esteem and self-compassion identified in the present article extend into adult-hood and if so whether they strengthen weaken or aremaintained at a similar level There is evidence that self-evaluations do indeed influence the capacity for self-compassion among adults suggesting that the process identi-fied in the present article is likely to continue into adulthood(Gilbert et al 2011 Kelly Carter Zuroff amp Borairi 2012Kelly et al 2014) However further research is needed todirectly test the longitudinal relations between self-esteemand self-compassion across the entire human developmentalspectrum

6 | CONCLUSION

In contrast to the initial validation studies of self-compassion which focused on the discriminant validity ofself-compassion with respect to self-esteem our study bringsself-esteem back into the spotlight as an important precursorto the development of self-compassion among young peopleAcross 4 years we found consistent evidence for self-esteemas an antecedent of the development of self-compassion butnot vice versa Our findings have an important implicationfor the design of self-compassion interventions To help ado-lescents become more self-compassionate enhancing theirsense of worthiness and deservingness may be as importantas directly teaching them to be kind to themselves

ACKNOWLEDGMENT

The author(s) disclosed receipt of the following financialsupport for the research authorship andor publication ofthis article Preparation of this manuscript was supportedby Grant DP140103874 from the Australian ResearchCouncil

DONALD ET AL | 9

CONFLICT OF INTERESTS

The author(s) declared no potential conflicts of interestwith respect to the research authorship andor publicationof this article

NOTE1 For completeness we ran these same analyses but without gender as acovariate and obtained identical parameter estimates

ORCID

James N Donald httporcidorg0000-0002-5940-2536Joseph Ciarrochi httporcidorg0000-0003-0471-8100Philip D Parker httporcidorg0000-0002-4604-8566Baljinder K Sahdra httporcidorg0000-0001-5064-7783

REFERENCES

Amato P R amp Keith B (1991) Parental Divorce and the Weil-Being of Children A Meta-Analysis Psychological Bulletin 11026ndash46

Atkins P W B amp Parker S K (2012) Understanding individualcompassion in organizations The role of appraisals and psycho-logical flexibility Academy of Management Review AMR-10 1ndash53 httpsdoi 105465amr20100490

Australian Bureau of Statistics (2012) Yearbook Australia Primaryand secondary education schooling structures (Document13010) Canberra Australia Government Printer

Baraldi A N amp Enders C K (2010) An introduction to modernmissing data analyses Journal of School Psychology 48 5ndash37httpsdoiorg101016jjsp200910001

Barry C T Loflin D C amp Doucette H (2015) Adolescent self-compassion Associations with narcissism self-esteem aggres-sion and internalizing symptoms in at-risk males Personality andIndividual Differences 77 118ndash123 httpsdoiorg101016jpaid201412036

Baumeister R F Bushman B J Campbell W K Baumeister RF Bushman B J amp Campbell W K (2000) Self-esteem nar-cissism and aggression Does violence result from low self-esteemor from threatened egotism Current Directions in PsychologicalScience 9 26ndash29 httpsdxdoiorg1011111467-872100053

Baumeister R F Campbell J D Krueger J I amp Vohs K D(2003) Does high self-esteem cause better performance interper-sonal success happiness or healthier lifestyles PsychologicalScience in the Public Interest 4 1ndash44 httpsdxdoiorg1011111529-100601431

Baumeister R F Heatherton T F amp Tice D M (1993) Whenego threats lead to self-regulation failure Negative consequencesof high self-esteem Journal of Personality and Social Psychol-ogy 64 141ndash156 httpsdxdoiorg1010370022-3514641141

Baumeister R F Smart L amp Boden J M (1996) Relation ofthreatened egotism to violence and aggression The dark side ofhigh self-esteem Psychological Review 103 5ndash33 httpsdxdoiorg1010370033-295X10315

Batson C D Eklund J H Chermok V L Hoyt J L amp OrtizB G (2007) An additional antecedent of empathic concern Val-uing the welfare of the person in need Journal of Personalityand Social Psychology 93 65ndash74 httpsdxdoiorg1010370022-351493165

Bluth K Gaylord S A Campo R A Mullarkey M C ampHobbs L (2016) Making friends with yourself A mixed meth-ods pilot study of a mindful self-compassion program for adoles-cents Mindfulness 7 479ndash492 httpsdoiorg101007s12671-015-0476-6

Bollen K A (1989) Structural equations with latent variables NewYork NY Wiley httpsdxdoiorg1010029781118619179

Carver C S amp Connor-Smith J (2010) Personality and copingAnnual Review of Psychology 61 679ndash704 httpsdoiorg101146annurevpsych093008100352

Chen F F (2007) Sensitivity of goodness of fit indexes to lack ofmeasurement invariance Structural Equation Modeling A Multi-disciplinary Journal 14 464ndash504 httpsdoiorg10108010705510701301834

Cheng H amp Furnham A (2004) Perceived parental rearing styleself-esteem and self-criticism as predictors of happiness Journalof Happiness Studies 5 1ndash21 httpsdoiorg101023BJOHS00000217043526705

Cheung G W amp Rensvold R B (2002) Evaluating goodness-of-fit indexes for testing measurement invariance Structural Equa-tion Modeling A Multidisciplinary Journal 9 233ndash255 httpsdoiorg101207S15328007SEM0902

Ciarrochi J Parker P Kashdan T B Heaven P C L amp BarkusE (2015) Hope and emotional well-being A six-year study todistinguish antecedents correlates and consequences Journal ofPositive Psychology 10 1ndash13 httpsdoiorg1010801743976020151015154

Ciarrochi J Parker P Sahdra B Marshall S Jackson C ampGloster A T (2016) The development of compulsive Internetuse and mental health A four-year study of adolescence Devel-opmental Psychology 52 272ndash283 httpsdxdoiorg101037dev0000070

Crocker J amp Park L E (2004) The costly pursuit of self-esteemPsychological Bulletin 130 392ndash414 httpsdoiorg1010370033-29091303392

Di Stefano C amp Motl R W (2006) Further investigating methodeffects associated with negatively worded items on self-report sur-veys Structural Equation Modeling A Multidisciplinary Journal13 440ndash464 httpsdoiorg101207s15328007sem1303_6

Donnellan M B Trzesniewski K H Robins R W Moffitt TE amp Caspi A (2005) Low self-esteem is related to aggressionantisocial behavior and delinquency Psychological Science 16328ndash335 httpsdoiorg101111j0956-7976200501535x

Enders C (2010) Applied missing data analysis New York NYGuilford Press httpsdxdoiorg1012691ajssm-4-4-1

Fiske S T Cuddy A J C Glick P amp Xu J (2002) A model of(often mixed) stereotype content Competence and warmth respec-tively follow from perceived status and competition Journal ofPersonality and Social Psychology 82 878ndash902 httpsdxdoiorg1010370022-3514826878

10 | DONALD ET AL

Chen (2007) who suggests invariance between nested mod-els exists if RMSEA 15

The first of the structural models was a fully forwardmodel in which estimates for all paths (both autoregressiveand cross-lagged) were estimated including all lags greaterthan 1 across multiple time points (SEM 1) Next all lagsgreater than 1 were removed (SEM 2) Finally we con-strained estimates across single-year lags to be equal therebytesting for developmental equilibrium (SEM 3) Diagram-matic illustrations of SEM 2 and SEM 3 across 3 rather than4 years (for the sake of simplicity) are shown in Figure 1

3 | RESULTS

31 | Preliminary analyses

Latent means and standard deviations for both self-compassion and self-esteem are shown in Table 1 As can beseen these were relatively consistent across time

As shown in Table 2 bivariate correlations among studyvariables across time points were in the 05ndash07 range forself-esteem and 03ndash06 range for self-compassion Intra-variable correlations were strongest at proximal time pointsand weakest at distal time points for both variables

We next tested the links between our two main variablesof interest and the three demographic variables discussedabove participant gender participant age and parentsrsquo mari-tal status Neither age nor parentsrsquo marital status significantly

predicted either self-compassion or self-esteem at any of thefour time points However gender consistently predictedboth self-esteem and self-compassion with females reportinglower self-esteem and self-compassion than males in each ofthe 4 years of the study For self-esteem standardized esti-mates of the difference between females and males rangedfrom 49 to 58 whereas for self-compassion standardizedestimates ranged from 31 to 43 with all pslt 0001

32 | Primary analyses

We next report the results from the ACL models As dis-cussed above we tested a series of increasingly restrictivemodels In all models we controlled for gender (but not par-ticipant age or parental marital status) given the evidencereported above for gender differences for both self-esteemand self-compassion across time Table 2 indicates that thefit indices for CFA 1 (the configural measurement model)and CFA 2 (the measurement model with factor loadingsconstrained to be equal across time) were acceptable withthe latter indicating measurement invariance for this modelof the data The fit indices for SEMs 1ndash3 are also displayedin Table 3 and the changes in these indices (ie SEM 1through to SEM 3) are within the thresholds outlined in theprevious section

As shown in Figure 2 self-esteem consistently predictedgreater self-compassion across the 4 years of the studywhereas the reciprocal relationship of self-compassion pre-dicting changes in self-esteem across time was not foundThis finding provides support for the self-compassion as con-sequence model

We next ran SEM 3 the developmental equilibriummodel in which single-year estimates were constrained to beequal Estimates for this model are displayed in Figure 31

We tested whether the cross-lagged estimates were signifi-cantly different from one another and found support for this(z5 837 plt 001) The correlation between disturbances atTime 2 (b5 39) suggests a medium-sized relationshipbetween changes in self-esteem and self-compassion acrosstime

a

c

bb

a

c

dd

(a) (b)

a

b

c

d

g

h

f

e

SE SE SE SE SE SE

SC SC SC SC SC SC

FIGURE 1 A conceptual diagram of structural model SEM 2 (Figure 1a) and SEM 3 (Figure 1b) SE5 self-esteem SC5 self-compassion SEM 2Path coefficients are freely estimated for each path indicated by arrows SEM 3 Autoregressive and cross-lagged path coefficients are constrained to beequal across time intervals

TABLE 1 Descriptive statistics for self-esteem and self-compassion at each time point

Self-compassion Self-esteem

n Mean SD n Mean SD

Year 9 1769 310 094 1888 070 033

Year 10 1772 303 090 1836 069 033

Year 11 1546 296 090 1576 066 034

Year 12 1506 305 089 1496 068 034

Note SD5 standard deviation Means and standard deviations are latent

6 | DONALD ET AL

To explore the extent to which specific components ofself-compassion relate longitudinally to self-esteem we con-ducted supplementary analyses Based on Neffrsquos (2003) con-ceptualization of self-compassion as comprising threepositive (mindfulness self-kindness and common humanity)and three negative (overidentification self-judgment andisolation) components each of these six self-compassioncomponents was separately modeled to explore their longitu-dinal relations with self-esteem The results of these supple-mentary analyses are consistent with our main findings withself-esteem consistently predicting self-compassion domainsacross the four waves of the study but inconsistentlymdashwithvery small effect sizesmdashfrom self-compassion to self-esteem(see these results in the Supplemental Material) It should beflagged however that the Self-Compassion ScalendashShortForm (Raes et al 2011) used in the present study has beenshown to perform best as a single-factor rather than a multi-dimensional instrument and this is consistent with theapproach we have taken in conducting the primary analysesabove

4 | DISCUSSION

This is the first study we are aware of to examine the longitu-dinal relations between self-esteem and self-compassion Wefound support for the self-compassion as consequence modelacross 4 years of data among a large adolescent sample Ourdevelopmental equilibrium model (SEM 3) suggests a stableprocess across time with self-esteem consistently explainingyear-on-year changes in self-compassion Notably a recipro-cal effect of self-compassion predicting changes in self-esteem was not found

These findings are consistent with our expectation thatpositive self-evaluations give rise to expressions of self-compassion when an individual is faced with difficulty Theself-compassion literature suggests that when individuals feelvulnerable and possess negative self-evaluations they areless likely to be self-compassionate as efforts to engage inself-compassion are perceived as threatening and undeserved(Gilbert et al 2011 Kelly et al 2014 Pauley amp McPher-son 2010) Our findings are consistent with these

TABLE 2 Bivariate correlations between study variables

1 2 3 4 5 6 7

1 Self-esteem (Year 9) mdash

2 Self-esteem (Year 10) 0622 mdash

3 Self-esteem (Year 11) 0608 0693 mdash

4 Self-esteem (Year 12) 0525 0560 0674 mdash

5 Self-compassion (Year 9) 0542 0371 0424 0358 mdash

6 Self-compassion (Year 10) 0432 0616 0515 0453 0412 mdash

7 Self-compassion (Year 11) 0447 0520 0633 0498 0496 0593 mdash

8 Self-compassion (Year 12) 0368 0387 0468 0595 0354 0472 0566

Note plt 001

TABLE 3 Fit indices for progressively more restrictive models

Model Description v2 df RMSEA CFI TLI

CFA 1 Configural CFA 1137896 5684 019 920 912

CFA 2 CFA with loading invariance 1156491 5748 019 918 911

SEM 1 SEM with all longitudinal paths estimated 1132116 5620 019 920 911

SEM 2 SEM with only single-year lags estimated 1145675 5634 019 918 910

SEM 3 SEM with developmental equilibrium 1148025 5640 019 918 909

Note RMSEA5 root mean square error of approximation CFI5 comparative fit index TLI5TuckerndashLewis index CFA5 confirmatory factor analytic modelSEM5 structural equation model

DONALD ET AL | 7

explanations showing that evaluations of worthiness of theself are important antecedents of the capacity to extend com-passion to oneself

Research on the antecedents of other-oriented compassionmay also help to explain our findings While responding incompassionate ways to others appraisals of deservingnesshave been identified as a core elementmdashalong with noticinganotherrsquos suffering feeling empathy for him or her and actingcompassionately (Atkins amp Parker 2012) Further there isevidence that when people appraise others as being responsi-ble for their own suffering they are less likely to extend com-passion to them (Rudolph Roesch Greitemeyer amp Weiner2004) Similarly when a person appraises another to be colduncooperative or nasty he or she is less likely to extend com-passion to him or her (Batson Eklund Chermok Hoyt ampOrtiz 2007 Fiske Cuddy Glick amp Xu 2002) Such proc-esses may occur in relation to the self as well When individu-als appraise themselves as responsible for their own sufferingor more broadly in a negative way (eg as being unworthy)they may be less likely to develop compassion toward them-selves Conversely having positive self-oriented evaluationsof deservingness and worth appears to be an important ante-cedent of the development of the capacity to extend compas-sion toward oneself when faced with difficulty

One other explanation for our findings is that self-compassion may function as a coping mechanism that indi-viduals with high self-esteem use to deal effectively with

rejection and failures Adolescents with high levels of self-esteem tend to engage in more adaptive and less avoidantcoping responses following setbacks (Mullis amp Chapman2000) Self-compassion involves engaging with and accept-ing onersquos own suffering rather than avoiding it so in thatsense it can be viewed as an approach form of coping (Car-ver amp Connor-Smith 2010 Neff et al 2005)

Further our findings have potentially important implica-tions for the understanding of how self-compassion is devel-oped among adolescents and perhaps more broadly Theoriesof self-compassion emphasize the behavior of noticing suffer-ing extending kindness toward oneself and recognizing suf-fering as common to humanity (Neff 2003 Neff et al2005) Our research suggests that understanding the kinds ofself-evaluations that young people make may be critical tohelping develop self-compassion In recent years researchexamining the efficacy of compassion-based interventionshas been growing (see Galante Galante Bekkers amp Gal-lacher 2014 for a review) including among young people(Bluth Gaylord Campo Mullarkey amp Hobbs 2016 Reddyet al 2013) Our findings suggest that for enhancing the effi-cacy of such interventions it may be important to understandand influence self-oriented evaluations in particular by ensur-ing that young people feel worthy of respect and kindness

Lastly the lack of support for the self-compassion asantecedent model is noteworthy We anticipated that becauseself-compassion involves cultivating positive self-orientedemotions and forgiving oneself for failures and mistakesthis would lead to positive self-oriented evaluations (ieenhanced self-esteem) However self-compassion research-ers have distinguished between self-compassion and self-esteem on the basis that the former does not involve cultivat-ing positive self-evaluations whereas the latter does (Neff2003 Neff amp Vonk 2009) Recent evidence has shown thatrather than reinforcing onersquos sense of self and identity self-compassion is associated with efforts to transcend it (Lindsayamp Creswell 2014 Neff amp Beretvas 2012 Neff amp Pommier2013 Welp amp Brown 2016) Further support for this comesfrom studies showing that other-oriented compassion

Self-esteem

Self-esteem

Self-esteem

Self-esteem

Self- compassion

Self-compassion

Self-compassion

Self- compassion

05

16

71

72

Grade 9 Grade 10 Grade 11 Grade 12

76

17

60

03

75

12

51

03 32 43 43 17

FIGURE 2 Path estimates from SEM 2with only single-year lags estimatedNote p lt 001

Self-esteem

Self-esteem

Self-compassion

Self- compassion

14 05

73

60

39 33

FIGURE 3 Path estimates from SEM 3 the developmental equilib-riummodel with mean standardized path estimates shownNote p lt 001

8 | DONALD ET AL

interventions lead to more pro-social behaviur and implicateregions in the brain associated with self-transcendence(McCall Steinbeis Ricard amp Singer 2014 Weng et al2013) This evidence suggests that cultivating self-compassion does not lead to self-evaluations of worthinessor deservingness perhaps because it leads to a more tran-scendent sense of self

5 | LIMITATIONS AND FUTUREDIRECTIONS

This study has several limitations First although this studyprovides valuable information regarding the temporal order-ing of the relations between self-compassion and self-esteemand although we tested for links between the study variablesand demographic covariates it was not possible to com-pletely rule out the possibility that additional exogenous vari-ables accounted for these results (Morgan amp Winship 2007)Experimental research is needed to test these relationships ina way that more robustly controls for potential confoundsFor example future self-compassion research could examinedifferential effects of compassion interventions where self-evaluations are explored and made explicit prior to self-compassion training and self-compassion interventionswhere this does not occur Another limitation is that we werenot able to test mediators of the link between self-esteem andself-compassion in the self-compassion as consequencemodel As discussed we expect that appraisals of deserving-ness may be an important mediator of this link Futureresearch could explicitly test this possibility by measuringfor example self-criticism and self-reassurance (GilbertClarke Hempel Miles amp Irons 2004) as potential mediatingvariables

Future research could also examine other possible con-textual variables that explain the relations between self-esteem and self-compassion over time For example socialsupport has been developmentally linked with self-esteem(Marshall et al 2013) and it may be that among individualswith relatively high social support the role of self-esteem indeveloping self-compassion is less important than for moresocially isolated adolescents Demographic factors such aseducation level or socioeconomic status may also moderaterelations between self-esteem and self-compassion across thedevelopmental spectrum Lastly it may be that differentforms of self-esteem influence the development of self-compassion in differential ways For example having vari-able or contingent self-esteem which have both been associ-ated with defensive and maladaptive behavior (KernisLakey amp Heppner 2008) may in fact undermine the devel-opment of self-compassion whereas relatively stable or non-contingent self-esteem enhances it Future research couldexplore these and other potential moderators

Additionally the present study used a convenience sam-ple of 17 Catholic schools across two Australian statesAlthough there is evidence that these schools were broadlyrepresentative of Australiarsquos socioeconomic standingincluded a mix of urban regional and rural schools and hadan almost even gender balance future research could test thelinks between self-esteem and self-compassion amongnationally representative samples Australia is an affluentsociety and future research could explore the extent to whichour findings can be replicated in lower socioeconomic sam-ples where the effects of self-esteem on self-compassion maybe different

Lastly the present study focused on adolescents (Years9ndash12) and our conclusions are therefore constrained to thisage group Future research is needed to understand whetherthe longitudinal relations between self-esteem and self-compassion identified in the present article extend into adult-hood and if so whether they strengthen weaken or aremaintained at a similar level There is evidence that self-evaluations do indeed influence the capacity for self-compassion among adults suggesting that the process identi-fied in the present article is likely to continue into adulthood(Gilbert et al 2011 Kelly Carter Zuroff amp Borairi 2012Kelly et al 2014) However further research is needed todirectly test the longitudinal relations between self-esteemand self-compassion across the entire human developmentalspectrum

6 | CONCLUSION

In contrast to the initial validation studies of self-compassion which focused on the discriminant validity ofself-compassion with respect to self-esteem our study bringsself-esteem back into the spotlight as an important precursorto the development of self-compassion among young peopleAcross 4 years we found consistent evidence for self-esteemas an antecedent of the development of self-compassion butnot vice versa Our findings have an important implicationfor the design of self-compassion interventions To help ado-lescents become more self-compassionate enhancing theirsense of worthiness and deservingness may be as importantas directly teaching them to be kind to themselves

ACKNOWLEDGMENT

The author(s) disclosed receipt of the following financialsupport for the research authorship andor publication ofthis article Preparation of this manuscript was supportedby Grant DP140103874 from the Australian ResearchCouncil

DONALD ET AL | 9

CONFLICT OF INTERESTS

The author(s) declared no potential conflicts of interestwith respect to the research authorship andor publicationof this article

NOTE1 For completeness we ran these same analyses but without gender as acovariate and obtained identical parameter estimates

ORCID

James N Donald httporcidorg0000-0002-5940-2536Joseph Ciarrochi httporcidorg0000-0003-0471-8100Philip D Parker httporcidorg0000-0002-4604-8566Baljinder K Sahdra httporcidorg0000-0001-5064-7783

REFERENCES

Amato P R amp Keith B (1991) Parental Divorce and the Weil-Being of Children A Meta-Analysis Psychological Bulletin 11026ndash46

Atkins P W B amp Parker S K (2012) Understanding individualcompassion in organizations The role of appraisals and psycho-logical flexibility Academy of Management Review AMR-10 1ndash53 httpsdoi 105465amr20100490

Australian Bureau of Statistics (2012) Yearbook Australia Primaryand secondary education schooling structures (Document13010) Canberra Australia Government Printer

Baraldi A N amp Enders C K (2010) An introduction to modernmissing data analyses Journal of School Psychology 48 5ndash37httpsdoiorg101016jjsp200910001

Barry C T Loflin D C amp Doucette H (2015) Adolescent self-compassion Associations with narcissism self-esteem aggres-sion and internalizing symptoms in at-risk males Personality andIndividual Differences 77 118ndash123 httpsdoiorg101016jpaid201412036

Baumeister R F Bushman B J Campbell W K Baumeister RF Bushman B J amp Campbell W K (2000) Self-esteem nar-cissism and aggression Does violence result from low self-esteemor from threatened egotism Current Directions in PsychologicalScience 9 26ndash29 httpsdxdoiorg1011111467-872100053

Baumeister R F Campbell J D Krueger J I amp Vohs K D(2003) Does high self-esteem cause better performance interper-sonal success happiness or healthier lifestyles PsychologicalScience in the Public Interest 4 1ndash44 httpsdxdoiorg1011111529-100601431

Baumeister R F Heatherton T F amp Tice D M (1993) Whenego threats lead to self-regulation failure Negative consequencesof high self-esteem Journal of Personality and Social Psychol-ogy 64 141ndash156 httpsdxdoiorg1010370022-3514641141

Baumeister R F Smart L amp Boden J M (1996) Relation ofthreatened egotism to violence and aggression The dark side ofhigh self-esteem Psychological Review 103 5ndash33 httpsdxdoiorg1010370033-295X10315

Batson C D Eklund J H Chermok V L Hoyt J L amp OrtizB G (2007) An additional antecedent of empathic concern Val-uing the welfare of the person in need Journal of Personalityand Social Psychology 93 65ndash74 httpsdxdoiorg1010370022-351493165

Bluth K Gaylord S A Campo R A Mullarkey M C ampHobbs L (2016) Making friends with yourself A mixed meth-ods pilot study of a mindful self-compassion program for adoles-cents Mindfulness 7 479ndash492 httpsdoiorg101007s12671-015-0476-6

Bollen K A (1989) Structural equations with latent variables NewYork NY Wiley httpsdxdoiorg1010029781118619179

Carver C S amp Connor-Smith J (2010) Personality and copingAnnual Review of Psychology 61 679ndash704 httpsdoiorg101146annurevpsych093008100352

Chen F F (2007) Sensitivity of goodness of fit indexes to lack ofmeasurement invariance Structural Equation Modeling A Multi-disciplinary Journal 14 464ndash504 httpsdoiorg10108010705510701301834

Cheng H amp Furnham A (2004) Perceived parental rearing styleself-esteem and self-criticism as predictors of happiness Journalof Happiness Studies 5 1ndash21 httpsdoiorg101023BJOHS00000217043526705

Cheung G W amp Rensvold R B (2002) Evaluating goodness-of-fit indexes for testing measurement invariance Structural Equa-tion Modeling A Multidisciplinary Journal 9 233ndash255 httpsdoiorg101207S15328007SEM0902

Ciarrochi J Parker P Kashdan T B Heaven P C L amp BarkusE (2015) Hope and emotional well-being A six-year study todistinguish antecedents correlates and consequences Journal ofPositive Psychology 10 1ndash13 httpsdoiorg1010801743976020151015154

Ciarrochi J Parker P Sahdra B Marshall S Jackson C ampGloster A T (2016) The development of compulsive Internetuse and mental health A four-year study of adolescence Devel-opmental Psychology 52 272ndash283 httpsdxdoiorg101037dev0000070

Crocker J amp Park L E (2004) The costly pursuit of self-esteemPsychological Bulletin 130 392ndash414 httpsdoiorg1010370033-29091303392

Di Stefano C amp Motl R W (2006) Further investigating methodeffects associated with negatively worded items on self-report sur-veys Structural Equation Modeling A Multidisciplinary Journal13 440ndash464 httpsdoiorg101207s15328007sem1303_6

Donnellan M B Trzesniewski K H Robins R W Moffitt TE amp Caspi A (2005) Low self-esteem is related to aggressionantisocial behavior and delinquency Psychological Science 16328ndash335 httpsdoiorg101111j0956-7976200501535x

Enders C (2010) Applied missing data analysis New York NYGuilford Press httpsdxdoiorg1012691ajssm-4-4-1

Fiske S T Cuddy A J C Glick P amp Xu J (2002) A model of(often mixed) stereotype content Competence and warmth respec-tively follow from perceived status and competition Journal ofPersonality and Social Psychology 82 878ndash902 httpsdxdoiorg1010370022-3514826878

10 | DONALD ET AL

To explore the extent to which specific components ofself-compassion relate longitudinally to self-esteem we con-ducted supplementary analyses Based on Neffrsquos (2003) con-ceptualization of self-compassion as comprising threepositive (mindfulness self-kindness and common humanity)and three negative (overidentification self-judgment andisolation) components each of these six self-compassioncomponents was separately modeled to explore their longitu-dinal relations with self-esteem The results of these supple-mentary analyses are consistent with our main findings withself-esteem consistently predicting self-compassion domainsacross the four waves of the study but inconsistentlymdashwithvery small effect sizesmdashfrom self-compassion to self-esteem(see these results in the Supplemental Material) It should beflagged however that the Self-Compassion ScalendashShortForm (Raes et al 2011) used in the present study has beenshown to perform best as a single-factor rather than a multi-dimensional instrument and this is consistent with theapproach we have taken in conducting the primary analysesabove

4 | DISCUSSION

This is the first study we are aware of to examine the longitu-dinal relations between self-esteem and self-compassion Wefound support for the self-compassion as consequence modelacross 4 years of data among a large adolescent sample Ourdevelopmental equilibrium model (SEM 3) suggests a stableprocess across time with self-esteem consistently explainingyear-on-year changes in self-compassion Notably a recipro-cal effect of self-compassion predicting changes in self-esteem was not found

These findings are consistent with our expectation thatpositive self-evaluations give rise to expressions of self-compassion when an individual is faced with difficulty Theself-compassion literature suggests that when individuals feelvulnerable and possess negative self-evaluations they areless likely to be self-compassionate as efforts to engage inself-compassion are perceived as threatening and undeserved(Gilbert et al 2011 Kelly et al 2014 Pauley amp McPher-son 2010) Our findings are consistent with these

TABLE 2 Bivariate correlations between study variables

1 2 3 4 5 6 7

1 Self-esteem (Year 9) mdash

2 Self-esteem (Year 10) 0622 mdash

3 Self-esteem (Year 11) 0608 0693 mdash

4 Self-esteem (Year 12) 0525 0560 0674 mdash

5 Self-compassion (Year 9) 0542 0371 0424 0358 mdash

6 Self-compassion (Year 10) 0432 0616 0515 0453 0412 mdash

7 Self-compassion (Year 11) 0447 0520 0633 0498 0496 0593 mdash

8 Self-compassion (Year 12) 0368 0387 0468 0595 0354 0472 0566

Note plt 001

TABLE 3 Fit indices for progressively more restrictive models

Model Description v2 df RMSEA CFI TLI

CFA 1 Configural CFA 1137896 5684 019 920 912

CFA 2 CFA with loading invariance 1156491 5748 019 918 911

SEM 1 SEM with all longitudinal paths estimated 1132116 5620 019 920 911

SEM 2 SEM with only single-year lags estimated 1145675 5634 019 918 910

SEM 3 SEM with developmental equilibrium 1148025 5640 019 918 909

Note RMSEA5 root mean square error of approximation CFI5 comparative fit index TLI5TuckerndashLewis index CFA5 confirmatory factor analytic modelSEM5 structural equation model

DONALD ET AL | 7

explanations showing that evaluations of worthiness of theself are important antecedents of the capacity to extend com-passion to oneself

Research on the antecedents of other-oriented compassionmay also help to explain our findings While responding incompassionate ways to others appraisals of deservingnesshave been identified as a core elementmdashalong with noticinganotherrsquos suffering feeling empathy for him or her and actingcompassionately (Atkins amp Parker 2012) Further there isevidence that when people appraise others as being responsi-ble for their own suffering they are less likely to extend com-passion to them (Rudolph Roesch Greitemeyer amp Weiner2004) Similarly when a person appraises another to be colduncooperative or nasty he or she is less likely to extend com-passion to him or her (Batson Eklund Chermok Hoyt ampOrtiz 2007 Fiske Cuddy Glick amp Xu 2002) Such proc-esses may occur in relation to the self as well When individu-als appraise themselves as responsible for their own sufferingor more broadly in a negative way (eg as being unworthy)they may be less likely to develop compassion toward them-selves Conversely having positive self-oriented evaluationsof deservingness and worth appears to be an important ante-cedent of the development of the capacity to extend compas-sion toward oneself when faced with difficulty

One other explanation for our findings is that self-compassion may function as a coping mechanism that indi-viduals with high self-esteem use to deal effectively with

rejection and failures Adolescents with high levels of self-esteem tend to engage in more adaptive and less avoidantcoping responses following setbacks (Mullis amp Chapman2000) Self-compassion involves engaging with and accept-ing onersquos own suffering rather than avoiding it so in thatsense it can be viewed as an approach form of coping (Car-ver amp Connor-Smith 2010 Neff et al 2005)

Further our findings have potentially important implica-tions for the understanding of how self-compassion is devel-oped among adolescents and perhaps more broadly Theoriesof self-compassion emphasize the behavior of noticing suffer-ing extending kindness toward oneself and recognizing suf-fering as common to humanity (Neff 2003 Neff et al2005) Our research suggests that understanding the kinds ofself-evaluations that young people make may be critical tohelping develop self-compassion In recent years researchexamining the efficacy of compassion-based interventionshas been growing (see Galante Galante Bekkers amp Gal-lacher 2014 for a review) including among young people(Bluth Gaylord Campo Mullarkey amp Hobbs 2016 Reddyet al 2013) Our findings suggest that for enhancing the effi-cacy of such interventions it may be important to understandand influence self-oriented evaluations in particular by ensur-ing that young people feel worthy of respect and kindness

Lastly the lack of support for the self-compassion asantecedent model is noteworthy We anticipated that becauseself-compassion involves cultivating positive self-orientedemotions and forgiving oneself for failures and mistakesthis would lead to positive self-oriented evaluations (ieenhanced self-esteem) However self-compassion research-ers have distinguished between self-compassion and self-esteem on the basis that the former does not involve cultivat-ing positive self-evaluations whereas the latter does (Neff2003 Neff amp Vonk 2009) Recent evidence has shown thatrather than reinforcing onersquos sense of self and identity self-compassion is associated with efforts to transcend it (Lindsayamp Creswell 2014 Neff amp Beretvas 2012 Neff amp Pommier2013 Welp amp Brown 2016) Further support for this comesfrom studies showing that other-oriented compassion

Self-esteem

Self-esteem

Self-esteem

Self-esteem

Self- compassion

Self-compassion

Self-compassion

Self- compassion

05

16

71

72

Grade 9 Grade 10 Grade 11 Grade 12

76

17

60

03

75

12

51

03 32 43 43 17

FIGURE 2 Path estimates from SEM 2with only single-year lags estimatedNote p lt 001

Self-esteem

Self-esteem

Self-compassion

Self- compassion

14 05

73

60

39 33

FIGURE 3 Path estimates from SEM 3 the developmental equilib-riummodel with mean standardized path estimates shownNote p lt 001

8 | DONALD ET AL

interventions lead to more pro-social behaviur and implicateregions in the brain associated with self-transcendence(McCall Steinbeis Ricard amp Singer 2014 Weng et al2013) This evidence suggests that cultivating self-compassion does not lead to self-evaluations of worthinessor deservingness perhaps because it leads to a more tran-scendent sense of self

5 | LIMITATIONS AND FUTUREDIRECTIONS

This study has several limitations First although this studyprovides valuable information regarding the temporal order-ing of the relations between self-compassion and self-esteemand although we tested for links between the study variablesand demographic covariates it was not possible to com-pletely rule out the possibility that additional exogenous vari-ables accounted for these results (Morgan amp Winship 2007)Experimental research is needed to test these relationships ina way that more robustly controls for potential confoundsFor example future self-compassion research could examinedifferential effects of compassion interventions where self-evaluations are explored and made explicit prior to self-compassion training and self-compassion interventionswhere this does not occur Another limitation is that we werenot able to test mediators of the link between self-esteem andself-compassion in the self-compassion as consequencemodel As discussed we expect that appraisals of deserving-ness may be an important mediator of this link Futureresearch could explicitly test this possibility by measuringfor example self-criticism and self-reassurance (GilbertClarke Hempel Miles amp Irons 2004) as potential mediatingvariables

Future research could also examine other possible con-textual variables that explain the relations between self-esteem and self-compassion over time For example socialsupport has been developmentally linked with self-esteem(Marshall et al 2013) and it may be that among individualswith relatively high social support the role of self-esteem indeveloping self-compassion is less important than for moresocially isolated adolescents Demographic factors such aseducation level or socioeconomic status may also moderaterelations between self-esteem and self-compassion across thedevelopmental spectrum Lastly it may be that differentforms of self-esteem influence the development of self-compassion in differential ways For example having vari-able or contingent self-esteem which have both been associ-ated with defensive and maladaptive behavior (KernisLakey amp Heppner 2008) may in fact undermine the devel-opment of self-compassion whereas relatively stable or non-contingent self-esteem enhances it Future research couldexplore these and other potential moderators

Additionally the present study used a convenience sam-ple of 17 Catholic schools across two Australian statesAlthough there is evidence that these schools were broadlyrepresentative of Australiarsquos socioeconomic standingincluded a mix of urban regional and rural schools and hadan almost even gender balance future research could test thelinks between self-esteem and self-compassion amongnationally representative samples Australia is an affluentsociety and future research could explore the extent to whichour findings can be replicated in lower socioeconomic sam-ples where the effects of self-esteem on self-compassion maybe different

Lastly the present study focused on adolescents (Years9ndash12) and our conclusions are therefore constrained to thisage group Future research is needed to understand whetherthe longitudinal relations between self-esteem and self-compassion identified in the present article extend into adult-hood and if so whether they strengthen weaken or aremaintained at a similar level There is evidence that self-evaluations do indeed influence the capacity for self-compassion among adults suggesting that the process identi-fied in the present article is likely to continue into adulthood(Gilbert et al 2011 Kelly Carter Zuroff amp Borairi 2012Kelly et al 2014) However further research is needed todirectly test the longitudinal relations between self-esteemand self-compassion across the entire human developmentalspectrum

6 | CONCLUSION

In contrast to the initial validation studies of self-compassion which focused on the discriminant validity ofself-compassion with respect to self-esteem our study bringsself-esteem back into the spotlight as an important precursorto the development of self-compassion among young peopleAcross 4 years we found consistent evidence for self-esteemas an antecedent of the development of self-compassion butnot vice versa Our findings have an important implicationfor the design of self-compassion interventions To help ado-lescents become more self-compassionate enhancing theirsense of worthiness and deservingness may be as importantas directly teaching them to be kind to themselves

ACKNOWLEDGMENT

The author(s) disclosed receipt of the following financialsupport for the research authorship andor publication ofthis article Preparation of this manuscript was supportedby Grant DP140103874 from the Australian ResearchCouncil

DONALD ET AL | 9

CONFLICT OF INTERESTS

The author(s) declared no potential conflicts of interestwith respect to the research authorship andor publicationof this article

NOTE1 For completeness we ran these same analyses but without gender as acovariate and obtained identical parameter estimates

ORCID

James N Donald httporcidorg0000-0002-5940-2536Joseph Ciarrochi httporcidorg0000-0003-0471-8100Philip D Parker httporcidorg0000-0002-4604-8566Baljinder K Sahdra httporcidorg0000-0001-5064-7783

REFERENCES

Amato P R amp Keith B (1991) Parental Divorce and the Weil-Being of Children A Meta-Analysis Psychological Bulletin 11026ndash46

Atkins P W B amp Parker S K (2012) Understanding individualcompassion in organizations The role of appraisals and psycho-logical flexibility Academy of Management Review AMR-10 1ndash53 httpsdoi 105465amr20100490

Australian Bureau of Statistics (2012) Yearbook Australia Primaryand secondary education schooling structures (Document13010) Canberra Australia Government Printer

Baraldi A N amp Enders C K (2010) An introduction to modernmissing data analyses Journal of School Psychology 48 5ndash37httpsdoiorg101016jjsp200910001

Barry C T Loflin D C amp Doucette H (2015) Adolescent self-compassion Associations with narcissism self-esteem aggres-sion and internalizing symptoms in at-risk males Personality andIndividual Differences 77 118ndash123 httpsdoiorg101016jpaid201412036

Baumeister R F Bushman B J Campbell W K Baumeister RF Bushman B J amp Campbell W K (2000) Self-esteem nar-cissism and aggression Does violence result from low self-esteemor from threatened egotism Current Directions in PsychologicalScience 9 26ndash29 httpsdxdoiorg1011111467-872100053

Baumeister R F Campbell J D Krueger J I amp Vohs K D(2003) Does high self-esteem cause better performance interper-sonal success happiness or healthier lifestyles PsychologicalScience in the Public Interest 4 1ndash44 httpsdxdoiorg1011111529-100601431

Baumeister R F Heatherton T F amp Tice D M (1993) Whenego threats lead to self-regulation failure Negative consequencesof high self-esteem Journal of Personality and Social Psychol-ogy 64 141ndash156 httpsdxdoiorg1010370022-3514641141

Baumeister R F Smart L amp Boden J M (1996) Relation ofthreatened egotism to violence and aggression The dark side ofhigh self-esteem Psychological Review 103 5ndash33 httpsdxdoiorg1010370033-295X10315

Batson C D Eklund J H Chermok V L Hoyt J L amp OrtizB G (2007) An additional antecedent of empathic concern Val-uing the welfare of the person in need Journal of Personalityand Social Psychology 93 65ndash74 httpsdxdoiorg1010370022-351493165

Bluth K Gaylord S A Campo R A Mullarkey M C ampHobbs L (2016) Making friends with yourself A mixed meth-ods pilot study of a mindful self-compassion program for adoles-cents Mindfulness 7 479ndash492 httpsdoiorg101007s12671-015-0476-6

Bollen K A (1989) Structural equations with latent variables NewYork NY Wiley httpsdxdoiorg1010029781118619179

Carver C S amp Connor-Smith J (2010) Personality and copingAnnual Review of Psychology 61 679ndash704 httpsdoiorg101146annurevpsych093008100352

Chen F F (2007) Sensitivity of goodness of fit indexes to lack ofmeasurement invariance Structural Equation Modeling A Multi-disciplinary Journal 14 464ndash504 httpsdoiorg10108010705510701301834

Cheng H amp Furnham A (2004) Perceived parental rearing styleself-esteem and self-criticism as predictors of happiness Journalof Happiness Studies 5 1ndash21 httpsdoiorg101023BJOHS00000217043526705

Cheung G W amp Rensvold R B (2002) Evaluating goodness-of-fit indexes for testing measurement invariance Structural Equa-tion Modeling A Multidisciplinary Journal 9 233ndash255 httpsdoiorg101207S15328007SEM0902

Ciarrochi J Parker P Kashdan T B Heaven P C L amp BarkusE (2015) Hope and emotional well-being A six-year study todistinguish antecedents correlates and consequences Journal ofPositive Psychology 10 1ndash13 httpsdoiorg1010801743976020151015154

Ciarrochi J Parker P Sahdra B Marshall S Jackson C ampGloster A T (2016) The development of compulsive Internetuse and mental health A four-year study of adolescence Devel-opmental Psychology 52 272ndash283 httpsdxdoiorg101037dev0000070

Crocker J amp Park L E (2004) The costly pursuit of self-esteemPsychological Bulletin 130 392ndash414 httpsdoiorg1010370033-29091303392

Di Stefano C amp Motl R W (2006) Further investigating methodeffects associated with negatively worded items on self-report sur-veys Structural Equation Modeling A Multidisciplinary Journal13 440ndash464 httpsdoiorg101207s15328007sem1303_6

Donnellan M B Trzesniewski K H Robins R W Moffitt TE amp Caspi A (2005) Low self-esteem is related to aggressionantisocial behavior and delinquency Psychological Science 16328ndash335 httpsdoiorg101111j0956-7976200501535x

Enders C (2010) Applied missing data analysis New York NYGuilford Press httpsdxdoiorg1012691ajssm-4-4-1

Fiske S T Cuddy A J C Glick P amp Xu J (2002) A model of(often mixed) stereotype content Competence and warmth respec-tively follow from perceived status and competition Journal ofPersonality and Social Psychology 82 878ndash902 httpsdxdoiorg1010370022-3514826878

10 | DONALD ET AL

explanations showing that evaluations of worthiness of theself are important antecedents of the capacity to extend com-passion to oneself

Research on the antecedents of other-oriented compassionmay also help to explain our findings While responding incompassionate ways to others appraisals of deservingnesshave been identified as a core elementmdashalong with noticinganotherrsquos suffering feeling empathy for him or her and actingcompassionately (Atkins amp Parker 2012) Further there isevidence that when people appraise others as being responsi-ble for their own suffering they are less likely to extend com-passion to them (Rudolph Roesch Greitemeyer amp Weiner2004) Similarly when a person appraises another to be colduncooperative or nasty he or she is less likely to extend com-passion to him or her (Batson Eklund Chermok Hoyt ampOrtiz 2007 Fiske Cuddy Glick amp Xu 2002) Such proc-esses may occur in relation to the self as well When individu-als appraise themselves as responsible for their own sufferingor more broadly in a negative way (eg as being unworthy)they may be less likely to develop compassion toward them-selves Conversely having positive self-oriented evaluationsof deservingness and worth appears to be an important ante-cedent of the development of the capacity to extend compas-sion toward oneself when faced with difficulty

One other explanation for our findings is that self-compassion may function as a coping mechanism that indi-viduals with high self-esteem use to deal effectively with

rejection and failures Adolescents with high levels of self-esteem tend to engage in more adaptive and less avoidantcoping responses following setbacks (Mullis amp Chapman2000) Self-compassion involves engaging with and accept-ing onersquos own suffering rather than avoiding it so in thatsense it can be viewed as an approach form of coping (Car-ver amp Connor-Smith 2010 Neff et al 2005)

Further our findings have potentially important implica-tions for the understanding of how self-compassion is devel-oped among adolescents and perhaps more broadly Theoriesof self-compassion emphasize the behavior of noticing suffer-ing extending kindness toward oneself and recognizing suf-fering as common to humanity (Neff 2003 Neff et al2005) Our research suggests that understanding the kinds ofself-evaluations that young people make may be critical tohelping develop self-compassion In recent years researchexamining the efficacy of compassion-based interventionshas been growing (see Galante Galante Bekkers amp Gal-lacher 2014 for a review) including among young people(Bluth Gaylord Campo Mullarkey amp Hobbs 2016 Reddyet al 2013) Our findings suggest that for enhancing the effi-cacy of such interventions it may be important to understandand influence self-oriented evaluations in particular by ensur-ing that young people feel worthy of respect and kindness

Lastly the lack of support for the self-compassion asantecedent model is noteworthy We anticipated that becauseself-compassion involves cultivating positive self-orientedemotions and forgiving oneself for failures and mistakesthis would lead to positive self-oriented evaluations (ieenhanced self-esteem) However self-compassion research-ers have distinguished between self-compassion and self-esteem on the basis that the former does not involve cultivat-ing positive self-evaluations whereas the latter does (Neff2003 Neff amp Vonk 2009) Recent evidence has shown thatrather than reinforcing onersquos sense of self and identity self-compassion is associated with efforts to transcend it (Lindsayamp Creswell 2014 Neff amp Beretvas 2012 Neff amp Pommier2013 Welp amp Brown 2016) Further support for this comesfrom studies showing that other-oriented compassion

Self-esteem

Self-esteem

Self-esteem

Self-esteem

Self- compassion

Self-compassion

Self-compassion

Self- compassion

05

16

71

72

Grade 9 Grade 10 Grade 11 Grade 12

76

17

60

03

75

12

51

03 32 43 43 17

FIGURE 2 Path estimates from SEM 2with only single-year lags estimatedNote p lt 001

Self-esteem

Self-esteem

Self-compassion

Self- compassion

14 05

73

60

39 33

FIGURE 3 Path estimates from SEM 3 the developmental equilib-riummodel with mean standardized path estimates shownNote p lt 001

8 | DONALD ET AL

interventions lead to more pro-social behaviur and implicateregions in the brain associated with self-transcendence(McCall Steinbeis Ricard amp Singer 2014 Weng et al2013) This evidence suggests that cultivating self-compassion does not lead to self-evaluations of worthinessor deservingness perhaps because it leads to a more tran-scendent sense of self

5 | LIMITATIONS AND FUTUREDIRECTIONS

This study has several limitations First although this studyprovides valuable information regarding the temporal order-ing of the relations between self-compassion and self-esteemand although we tested for links between the study variablesand demographic covariates it was not possible to com-pletely rule out the possibility that additional exogenous vari-ables accounted for these results (Morgan amp Winship 2007)Experimental research is needed to test these relationships ina way that more robustly controls for potential confoundsFor example future self-compassion research could examinedifferential effects of compassion interventions where self-evaluations are explored and made explicit prior to self-compassion training and self-compassion interventionswhere this does not occur Another limitation is that we werenot able to test mediators of the link between self-esteem andself-compassion in the self-compassion as consequencemodel As discussed we expect that appraisals of deserving-ness may be an important mediator of this link Futureresearch could explicitly test this possibility by measuringfor example self-criticism and self-reassurance (GilbertClarke Hempel Miles amp Irons 2004) as potential mediatingvariables

Future research could also examine other possible con-textual variables that explain the relations between self-esteem and self-compassion over time For example socialsupport has been developmentally linked with self-esteem(Marshall et al 2013) and it may be that among individualswith relatively high social support the role of self-esteem indeveloping self-compassion is less important than for moresocially isolated adolescents Demographic factors such aseducation level or socioeconomic status may also moderaterelations between self-esteem and self-compassion across thedevelopmental spectrum Lastly it may be that differentforms of self-esteem influence the development of self-compassion in differential ways For example having vari-able or contingent self-esteem which have both been associ-ated with defensive and maladaptive behavior (KernisLakey amp Heppner 2008) may in fact undermine the devel-opment of self-compassion whereas relatively stable or non-contingent self-esteem enhances it Future research couldexplore these and other potential moderators

Additionally the present study used a convenience sam-ple of 17 Catholic schools across two Australian statesAlthough there is evidence that these schools were broadlyrepresentative of Australiarsquos socioeconomic standingincluded a mix of urban regional and rural schools and hadan almost even gender balance future research could test thelinks between self-esteem and self-compassion amongnationally representative samples Australia is an affluentsociety and future research could explore the extent to whichour findings can be replicated in lower socioeconomic sam-ples where the effects of self-esteem on self-compassion maybe different

Lastly the present study focused on adolescents (Years9ndash12) and our conclusions are therefore constrained to thisage group Future research is needed to understand whetherthe longitudinal relations between self-esteem and self-compassion identified in the present article extend into adult-hood and if so whether they strengthen weaken or aremaintained at a similar level There is evidence that self-evaluations do indeed influence the capacity for self-compassion among adults suggesting that the process identi-fied in the present article is likely to continue into adulthood(Gilbert et al 2011 Kelly Carter Zuroff amp Borairi 2012Kelly et al 2014) However further research is needed todirectly test the longitudinal relations between self-esteemand self-compassion across the entire human developmentalspectrum

6 | CONCLUSION

In contrast to the initial validation studies of self-compassion which focused on the discriminant validity ofself-compassion with respect to self-esteem our study bringsself-esteem back into the spotlight as an important precursorto the development of self-compassion among young peopleAcross 4 years we found consistent evidence for self-esteemas an antecedent of the development of self-compassion butnot vice versa Our findings have an important implicationfor the design of self-compassion interventions To help ado-lescents become more self-compassionate enhancing theirsense of worthiness and deservingness may be as importantas directly teaching them to be kind to themselves

ACKNOWLEDGMENT

The author(s) disclosed receipt of the following financialsupport for the research authorship andor publication ofthis article Preparation of this manuscript was supportedby Grant DP140103874 from the Australian ResearchCouncil

DONALD ET AL | 9

CONFLICT OF INTERESTS

The author(s) declared no potential conflicts of interestwith respect to the research authorship andor publicationof this article

NOTE1 For completeness we ran these same analyses but without gender as acovariate and obtained identical parameter estimates

ORCID

James N Donald httporcidorg0000-0002-5940-2536Joseph Ciarrochi httporcidorg0000-0003-0471-8100Philip D Parker httporcidorg0000-0002-4604-8566Baljinder K Sahdra httporcidorg0000-0001-5064-7783

REFERENCES

Amato P R amp Keith B (1991) Parental Divorce and the Weil-Being of Children A Meta-Analysis Psychological Bulletin 11026ndash46

Atkins P W B amp Parker S K (2012) Understanding individualcompassion in organizations The role of appraisals and psycho-logical flexibility Academy of Management Review AMR-10 1ndash53 httpsdoi 105465amr20100490

Australian Bureau of Statistics (2012) Yearbook Australia Primaryand secondary education schooling structures (Document13010) Canberra Australia Government Printer

Baraldi A N amp Enders C K (2010) An introduction to modernmissing data analyses Journal of School Psychology 48 5ndash37httpsdoiorg101016jjsp200910001

Barry C T Loflin D C amp Doucette H (2015) Adolescent self-compassion Associations with narcissism self-esteem aggres-sion and internalizing symptoms in at-risk males Personality andIndividual Differences 77 118ndash123 httpsdoiorg101016jpaid201412036

Baumeister R F Bushman B J Campbell W K Baumeister RF Bushman B J amp Campbell W K (2000) Self-esteem nar-cissism and aggression Does violence result from low self-esteemor from threatened egotism Current Directions in PsychologicalScience 9 26ndash29 httpsdxdoiorg1011111467-872100053

Baumeister R F Campbell J D Krueger J I amp Vohs K D(2003) Does high self-esteem cause better performance interper-sonal success happiness or healthier lifestyles PsychologicalScience in the Public Interest 4 1ndash44 httpsdxdoiorg1011111529-100601431

Baumeister R F Heatherton T F amp Tice D M (1993) Whenego threats lead to self-regulation failure Negative consequencesof high self-esteem Journal of Personality and Social Psychol-ogy 64 141ndash156 httpsdxdoiorg1010370022-3514641141

Baumeister R F Smart L amp Boden J M (1996) Relation ofthreatened egotism to violence and aggression The dark side ofhigh self-esteem Psychological Review 103 5ndash33 httpsdxdoiorg1010370033-295X10315

Batson C D Eklund J H Chermok V L Hoyt J L amp OrtizB G (2007) An additional antecedent of empathic concern Val-uing the welfare of the person in need Journal of Personalityand Social Psychology 93 65ndash74 httpsdxdoiorg1010370022-351493165

Bluth K Gaylord S A Campo R A Mullarkey M C ampHobbs L (2016) Making friends with yourself A mixed meth-ods pilot study of a mindful self-compassion program for adoles-cents Mindfulness 7 479ndash492 httpsdoiorg101007s12671-015-0476-6

Bollen K A (1989) Structural equations with latent variables NewYork NY Wiley httpsdxdoiorg1010029781118619179

Carver C S amp Connor-Smith J (2010) Personality and copingAnnual Review of Psychology 61 679ndash704 httpsdoiorg101146annurevpsych093008100352

Chen F F (2007) Sensitivity of goodness of fit indexes to lack ofmeasurement invariance Structural Equation Modeling A Multi-disciplinary Journal 14 464ndash504 httpsdoiorg10108010705510701301834

Cheng H amp Furnham A (2004) Perceived parental rearing styleself-esteem and self-criticism as predictors of happiness Journalof Happiness Studies 5 1ndash21 httpsdoiorg101023BJOHS00000217043526705

Cheung G W amp Rensvold R B (2002) Evaluating goodness-of-fit indexes for testing measurement invariance Structural Equa-tion Modeling A Multidisciplinary Journal 9 233ndash255 httpsdoiorg101207S15328007SEM0902

Ciarrochi J Parker P Kashdan T B Heaven P C L amp BarkusE (2015) Hope and emotional well-being A six-year study todistinguish antecedents correlates and consequences Journal ofPositive Psychology 10 1ndash13 httpsdoiorg1010801743976020151015154

Ciarrochi J Parker P Sahdra B Marshall S Jackson C ampGloster A T (2016) The development of compulsive Internetuse and mental health A four-year study of adolescence Devel-opmental Psychology 52 272ndash283 httpsdxdoiorg101037dev0000070

Crocker J amp Park L E (2004) The costly pursuit of self-esteemPsychological Bulletin 130 392ndash414 httpsdoiorg1010370033-29091303392

Di Stefano C amp Motl R W (2006) Further investigating methodeffects associated with negatively worded items on self-report sur-veys Structural Equation Modeling A Multidisciplinary Journal13 440ndash464 httpsdoiorg101207s15328007sem1303_6

Donnellan M B Trzesniewski K H Robins R W Moffitt TE amp Caspi A (2005) Low self-esteem is related to aggressionantisocial behavior and delinquency Psychological Science 16328ndash335 httpsdoiorg101111j0956-7976200501535x

Enders C (2010) Applied missing data analysis New York NYGuilford Press httpsdxdoiorg1012691ajssm-4-4-1

Fiske S T Cuddy A J C Glick P amp Xu J (2002) A model of(often mixed) stereotype content Competence and warmth respec-tively follow from perceived status and competition Journal ofPersonality and Social Psychology 82 878ndash902 httpsdxdoiorg1010370022-3514826878

10 | DONALD ET AL

interventions lead to more pro-social behaviur and implicateregions in the brain associated with self-transcendence(McCall Steinbeis Ricard amp Singer 2014 Weng et al2013) This evidence suggests that cultivating self-compassion does not lead to self-evaluations of worthinessor deservingness perhaps because it leads to a more tran-scendent sense of self

5 | LIMITATIONS AND FUTUREDIRECTIONS

This study has several limitations First although this studyprovides valuable information regarding the temporal order-ing of the relations between self-compassion and self-esteemand although we tested for links between the study variablesand demographic covariates it was not possible to com-pletely rule out the possibility that additional exogenous vari-ables accounted for these results (Morgan amp Winship 2007)Experimental research is needed to test these relationships ina way that more robustly controls for potential confoundsFor example future self-compassion research could examinedifferential effects of compassion interventions where self-evaluations are explored and made explicit prior to self-compassion training and self-compassion interventionswhere this does not occur Another limitation is that we werenot able to test mediators of the link between self-esteem andself-compassion in the self-compassion as consequencemodel As discussed we expect that appraisals of deserving-ness may be an important mediator of this link Futureresearch could explicitly test this possibility by measuringfor example self-criticism and self-reassurance (GilbertClarke Hempel Miles amp Irons 2004) as potential mediatingvariables

Future research could also examine other possible con-textual variables that explain the relations between self-esteem and self-compassion over time For example socialsupport has been developmentally linked with self-esteem(Marshall et al 2013) and it may be that among individualswith relatively high social support the role of self-esteem indeveloping self-compassion is less important than for moresocially isolated adolescents Demographic factors such aseducation level or socioeconomic status may also moderaterelations between self-esteem and self-compassion across thedevelopmental spectrum Lastly it may be that differentforms of self-esteem influence the development of self-compassion in differential ways For example having vari-able or contingent self-esteem which have both been associ-ated with defensive and maladaptive behavior (KernisLakey amp Heppner 2008) may in fact undermine the devel-opment of self-compassion whereas relatively stable or non-contingent self-esteem enhances it Future research couldexplore these and other potential moderators

Additionally the present study used a convenience sam-ple of 17 Catholic schools across two Australian statesAlthough there is evidence that these schools were broadlyrepresentative of Australiarsquos socioeconomic standingincluded a mix of urban regional and rural schools and hadan almost even gender balance future research could test thelinks between self-esteem and self-compassion amongnationally representative samples Australia is an affluentsociety and future research could explore the extent to whichour findings can be replicated in lower socioeconomic sam-ples where the effects of self-esteem on self-compassion maybe different

Lastly the present study focused on adolescents (Years9ndash12) and our conclusions are therefore constrained to thisage group Future research is needed to understand whetherthe longitudinal relations between self-esteem and self-compassion identified in the present article extend into adult-hood and if so whether they strengthen weaken or aremaintained at a similar level There is evidence that self-evaluations do indeed influence the capacity for self-compassion among adults suggesting that the process identi-fied in the present article is likely to continue into adulthood(Gilbert et al 2011 Kelly Carter Zuroff amp Borairi 2012Kelly et al 2014) However further research is needed todirectly test the longitudinal relations between self-esteemand self-compassion across the entire human developmentalspectrum

6 | CONCLUSION

In contrast to the initial validation studies of self-compassion which focused on the discriminant validity ofself-compassion with respect to self-esteem our study bringsself-esteem back into the spotlight as an important precursorto the development of self-compassion among young peopleAcross 4 years we found consistent evidence for self-esteemas an antecedent of the development of self-compassion butnot vice versa Our findings have an important implicationfor the design of self-compassion interventions To help ado-lescents become more self-compassionate enhancing theirsense of worthiness and deservingness may be as importantas directly teaching them to be kind to themselves

ACKNOWLEDGMENT

The author(s) disclosed receipt of the following financialsupport for the research authorship andor publication ofthis article Preparation of this manuscript was supportedby Grant DP140103874 from the Australian ResearchCouncil

DONALD ET AL | 9

CONFLICT OF INTERESTS

The author(s) declared no potential conflicts of interestwith respect to the research authorship andor publicationof this article

NOTE1 For completeness we ran these same analyses but without gender as acovariate and obtained identical parameter estimates

ORCID

James N Donald httporcidorg0000-0002-5940-2536Joseph Ciarrochi httporcidorg0000-0003-0471-8100Philip D Parker httporcidorg0000-0002-4604-8566Baljinder K Sahdra httporcidorg0000-0001-5064-7783

REFERENCES

Amato P R amp Keith B (1991) Parental Divorce and the Weil-Being of Children A Meta-Analysis Psychological Bulletin 11026ndash46

Atkins P W B amp Parker S K (2012) Understanding individualcompassion in organizations The role of appraisals and psycho-logical flexibility Academy of Management Review AMR-10 1ndash53 httpsdoi 105465amr20100490

Australian Bureau of Statistics (2012) Yearbook Australia Primaryand secondary education schooling structures (Document13010) Canberra Australia Government Printer

Baraldi A N amp Enders C K (2010) An introduction to modernmissing data analyses Journal of School Psychology 48 5ndash37httpsdoiorg101016jjsp200910001

Barry C T Loflin D C amp Doucette H (2015) Adolescent self-compassion Associations with narcissism self-esteem aggres-sion and internalizing symptoms in at-risk males Personality andIndividual Differences 77 118ndash123 httpsdoiorg101016jpaid201412036

Baumeister R F Bushman B J Campbell W K Baumeister RF Bushman B J amp Campbell W K (2000) Self-esteem nar-cissism and aggression Does violence result from low self-esteemor from threatened egotism Current Directions in PsychologicalScience 9 26ndash29 httpsdxdoiorg1011111467-872100053

Baumeister R F Campbell J D Krueger J I amp Vohs K D(2003) Does high self-esteem cause better performance interper-sonal success happiness or healthier lifestyles PsychologicalScience in the Public Interest 4 1ndash44 httpsdxdoiorg1011111529-100601431

Baumeister R F Heatherton T F amp Tice D M (1993) Whenego threats lead to self-regulation failure Negative consequencesof high self-esteem Journal of Personality and Social Psychol-ogy 64 141ndash156 httpsdxdoiorg1010370022-3514641141

Baumeister R F Smart L amp Boden J M (1996) Relation ofthreatened egotism to violence and aggression The dark side ofhigh self-esteem Psychological Review 103 5ndash33 httpsdxdoiorg1010370033-295X10315

Batson C D Eklund J H Chermok V L Hoyt J L amp OrtizB G (2007) An additional antecedent of empathic concern Val-uing the welfare of the person in need Journal of Personalityand Social Psychology 93 65ndash74 httpsdxdoiorg1010370022-351493165

Bluth K Gaylord S A Campo R A Mullarkey M C ampHobbs L (2016) Making friends with yourself A mixed meth-ods pilot study of a mindful self-compassion program for adoles-cents Mindfulness 7 479ndash492 httpsdoiorg101007s12671-015-0476-6

Bollen K A (1989) Structural equations with latent variables NewYork NY Wiley httpsdxdoiorg1010029781118619179

Carver C S amp Connor-Smith J (2010) Personality and copingAnnual Review of Psychology 61 679ndash704 httpsdoiorg101146annurevpsych093008100352

Chen F F (2007) Sensitivity of goodness of fit indexes to lack ofmeasurement invariance Structural Equation Modeling A Multi-disciplinary Journal 14 464ndash504 httpsdoiorg10108010705510701301834

Cheng H amp Furnham A (2004) Perceived parental rearing styleself-esteem and self-criticism as predictors of happiness Journalof Happiness Studies 5 1ndash21 httpsdoiorg101023BJOHS00000217043526705

Cheung G W amp Rensvold R B (2002) Evaluating goodness-of-fit indexes for testing measurement invariance Structural Equa-tion Modeling A Multidisciplinary Journal 9 233ndash255 httpsdoiorg101207S15328007SEM0902

Ciarrochi J Parker P Kashdan T B Heaven P C L amp BarkusE (2015) Hope and emotional well-being A six-year study todistinguish antecedents correlates and consequences Journal ofPositive Psychology 10 1ndash13 httpsdoiorg1010801743976020151015154

Ciarrochi J Parker P Sahdra B Marshall S Jackson C ampGloster A T (2016) The development of compulsive Internetuse and mental health A four-year study of adolescence Devel-opmental Psychology 52 272ndash283 httpsdxdoiorg101037dev0000070

Crocker J amp Park L E (2004) The costly pursuit of self-esteemPsychological Bulletin 130 392ndash414 httpsdoiorg1010370033-29091303392

Di Stefano C amp Motl R W (2006) Further investigating methodeffects associated with negatively worded items on self-report sur-veys Structural Equation Modeling A Multidisciplinary Journal13 440ndash464 httpsdoiorg101207s15328007sem1303_6

Donnellan M B Trzesniewski K H Robins R W Moffitt TE amp Caspi A (2005) Low self-esteem is related to aggressionantisocial behavior and delinquency Psychological Science 16328ndash335 httpsdoiorg101111j0956-7976200501535x

Enders C (2010) Applied missing data analysis New York NYGuilford Press httpsdxdoiorg1012691ajssm-4-4-1

Fiske S T Cuddy A J C Glick P amp Xu J (2002) A model of(often mixed) stereotype content Competence and warmth respec-tively follow from perceived status and competition Journal ofPersonality and Social Psychology 82 878ndash902 httpsdxdoiorg1010370022-3514826878

10 | DONALD ET AL

CONFLICT OF INTERESTS

The author(s) declared no potential conflicts of interestwith respect to the research authorship andor publicationof this article

NOTE1 For completeness we ran these same analyses but without gender as acovariate and obtained identical parameter estimates

ORCID

James N Donald httporcidorg0000-0002-5940-2536Joseph Ciarrochi httporcidorg0000-0003-0471-8100Philip D Parker httporcidorg0000-0002-4604-8566Baljinder K Sahdra httporcidorg0000-0001-5064-7783

REFERENCES

Amato P R amp Keith B (1991) Parental Divorce and the Weil-Being of Children A Meta-Analysis Psychological Bulletin 11026ndash46

Atkins P W B amp Parker S K (2012) Understanding individualcompassion in organizations The role of appraisals and psycho-logical flexibility Academy of Management Review AMR-10 1ndash53 httpsdoi 105465amr20100490

Australian Bureau of Statistics (2012) Yearbook Australia Primaryand secondary education schooling structures (Document13010) Canberra Australia Government Printer

Baraldi A N amp Enders C K (2010) An introduction to modernmissing data analyses Journal of School Psychology 48 5ndash37httpsdoiorg101016jjsp200910001

Barry C T Loflin D C amp Doucette H (2015) Adolescent self-compassion Associations with narcissism self-esteem aggres-sion and internalizing symptoms in at-risk males Personality andIndividual Differences 77 118ndash123 httpsdoiorg101016jpaid201412036

Baumeister R F Bushman B J Campbell W K Baumeister RF Bushman B J amp Campbell W K (2000) Self-esteem nar-cissism and aggression Does violence result from low self-esteemor from threatened egotism Current Directions in PsychologicalScience 9 26ndash29 httpsdxdoiorg1011111467-872100053

Baumeister R F Campbell J D Krueger J I amp Vohs K D(2003) Does high self-esteem cause better performance interper-sonal success happiness or healthier lifestyles PsychologicalScience in the Public Interest 4 1ndash44 httpsdxdoiorg1011111529-100601431

Baumeister R F Heatherton T F amp Tice D M (1993) Whenego threats lead to self-regulation failure Negative consequencesof high self-esteem Journal of Personality and Social Psychol-ogy 64 141ndash156 httpsdxdoiorg1010370022-3514641141

Baumeister R F Smart L amp Boden J M (1996) Relation ofthreatened egotism to violence and aggression The dark side ofhigh self-esteem Psychological Review 103 5ndash33 httpsdxdoiorg1010370033-295X10315

Batson C D Eklund J H Chermok V L Hoyt J L amp OrtizB G (2007) An additional antecedent of empathic concern Val-uing the welfare of the person in need Journal of Personalityand Social Psychology 93 65ndash74 httpsdxdoiorg1010370022-351493165

Bluth K Gaylord S A Campo R A Mullarkey M C ampHobbs L (2016) Making friends with yourself A mixed meth-ods pilot study of a mindful self-compassion program for adoles-cents Mindfulness 7 479ndash492 httpsdoiorg101007s12671-015-0476-6

Bollen K A (1989) Structural equations with latent variables NewYork NY Wiley httpsdxdoiorg1010029781118619179

Carver C S amp Connor-Smith J (2010) Personality and copingAnnual Review of Psychology 61 679ndash704 httpsdoiorg101146annurevpsych093008100352

Chen F F (2007) Sensitivity of goodness of fit indexes to lack ofmeasurement invariance Structural Equation Modeling A Multi-disciplinary Journal 14 464ndash504 httpsdoiorg10108010705510701301834

Cheng H amp Furnham A (2004) Perceived parental rearing styleself-esteem and self-criticism as predictors of happiness Journalof Happiness Studies 5 1ndash21 httpsdoiorg101023BJOHS00000217043526705

Cheung G W amp Rensvold R B (2002) Evaluating goodness-of-fit indexes for testing measurement invariance Structural Equa-tion Modeling A Multidisciplinary Journal 9 233ndash255 httpsdoiorg101207S15328007SEM0902

Ciarrochi J Parker P Kashdan T B Heaven P C L amp BarkusE (2015) Hope and emotional well-being A six-year study todistinguish antecedents correlates and consequences Journal ofPositive Psychology 10 1ndash13 httpsdoiorg1010801743976020151015154

Ciarrochi J Parker P Sahdra B Marshall S Jackson C ampGloster A T (2016) The development of compulsive Internetuse and mental health A four-year study of adolescence Devel-opmental Psychology 52 272ndash283 httpsdxdoiorg101037dev0000070

Crocker J amp Park L E (2004) The costly pursuit of self-esteemPsychological Bulletin 130 392ndash414 httpsdoiorg1010370033-29091303392

Di Stefano C amp Motl R W (2006) Further investigating methodeffects associated with negatively worded items on self-report sur-veys Structural Equation Modeling A Multidisciplinary Journal13 440ndash464 httpsdoiorg101207s15328007sem1303_6

Donnellan M B Trzesniewski K H Robins R W Moffitt TE amp Caspi A (2005) Low self-esteem is related to aggressionantisocial behavior and delinquency Psychological Science 16328ndash335 httpsdoiorg101111j0956-7976200501535x

Enders C (2010) Applied missing data analysis New York NYGuilford Press httpsdxdoiorg1012691ajssm-4-4-1

Fiske S T Cuddy A J C Glick P amp Xu J (2002) A model of(often mixed) stereotype content Competence and warmth respec-tively follow from perceived status and competition Journal ofPersonality and Social Psychology 82 878ndash902 httpsdxdoiorg1010370022-3514826878

10 | DONALD ET AL

Fredrickson B L (2001) The role of positive emotions in positivepsychology The broaden-and-build theory of positive emotionsAmerican Psychologist 56 218ndash226 httpsdxdoiorg1010370003-066X563218

Galante J Galante I Bekkers M-J amp Gallacher J (2014) Effectof kindness-based meditation on health and well-being A system-atic review and meta-analysis Journal of Consulting and ClinicalPsychology 82 1101ndash1114 httpsdoiorg101037a0037249

Gelman A amp Hill J (2007) Data analysis using regression andmultilevelhierarchical models (Vol 1) New York NY Cam-bridge University Press httpsdoiorg101017CBO9780511790942

Gilbert P Clarke M Hempel S Miles J N V amp Irons C(2004) Criticizing and reassuring oneself An exploration offorms style and reasons in female students British Journal ofClinical Psychology 43 31ndash50 httpsdoiorg101348014466504772812959

Gilbert P McEwan K Matos M amp Rivis A (2011) Fears ofcompassion Development of three self-report measures Psychol-ogy and Psychotherapy 84 239ndash255 httpsdoiorg101348147608310X526511

Goyen M J amp Anshel M H (1998) Sources of acute competitivestress and use of coping strategies as a function of age and gen-der Journal of Applied Developmental Psychology 19 469ndash486httpsdoiorg101016S0193-3973(99)80051-3

Heaven P C L Ciarrochi J amp Hurrell K (2010) The distinctive-ness and utility of a brief measure of alexithymia for adolescentsPersonality and Individual Differences 49 222ndash227 httpsdoiorg101016jpaid201003039

Hu L amp Bentler P M (1999) Cutoff criteria for fit indexes incovariance structure analysis Conventional criteria versus newalternatives Structural Equation Modeling A MultidisciplinaryJournal 6 1ndash55 httpsdoiorg10108010705519909540118

Kelly A C Carter J C Zuroff D C amp Borairi S (2012) (EDE-Q Reliability) Self-compassion and fear of self-compassion inter-act to predict response to eating disorders treatment A prelimi-nary investigation Psychotherapy Research 23 1ndash13 httpsdoiorg101080105033072012717310

Kelly A C Vimalakanthan K amp Carter J C (2014) Understand-ing the roles of self-esteem self-compassion and fear of self-compassion in eating disorder pathology An examination offemale students and eating disorder patients Eating Behaviors15 388ndash391 httpsdoiorg101016jeatbeh201404008

Kernis M H Lakey C E amp Heppner W L (2008) Secure versusfragile high self-esteem as a predictor of verbal defensivenessconverging findings across three different markers Journal ofPersonality 76 477ndash512 httpsdoiorg101111j1467-6494200800493x

Kling K C Hyde J S Showers C J amp Buswell B N (1999)Gender differences in self-esteem A meta-analysis PsychologicalBulletin 125 470ndash500 httpsdoiorg1010370033-29091254470

Leary M R Tate E B Adams C E Allen A B amp Hancock J(2007) Self-compassion and reactions to unpleasant self-relevantevents The implications of treating oneself kindly Journal ofPersonality and Social Psychology 92 887ndash904 httpsdoiorg1010370022-3514925887

Lindsay E K amp Creswell J D (2014) Helping the self helpothers Self-affirmation increases self-compassion and pro-socialbehaviors Frontiers in Psychology 5 1ndash9 httpsdoiorg103389fpsyg201400421

Little T D (2013) Longitudinal structural equation modeling NewYork NY Guilford Press

Marsh H W amp Craven R G (2006) Reciprocal effects of self-concept and performance from a multidimensional perspectiveBeyond seductive pleasure and unidimensional perspectives Per-spectives on Psychological Science 1 133ndash163 httpsdxdoiorg101111j1745-6916200600010x

Marshall S L Parker P D Ciarrochi J amp Heaven P C L(2013) Is self-esteem a cause or consequence of social supportA 4-year longitudinal study Child Development 85 1275ndash1291httpsdoiorg101111cdev12176

McCall C Steinbeis N Ricard M amp Singer T (2014) Compas-sion meditators show less anger less punishment and more com-pensation of victims in response to fairness violations Frontiersin Behavioral Neuroscience 8 424ndash434 httpsdoiorg103389fnbeh201400424

McDonald R P amp Marsh H W (1990) Choosing a multivariatemodel Noncentrality and goodness of fit Psychological Bulletin107 247ndash255 httpsdoiorg1010370033-29091072247

Millsap R E (2011) Statistical approaches to measurement invari-ance New York NY Routledge httpsdxdoiorg1043249780203821961

Morgan S L amp Winship C (2007) Counterfactuals and causalinference Methods and principles for social research (2nd ed)Cambridge Cambridge University Press

Mullis R L amp Chapman P (2000) Age Gender and Self-EsteemDifferences in Adolescent Coping Styles Journal of Social Psy-chology 140 539ndash541

Muris P Meesters C Pierik A amp De Kock B (2016) Good forthe self Self-compassion and other self-related constructs in rela-tion to symptoms of anxiety and depression in non-clinicalyouths Journal of Child and Family Studies 25 607ndash617httpsdoiorg101007s10826-015-0235-2

Neff K D (2003) The development and validation of a scale tomeasure self-compassion Self and Identity 2 223ndash250 httpsdoiorg10108015298860309027

Neff K D amp Beretvas S N (2012) The role of self-compassion inromantic relationships Self and Identity 12 37ndash41 httpsdoiorg101080152988682011639548

Neff K D Hsieh Y-P amp Dejitterat K (2005) Self-compassionachievement goals and coping with academic failure Self andIdentity 4 263ndash287 httpsdoiorg10108013576500444000317

Neff K D Kirkpatrick K L amp Rude S S (2007) Self-compas-sion and adaptive psychological functioning Journal of Researchin Personality 41 139ndash154 httpsdoiorg101016jjrp200603004

Neff K amp McGehee P (2010) Self-compassion and psychologicalresilience among adolescents and young adults Self and Identity9 225ndash240 httpsdoiorg10108015298860902979307

Neff K D amp Pommier E (2013) The relationship between self-compassion and other-focused concern among college

DONALD ET AL | 11

undergraduates community adults and practicing meditators Selfand Identity 12 160ndash176 httpsdoiorg101080152988682011649546

Neff K D amp Vonk R (2009) Self-compassion versus global self-esteem Two different ways of relating to oneself Journal of Per-sonality 77 23ndash50 httpsdoiorg101111j1467-6494200800537x

Nussbaum A D amp Dweck C S (2008) Defensiveness versusremediation Self-theories and modes of self-esteem maintenancePersonality amp Social Psychology Bulletin 34 599ndash612 httpsdoiorg1011770146167207312960

Orth U Robins R W amp Widaman K F (2012) Life-span devel-opment of self-esteem and its effects on important life outcomesJournal of Personality and Social Psychology 102 1271ndash1288httpsdoiorg101037a0025558

Pauley G amp McPherson S (2010) The experience and meaning ofcompassion and self-compassion for individuals with depressionor anxiety Psychology and Psychotherapy Theory Researchand Practice 83 129ndash143 httpsdoiorg101348147608309X471000

R Core Team (2017) R A language and environment for statisticalcomputing Vienna Austria R Foundation for Statistical Comput-ing Retrieved from httpwwwR-projectorg

Raes F Pommier E Neff K D amp Van Gucht D (2011) Con-struction and factorial validation of a short form of the Self-Compassion Scale Clinical Psychology amp Psychotherapy 18250ndash255 httpsdoiorg101002cpp702

Reddy S D Tenzin L Brooke N Silva B O Pace T W WCole S P Craighead L W (2013) Cognitive-based com-passion training A promising prevention strategy for at-risk ado-lescents Journal of Child and Family Studies 22 219ndash230httpsdoiorg101007s10826-012-9571-7

Reilly E D Rochlen A B amp Awad G H (2014) Menrsquos self-compassion and self-esteem The moderating roles of shame andmasculine norm adherence Psychology of Men amp Masculinity15 22ndash28 httpsdoiorg101037a0031028

Rosenberg M (1979) Conceiving the self New York NY BasicBooks

Rosenberg M Schooler C Schoenbach C amp Rosenberg F(1995) Global self-esteem and specific self-esteem Different con-cepts different outcomes American Psychological Review 60141ndash156 httpsdoiorg1023072096350

Rosseel Y (2012) lavaan An R package for structural equationmodeling Journal of Statistical Software 48 1ndash36 httpsdxdoiorg1018637jssv048i02

Rudolph U Roesch S C Greitemeyer T amp Weiner B (2004) Ameta-analytic review of help giving and aggression from an attri-butional perspective Contributions to a general theory of

motivation Cognition amp Emotion 18 815ndash848 httpsdxdoiorg10108002699930341000248

Ryan R M amp Brown K W (2003) Why we donrsquot need self-esteem On fundamental needs contingent love and mindfulnessPsychological Inquiry 14 71ndash76

Sowislo J F amp Orth U (2013) Does low self-esteem predictdepression and anxiety A meta-analysis of longitudinal studiesPsychological Bulletin 139 213ndash240 httpsdoiorg101037a0028931

Trzesniewski K H Donnellan M B Moffitt T E Robins RW Poulton R amp Caspi A (2006) Low self-esteem during ado-lescence predicts poor health criminal behavior and limited eco-nomic prospects during adulthood Developmental Psychology42 381ndash390 httpsdoiorg1010370012-1649422381

Twenge J M amp Campbell W K (2001) Age and Birth CohortDifferences in Self-Esteem A Cross-Temporal Meta-AnalysisPersonality and Social Psychology Review 5 321ndash344

Welp L R amp Brown C M (2016) Self-compassion empathy andhelping intentions Journal of Positive Psychology 9 54ndash65httpsdoiorg101080174397602013831465

Weng H Y Fox A S Shackman A J Stodola D E CaldwellJ Z K Olson M C Davidson R J (2013) Compassiontraining alters altruism and neural responses to suffering Psycho-logical Science 24 1171ndash1180 httpsdoiorg1011770956797612469537

Weston R amp Gore P (2006) A brief guide to structural equationmodeling The Counseling Psychologist 34 719ndash751 httpsdoiorg1011770011000006286345

SUPPORTING INFORMATION

Additional Supporting Information may be found online inthe supporting information tab for this article

Table SM1Table SM2Table SM3

How to cite this article Donald JN Ciarrochi JParker PD Sahdra BK Marshall SL Guo J A worthyself is a caring self Examining the developmental rela-tions between self-esteem and self-compassion in ado-lescents J Pers 2017001ndash12 httpsdoiorg101111jopy12340

12 | DONALD ET AL