Self-Esteem Among Children in Grade R in an Urban South African School

This paper presents the first assessment of the Behavioural Rating Scale of Presented Self-Esteem (Haltiwanger, 1989) in South Africa. The analyses are based on teachers’ evaluation of self-esteem of 57 young isiZulu and Sesotho-speaking children attending a South African government-funded urban primary school. Although we found Cronbach’s Alpha to be very high (α = .96), an exploratory factor analysis revealed a possible two-factor solution. However, the second factor did not match the two-factor solution reported in previous research (Fuchs-Beauchamp, 1996) and explained only a small amount of total variance. No self-esteem differences were detected between boys and girls, or between isiZuluand Sesotho-speakers. The association between subjective summary ratings of self-esteem by teachers and the PSE scores in Soweto matches the associations measured in the US by Haltiwanger (1989). Interestingly, teachers’ subjective assessment of children’s future leadership status correlated positively with evaluation of the children’s self-esteem, while teachers’ subjective assessment of being burdened by major problems in the children’s future did not. Measurement issues relating to ecological validity, culture-sensitivity, and subsequent work on self-esteem of children and education in South Africa are discussed.


Introduction
Self-esteem is an important predictor of outcomes such as aspirations, performance, and life satisfaction (Baumeister, Campbell, Krueger, & Vohs, 2003;Diener & Diener, 1995;Schoon, 2001).The present paper assesses self-esteem in young children in a South African township.This assessment is unique in various ways.It is challenging to assess self-esteem in young children and this paper explores the first use of the Behavioural Rating Scale of Presented Self-Esteem (PSE, Haltiwanger, 1989) in a South African urban foundation phase school context, thereby contributing to the growing list of applications of the PSE in different countries and cultural contexts.Beyond the analysis of cross-cultural equivalence of PSE, this paper aims at contributing to a better understanding of the yet unresolved role of self-esteem in South Africa (Eaton & Louw, 2000).
Self-esteem is defined as the joint product of evaluations across salient attributes of the self and the appraisal of others.Self-esteem is therefore an important and evaluative aspect of the self-concept, where the self-concept refers to a person's beliefs and ideas of his or her own self.If information about the self contains a value judgment, it becomes relevant for self-esteem (Greenwald, et al., 2002;Rosenberg, 1979).Self-esteem reflects an evaluation of one's worth and importance (Rosenberg, 1979), and it is related directly and indirectly to school performance (Marsh & Craven, 2006;Marsh & O'Mara, 2008), aspirations (Sheldon, Ryan, Deci, & Kasser, 2004), wellbeing and happiness (Baumeister, et al., 2003;Bergman & Scott, 2001), and many other desirable outcomes.Interestingly, self-esteem does not necessitate a definitional requirement of accuracy (Baumeister, et al., 2003).For instance, individuals with high self-esteem tend to overestimate their cognitive ability and their self-evaluation does not correlate with actual cognitive abilities (Gabriel, Critelli, & Ee, 1994).Self-esteem emerges from a reflective weighing of successes and aspirations (Hart, Atkins, & Tursi, 2006), and it seems to depend on the degree to which actual successes match goals and aspirations (Higgins, 1987;James, 1890).
Self-esteem can be considered as a state or a trait.Researcher who conceptualize self-esteem as a state construct self-esteem as continually fluctuating in response to environmental stimuli (Bergman & Scott, 2001;Kernis, Grannemann, & Barclay, 1992;Leary & Baumeister, 2000).Researchers considering self-esteem as a trait understand it as a personality construct that is relatively stable over time (Rosenberg, 1979).Meta-analyses support a trait-like view as stability coefficients over the life-span of self-esteem are comparable to those found in other personality traits (Trzesniewski, Donnellan, & Robins, 2003), even though it is least stable during childhood (Trzesniewski, Robins, Roberts, & Caspi, 2003).Despite low stability during childhood, children's self-esteem is associated with children's temperament (Zentner & Bates, 2008).During childhood, self-esteem is comparatively high and drops during adolescence (Robins, Trzesniewski, Tracy, Gosling, & Potter, 2002).Research by Coopersmith (1967) found three factors to positively affect children's self-esteem: 1) unconditional positive regard from parents and others, 2) existence of clear and stable standards, and 3) giving the child freedom, autonomy, and respect for behaviour that lies within the established standards.Self-esteem is associated with important outcomes and correlates, especially with regard to further development.People with high self-esteem tend to be clear and consistent about themselves, resulting in a concise self-concept (Campbell, 1990, Campbell & Lavallee, 1993), and they tend to have higher educational and occupational goals (Baumeister, et al., 2003;Schoon, 2001), performance, persistence (Baumeister, et al., 2003), life satisfaction (Diener & Diener, 1995), etc.But the relevance and value of self-esteem as well as its correlates differ across cultures.For example, Diener and Diener (1995) found self-esteem and life satisfaction to be correlated in several nations but the strength of association differs across national cultures.They found the correlation to be dependent on predominant values in a society, i.e. individualism versus collectivism1 or homogeneity2 (Diener & Diener, 1995).
Apart from important cross-cultural issues related to research design such as equivalence (Hambleton, 2005;Harkness, van de Vijver, & Mohler, 2003;Harkness, Villar, & Edwards, 2010), it is important to be aware of cultural differences when investigating psychological phenomena, especially individual differences.So-called western societies value high self-esteem whereas Asian societies, for example, do not (Shaffer, 2009).South Africa is described as a rather collectivistic society with high heterogeneity ratings (Diener & Diener, 1995;Eaton & Louw, 2000).In collectivistic cultures, the self is defined primarily in terms of its relationships with significant others and therefore "connectedness" and "embeddedness" in meaningful relationships are central to the self (Kitayama, 2006).Not only the meaning and role of the self may differ across cultures but also associated behaviours.For instance, people in collectivistic societies have longer, more profound, but fewer interactions (Kitayama, 2006), and they differ in emotional display rules such as the meaning and frequency of smiling (Matsumoto, Takeuchi, Andayani, Kouznetsova, & Krupp, 1998).Although declared an important research field, the role of self-esteem within South African society remains unclear (Eaton & Louw, 2000).

Description of the Behavioural Rating Scale of Presented Self-Esteem Scale
Some researchers conclude young children's self-concept to be too undifferentiated (Harter & Pike, 1984), while others were able to find defined and differentiated self-concepts in children from age 5 (Marsh, Craven & Debus, 1991).According to Harter (1999Harter ( , 2006)), children possess a sense of self but they cannot verbalize it adequately and are therefore not yet able to report on their own self-esteem.Children nevertheless display self-esteem through their actions.This is labelled behaviourally-manifest self-esteem and is assessed with the Behavioural Rating Scale of Presented Self-Esteem in Young Children (PSE, Haltiwanger, 1989).The PSE was designed and has been tested for children from age 4 to 7 in different national and cultural contexts.Teachers or primary caregivers (e.g.parents) rate the behaviour of a child on a 4-point ordinal scale.A respondent indicates whether the child trusts his or her own ideas, knows what she or he wants, is able to make choices and decisions or if the child doesn't trust his or her own ideas, if she or he acts uncertain when making decisions, etc.
The 15 items are designed to measure inferred self-esteem based on questions relating to initiative, preference for challenges, social approach, social-emotional expression, and coping skills.Children with high self-esteem are described as approaching the world with confidence, showing pride in their work, setting goals independently, taking initiative, etc. Children with low self-esteem are described as shying away from challenging tasks, distrusting their own ideas, giving up easily when frustrated, being indecisive, etc. (Haltiwanger, 1989).
The scale showed high reliability in previous studies.For example, Cronbach's Alpha was very high (e.g.α = .98)(Fuchs-Beauchamp, 1996;Haltiwanger, 1989;Verschueren, et al., 1998).The scale also showed considerable stability over a three year period (r = .59)(Verschueren, et al., 1998).Haltiwanger (1989) found the scale to be strongly correlated with teachers' summary ratings of the self-esteem of the child (r = .65),although teachers' judgment about the intelligence of the child correlated only moderately with the PSE.Another study showed the PSE to be significantly correlated with self-assessed competence of the child but not with the child's perception of social acceptance (Verschueren, et al., 1998).
Previous investigations about the factor structure of the PSE demonstrated two possible factor solutions: Fuchs-Beauchamp (1996) suggested a two-factor solution with an Approach Confidence Factor (10 items) and a Social-Emotional Expression Factor (5 items).Verschueren and colleagues (1998) found a single-factor solution to be the most appropriate.A study in Finland demonstrated the scale to work in a non-US-American context (Lemola, et al., 2010), but two other studies found significant gender differences with higher mean scores in self-esteem for girls (Fuchs-Beauchamp, 1996;Verschueren, et al., 1998).
This article reports on the first assessment of the PSE in South Africa and how, in future studies, it may be used to predict specific outcomes such as various aspects of school performance or social development.While it has been well-established that self-esteem is related to school performance and social development among adolescents and adults (e.g.Harter, 1999;Marsh, Dowson, Pietsch, & Walker, 2004;Marsh & O'Mara, 2008), little is known about the nature of these relationships among young children.The present study aims at establishing the usefulness of this scale in the South African context in order to predict subsequent performance including cognitive, emotional, or linguistic development.
Children of the present study visit a school in Soweto.The learners in the school tend to speak two of the South African national languages, namely Sesotho and isiZulu.Most of the children in this school are from a lower or lower-middle class township background.Poverty and its associated effects have been shown to have negative effects on children's development (Donald, Lazarus, & Lolwana, 2010;Fleisch, 2008;Kamper, 2008).The Soweto Panel Research Programme (SPRP) aims at systematically researching resources of and obstacles to development, of which self-esteem may be one of the resources for resilience against adversity and hardship.

Method
This research is part of a longitudinal panel research project in the context of a primary school in Soweto, South Africa.Principal care providers signed a consent form indicating that they permitted their children to participate in this project.
In July 2010, two teachers of the school were interviewed about the self-esteem of the children in their class as measured by the PSE.Of the 57 children assessed, 33 speak isiZulu and 24 Sesotho.The interviews with the teachers consisted of three parts: First, teachers responded to the Behavioural Rating Scale of Presented Self-Esteem.Second, we assessed teachers' subjective summary ratings to explore whether the strength of association between the self-esteem scores and summary ratings in an American context as identified by Haltiwanger (1989) was comparable to that obtained in Soweto.Accordingly, we asked teachers about the children's self-esteem directly, using the following questions: Who are the children with the highest self-esteem in this class?And who are the children with the lowest self-esteem in this class?Third, we were also interested in how self-esteem may be associated with teachers' evaluations of the children's future.We therefore asked the teachers: Which child will become a leader?And which child will have major problems in life?To our surprise, the teachers showed no hesitation in assigning many children into these two groups.

Results
The mean score of the 57 children of the PSE was 3.26 (SD = .72).
As in most applications of the PSE, we did not detect significant differences in the mean score between girls (M = 3.3) and boys (M = 3.2; t(53)= 0.63 , p >.05).We also did not detect significant difference in self-esteem scores between language (M = 3.3 for isiZulu-speakers; M = 3.2 for Sesotho-speakers; t(55) = 0.83, p > .05).
We found reliability to be considerably high with a Cronbach's Alpha coefficient of .96for the sample (α = .94for isiZulu-speakers, n = 33; α = .96for Sesotho-speakers, n = 24), replicating the reliability scores from studies in other countries and cultural contexts.

PSE Item
M SD Exploring the structure of self-esteem as measured by the PSE, we also conducted an exploratory factor analysis (EFA) with the 15-item scale.We initially ran an EFA using an eigenvalue-greater-than-one criterion for factor extraction, allowing for oblique rotation.With this method, two highly correlated factors were extracted (r = .65).
The rotated factor loadings indicated item 2 (smiles infrequently; face often shows sadness or negative feelings vs. smiles readily; face does not often show sadness or negative feelings), item 11 (remains in group activities and gets involved; does not withdraw versus withdraws from group activities; doesn't get involved), and item 15 (shows pride in his/her work or accomplishments; does not show pride in his/her work or accomplishments) to load on the second factor (refer to table 2 for rotated factor loadings).This factor structure is incompatible with the factor structure found by Fuchs-Beauchamp (1996).Finally, while the first factor accounts for 62% of the total variance, the second factor accounts for only 9% of the variance.Based on these initial results, we reran a factor analysis with a single factor extraction as proposed by Verschueren et al. (1998).Based on the evidence from the exploratory factor analysis, i.e. that the items of the second factor load highly on the first factor, the second factor accounts for only a small amount of the total variance extracted, the two factors correlate highly with each other, and the second factor is incompatible with the factor structure found by other researchers, the single-factor solution is the most appropriate for our data3 .This single factor accounted for 62% of the variance, which is a good extraction of total variance and it also represents excellent parsimony.Factor loadings reached a mean of .78.Obtained factor loadings ranged from .52 to .92 (cf. Table 2), which suggests that all items can be judged as good to excellent indicators for this single factor.In the next analysis step, we explored the extent to which a subjective summary rating of self-esteem by teachers relates to the children's PSE score.Biserial correlations were used to analyse the additional questions, which we asked the teachers after they completed the PSE for each child.The mean PSE score correlated with the teachers' subjective summary ratings of the children's high self-esteem (r = .53,p < .001).The teachers' subjective summary rating of the children's low self-esteem correlated significantly higher with the mean PSE score (r = -.76,p < .001).Thus, the association between subjective summary ratings of self-esteem of children and their self-esteem score in America as obtained by Haltiwanger (r = .65;1989)is similar to that obtained in our analysis from school children in Soweto.
In the final analysis step, we examined the relationship between self-esteem as measured by the PSE with a subjective assessment of the teachers in relation to which child could become a future leader and which child would have serious problems throughout its life.We were surprised that teachers unambiguously und unhesitatingly identified among the 5-year olds those who will become leaders and those who will have major problems throughout their life, a finding that merits further empirical study in a different research project.Future leadership as assessed by teachers correlated moderately with the PSE (r = .28,p < .05).Interestingly, those identified as having a problem-laden life and self-esteem did not correlate significantly (r = -.11,p = .35).

Discussion
The findings of the first application of PSE in South Africa, and, as far as our literature review showed, the first self-esteem measurement of young children in a township context, show that PSE is applicable in the Soweto school setting.We found Cronbach's Alpha to be high and similar to European and North American applications of PSE.Furthermore, we found that a single factor structure was statistically most appropriate and conceptually most interpretable.The results of this study correspond well with criteria reported by previous researchers (Fuchs-Beauchamp, 1996;Haltiwanger, 1989;Lemola, et al., 2010;Verschueren, et al., 1998).The fact that items 2, 10, 11 and 15 could form a second factor could be explained with regard to the cultural background of our study.These four items may measure another facet of self-esteem than intended by the test developer.Especially items 2 (smiling) and 10 (eye contact) could be sensitive to membership in a collectivistic culture such as South Africa, as smiling and making eye contact could have other implications in collectivistic societies than in individualistic societies, where the instrument was developed (Matsumoto, et al., 1998).It may be that smiling and making eye contact are part of a collective norm within a collectivistic society.These items may therefore be inadequate indicators for self-esteem in this subgroup.Alternatively, this finding may also be linked to the young age of the children in our sample.
The correlations between PSE and teachers' subjective summary rating of the children's self-esteem were similar to those reported by Haltiwanger (1989).Interestingly, we found a higher correlation between the mean PSE scores and the question "Who are the children with the lowest self-esteem in this class?"than with "Who are the children with the highest self-esteem in this class?"This result suggests that teachers are more consistent in their evaluation of lower self-esteem than they are with higher levels of self-esteem.The finding may be linked to a general tendency of teachers toward a positive evaluation of children.Especially those children who do not satisfy a particular behavioural norm, particularly in relation to following instructions or being courteous, seem to be evaluated particularly negatively in our study.
Finally, and to our surprise, none of the teachers had difficulties predicting future leadership or lifelong major problems among the 5-year olds.Interestingly, subjective rating of a child becoming a future leader correlated positively with self-esteem, although not very strongly, while major problems in life did not.
As the PSE seems to work well in the South African township context, future research may use this self-esteem measure to investigate its relationship with school performance, educational and job aspirations (of the children themselves as well as their care providers), health status, etc.In many international studies, self-esteem has been shown to be a significant and stable predictor of a multitude of desirable social, psychological, and economic outcomes, and this study has made the first step in preparing investigations in this regard in South African townships.

Table 2 :
Factor Loadings of PSE for Two Factor and Single Factor Solution, Items Sorted Descending for Single Factor Solution