J. ChiU Psychol. Psychiai. Vol. 33. No. 7, pp. 1255-1261, 1992 Printed in Great Britain.

0021-9630/92 J5.00 + 0.00 Pergamon Press Ltd © 1992 Association for Child Psychology and Psychiatry

Parent-Teacher Agreement on Kindergarteners' Behavior Problems: a Research Note Claude Gagnon, Frank Vitaro and Richard E. Tremblay Abstract—The present study investigated parameters of agreement as well of disagreement among mothers' and teachers' ratings of the behavior of kindergarten boys and girls. Findings indicated a generally low to moderate level of agreement between these two types of informants. They also suggested that demographic and background characteristics can effect size and direction of level of agreement. Although proportions of children rated as having the listed behavior problems varied in magnitude according to mothers and teachers, the rank order ofthe proportion of children with those behavior problems was rather similar for both informants. Such similar rankings of prevalence rates may reflect parents' and teachers' agreement on the consistency of children's behavior across the home and school settings. Keywords: Parent-teacher agreement, child behavior problems, assessment, kindergarten, child psychopathology

In a meta-analysis of 119 studies concerning children with behavioral and emotional disorders, Achenbach, McConaughy and Howell (1987) reported a mean correlation of .27 between parent and teacher assessments. Instead of concluding that one or both sources of information are unreliable or invalid, the authors suggested that disagreements between informants were as instructive as agreements since they could highlight variations in judgments of the child's functioning across situations and interaction partners. Parents and teachers indeed maintain different relationships with a child and observe her/him under different conditions. Given this viewpoint, discrepancies between informants can be expected. Factors which effect agreement between informants thus become as important to investigate as level of agreement perse. For example, Verhulst and Akkerhuis (1989) found that the size of correlations between parents' and teachers' ratings of children's behavioral/emotional problems, although generally low to moderate, varied according to age of the child (lower for

Accepted manuscript received 28 October 1991

Research Unit on Children's Psycho-Social Maladjustment, University of Montreal, Canada. Requests for reprints to: Claude Gagnon, Research Unit on Children's Psycho-Social Maladjustment, School of Psycho-Education, University of Montreal, 750, Gouin E., Montreal, Quebec, Canada H2C 1A6.

1255

1256

C. Gagnon et at.

younger children) and type of problem assessed (higher for externalizing than for internalizing problems). Verhulst and Akkerhuis (1989) as well as Achenbach et al. (1987) found that the child's gender did not influence parent-teacher agreement whereas Touliatos and Lindholm (1981) reported higher inter-rater agreement for boys than for girls. However, to our knowledge, the influence of factors such as socioeconomic and family status on parent-teacher agreement has not been assessed although prevalence rates of children's behavior problems have been shown to vary according to such factors (Williams, Anderson, Mcgee & Silva, 1990; Offord, Boyle & Racine, 1991). A first objective of this research report is to further examine the influence of demographic and background characteristics of informants on their level of agreement. A second objective pertains to a comparison across informants of prevalence rates and prevalence rankings of childhood disruptive behaviors. Loeber, Green, Lahey and Stouthamer-Loeber (1991) have recendy shown that although there were signiflcant differences in the level of disruptive child behaviors reported by parents and by teachers in a sample of clinic-referred prepubertal boys, the ranking of prevalence of specific child behavior problems was strikingly similar. It remains to be determined whether such findings can be replicated using community samples of younger children of both genders, and rating scales instead of structured interviews.

Method Subjects

In kindergarten, 1924 children (1024 boys, 900 girls) were rated by their teacher and their mother using a behavior checklist. All ratings took place in the spring, following eight months of teacher contact with the children. The mean age for the entire sample was 6.0 years (SD = .31). This sample originated from one large metropolis, one medium size city and one small size city in the province of Quebec, Canada. The response rate from teachers and parents was 81 and 76%, respectively. Measures

A French version ofthe Preschool Behavior Questionnaire (Behar & Stringfield, 1974) was completed by teachers and mothers. This questionnaire (PBQ) has been widely used for more than a decade with preschool through second grade children as a means of evaluating behavior problems (Eriekson, Sroufe & Egeland, 1985; Kinard & Reinherz, 1984; Rubin, Moller & Emptage, 1986; Tremblay, DesmaraisGervais, Gagnon & Charlebois, 1987). Acceptance of the questionnaire stems largely from data demonstrating its strong inter-rater and test-retest reliability (Behar, 1977), as well as its external vEilidity as an instrument for identifying preschool and early school age children who are experiencing social problems (Hoge, Meginbir, Khan & Weatherall, 1985; Rubin et al., 1986). Following results of factor analyses from a number of different samples (Fowler & Park, 1979; Hoge et al., 1985; Moller & Rubin, 1988; Tremblay et al., 1987), a two-principal component solution has been shown to be stable for the teacher version ofthe PBQ across gender, ages, different socioeconomic populations and cultures. The first factor consists of disruptive and externalizing (i.e., aggressive-disruptive) behaviors (13 items; see Table 2) while the second factor is made up of "anxious" or internalizing (i.e., anxiety-social withdrawal) items (worried, solitary, distressed, fearful, cries, stares into space). As for the mother's version ofthe PBQ, factor analysis involving a different sample of 768 first and second grade subjects yielded the same factor structure as the teacher version (Coutu, Vitaro & Pelletier, 1989). Test-retest reliability checks for the mother's questionnaire over a 1-month interval yielded rs greater than or equal to .65 for both scales and both genders (average r = .76). Thus the same two scales were used for both informants. The mothers were also asked questions regarding

Parent-teacher agreement

1257

family status (one- or two-parent faimily), number of children in the family (one, more than one), mother's level of education (high school not completed; high school completed without higher education; some higher education) and father's occupational prestige (mother's when no information was available for the father). Data on parents' occupation were quantified according to Blishen and McRoberts (1976) socioeconomic index, which ranges from 14.4 (hunting and trapping) to 75.3 (administrators). To facilitate data analyses, this scale was divided into three levels of occupational prestige: (1) 60 and over; (2) 30-59; (3) less than 30.

Results Variations in level of agreement between informants

Degrees of agreement between informants have been assessed through separate Pearson's product-moment correlations for different demographic and background characteristics. Although statistically significant {p< .001), inter-rater correlations were low to moderate: .26 for girls and .39 for boys, for externalizing behavior; .21 for girls and .22 for boys for internalizing behavior. The difference between rs for boys (.39) and for girls (.26) on the externalizing scale was statistically significant {z = 3.39, p < .001), although there was no difference for the internalizing scale. The rs on the externalizing scale also varied as a function of fathers' occupational prestige, mothers' level of education, family status and number of children in the family. Correlation coefficients between mothers and teachers on the externsdizing scale for boys were .29 for the lowest level of occupational prestige and .44 for the highest; the difference between the two was statistically significant {z = 2.0,p< .05). Motherteacher correlation for girls' externalizing behavior ratings was higher (z = 2.01, p < .05) when mothers had completed high school (r= .32) than when they had not (r = .18); the correlation coefficient was .21 when mothers had some higher education, but the difference with the highest group was not statistically significant. The level of agreement between informants was higher (r= .48) for externalizing behavior for boys when there was only one child in the family than when there was more than one {r = .36; z = 2.17, p< .05); the equivalent correlation coefficients for girls were .34 and .23, respectively, but the difference was not statistically significant. Differences across informants in level of ratings

In order to assess differences between parents' and teachers' ratings, a 2 (Informants) x 2 (Gender) x 3 (Occupational Prestige) MANOVA was performed on the externalizing and the internalizing scores from the teacher and the mother rating scales. This allowed for an examination ofthe hypothesis that low agreement in the correlational analyses was, in part, a function of the differences in the level of rating of externalizing and internalizing problems by each informant. Gender and Occupational Prestige were added to the MANOVA to test whether the differences between informants were modulated by the children's gender or the parents' occupational prestige. The multivariate results indicated an overall significant difference between informants, Hotellings i^(2, 1914) = 547.96, p < .001. Subsequent univariate i^ tests revealed that the difference between mothers' and teachers' ratings was evident for the externalizing and the internalizing scales, F {1, 1915) = 822.73 and 577.81, respectively, p< .001. No interaction effects were found. As shown in

1258

C. Gagnon «( ai.

Table 1, mothers attributed mean scores about twice as high on the externalizing scales and one and a half times as high on the internalizing scale compared to the teachers. This result was obtained across genders and occupational prestige categories.

Table 1. Means and SD (in brackets) of externalizing and internalizing scores according to informants, levels of occupational prestige and gender Scales Occupational prestige

Externalizing Boys Girls («=1024) (n = 900)

Internalizing Boys Girls (n=1024) (n = 900)

Teachers

Level 1 Level 2 Level 3

3.90 (4.48) 5.47(6.07) 5.19 (5.91)

2.15 (3.34) 2.17(3.66) 2.37 (3.48)

2.43 (2.27) 2.41(2.57) 2.72 (2.64)

2.03 (2.26) 2.12(2.23) 2.32 (2.42)

Mothers

Level 1 Level 2 Level 3

7.13(3.76) 7.89(4.18) 8.16(4.48)

5.99(3.88) 6.27(3.77) 7.02(4.02)

3.86(2.14) 3.99(2.31) 4.07(2.36)

3.67(2.05) 3.84(2.15) 4.18(2.37)

Informants

Ranking of prevalence rates

The next series of analyses were aimed at verifying if proportions of children exhibiting behavior problems according to each informant, although different in magnitude, were similar in rank ordering. For this purpose, responses 1 and 2 ("applies sometimes" and "frequently applies") from each informant were collapsed and proportions of children with 1 or 2 scores on each item were computed. These computed proportions were subsequently rank ordered according to informant and gender of the child. Table 2 shows these proportions (i.e. percentages) and rank orderings. One can observe a substantial correspondance in the ranking of proportions of children exhibiting externalizing behaviors according to mothers and teachers, for boys and for girls {Rhos oi .11 and .81, respectively, p< .01). Although differences in prevalence rates were significant, prevalence rankings were very similar. However, two rank differences are notable. For boys, there is a rank difference of 5 for the item "irritable, quick to fly off the handle"; the difference in prevalence rates between informants is also quite large (72.8% according to mothers vs 27.3% according to teachers). For girls, a similar large difference can be seen for the item "disobedient" (rank difference of 4; a prevalence rate of 74.5% according to mothers and 19.4% according to teachers). Ranking similarities in prevalence rates are less striking for proportions of children with internalizing behaviors. Rhos were lower (.43 for boys and .26 for girls), and not statistically significant. Given the small number of possible comparisons (n = 6), the rank order correlations were affected by the rank inversion of one behavioral item: mothers reported high prevalence rates of crying behavior for boys and girls (second and first ranks, respectively) whereas teachers rated this behavior in fifth place. Rankings for the proportions of children on the internalizing behaviors were quite similar.

1259

Parent-teacher agreement

Table 2. Prevalence rankings of externalizing behaviors according to teachers' and mothers' ratings^ Mothers' Teachers' ratings Rank diff.* ratings

Rank

Gender

Rank

Behavior

Boys

1 2 3 4 5 6 7 8 9 10 11 12 13 Spearman's

Restless Squirmy Blames Disobedient Fights Does not share Tells lies Irritable Bullies Inconsiderate Kicks, hits Not liked Destroys Rho = .77, p< .001

50.8 47.3 43.2 41.8 41.2 34.0 28.9 27.3 24.8 21.4 19.8 18.4 18.2

0 -4 -1 +2 -2 -2 +2 +5 -3 - 1 +2 - 1 +3

81.6 57.2 70.5 78.5 49.0 45.4 69.3 72.8 19.4 21.5 38.2 IB.4 36.0

1 6 4 2 7 8 5 3 12 11 9 13 10

Girls

1 2 3 4 5 6 7 8 9 10 11 12 13 Spearman's

Blames Squirmy Restless Does not share Disobedient Tells lies Irritable Inconsiderate Fights Not liked Bullies Kicks, hits Destroys Rho = . 8 1 , p

Parent-teacher agreement on kindergarteners' behavior problems: a research note.

The present study investigated parameters of agreement as well of disagreement among mothers' and teachers' ratings of the behavior of kindergarten bo...
456KB Sizes 0 Downloads 0 Views