SENSE - Student self-concept and school achievement: bidirectional relations and effects of social comparisons and grading

About the project

The purpose of the research project is to strengthen the knowledge of the relations between students´ self-perception of competence and academic achievement, and to investigate social comparisons and grading as determinants of students’ self-concept. Reciprocal relations between self-concept and achievement will be studied as well as mediating effects of academic motivation. Moderating effects of background variables such as gender, age, and cognitive ability will be investigated.

Effects on self-concept of social comparisons in schools with different levels of performance, along with differential effects of grading as a function of student´ characteristics, are other issues in focus. In the project a number of empirical studies will be conducted using large-scale data from the “Utvärdering genom uppföljning” (UGU) national infrastructure which contains data for 10 nationally representative samples of students born between 1948 and 2004. The database contains a large number of variables on individual level from questionnaires and register data as well as aggregated data. The data allow both longitudinal analyses and comparisons between different birth cohorts. The results from the project will add substantially to the research field by using powerful analytical approaches with large-scale data, taking into account both mediating and moderating effects when investigating relations between students´ self-perception of competence and academic achievement.

Research Focus

The project is divided into three subprojects, each with a specific focus. However, the subprojects are linked both by the strong focus on the importance of students´ self-beliefs for achievement and by using data from the longitudinal project Evaluation Through Follow-up (UGU). 

Subproject 1

The construct validity and predictive power of measures of self-concept and self-efficacy for academic achievement: moderation of gender and cognitive ability and mediation of academic motivation.

A first set of studies in Subproject 1 focusses on conceptual and measurement distinctions between self-concept and self-efficacy, while a second set investigates longitudinal, bidirectional, relations between self-perceptions of competence and achievement, taking mediation and moderation effects into account. Self-concept has been shown to predict academic success and different measures of self efficacy have also been shown to have predictive power with respect to later academic success (Bong & Skaalvik, 2003). However, no consensus has been reached in terms of how these constructs predict academic achievement and their relative importance for academic achievement.

It has been suggested that self-concept is a better predictor for academic achievement for younger students who have not yet developed a clear sense of their competencies on specific tasks, while self efficacy is a better predictor for older students who have developed their self-perception of competence in specific tasks due to feed-back from significant others, and experiences of learning (Bong & Skaalvik, 2003). However, when achievement on more narrow and specific tasks is predicted, self-efficacy assessments with respect to similar tasks typically form better predictors than assessments of self-concept. This suggests that measures of self-efficacy to a larger extent than measures of self-concept are influenced by domain- and task specific factors. This hypothesis will be investigated through fitting higher-order confirmatory factor analysis (CFA) models to self-concept items (e.g., “How good do you think you are in Swedish” and “How good do you think you are in Mathematics”) and self-efficacy items (e.g., “How well do you read and understand a text in Swedish” and “How well do you do mental arithmetic”). We expect that the self-concept items will fit a one-factor self-concept model while the self-efficacy items for Swedish, English and mathematics will fit a model with three first-order factors and one general self-efficacy factor.

These two models will then be related to one another to investigate overlap among the factors, and by adding outcome variables in the form of grades and national tests the predictive power of self-concept and self-efficacy can be compared. Grade 6 data is available for the cohorts born in 1992, 1998 and 2004; Grade 9 and Grade 12 data is available for the 1992 and 1998 cohorts; and Grade 9 data will be available for the 2004 cohort in 2020. These data provide a rich material to investigate issues of measurement of self-concept/self-efficacy, such as dimensionality and measurement invariance across cohorts. For all cohorts information is available for more than one grade, which allows for investigation of stability and change of self-concept and self-efficacy over time. Using crosslagged panel analyses with structural equation modeling (SEM) techniques it also is possible to investigate to what extent self-perceptions of competence in earlier grades influence achievement in later grades, and to what extent achievement in earlier grades influence selfperceptions of competence in later grades. Given that the 1992 and 1998 cohorts do not have information about school achievement in grade 6 the investigation of such bidirectional influences between self-concept/self-efficacy and school achievement will be somewhat limited, but the measures of cognitive abilities in grade 6 will be used as proxy measures of early achievement. The longitudinal design can also be extended to include further variables hypothesized to mediate relations between self concept/self-efficacy and achievement. The UGU questionnaires include indicators of intrinsic and extrinsic motivation, which are of great theoretical interest as mediators in this context.

Other potentially interesting mediators are perceived demands of schooling and experiences of stress and worrying. It is also of great interest to investigate gender differences with respect to achievement, self-perceptions, perceived demands and worrying and particularly so if the patterns and sizes of longitudinal relations among variables are different for males and females. A straightforward way to investigate such moderating effects is to split the sample by gender and use a two-group SEM approach. Moderating effects of other student characteristics such as educational and migration background will also be investigated.

Subproject 2

The influence of social comparisons and reference groups on students´ academic motivation and achievement.

In subproject 2, studies are planned of the BFLPE. We will investigate the effect of average school-level achievement for students´ self-concept and self-efficacy, and test for a negative influence, given individual characteristics. Although studies have been done with regard to BFLPE, inconsistences exist in the reported effect sizes and difficulties in finding significant moderators (Fang et al., 2018).

Some results indicate that BFLPE is consistent over ability levels (Marsh & Hau, 2003; Nagengast & Marsh, 2012), while other results suggest that high-achieving students are not affected by the BFLPE to the same extent as low-achieving students (Marsh & Rowe, 1996; Trautwein et al., 2009). Moreover, it has been suggested that the BFLPE is an age-based process and an intercultural phenomenon, which is stronger among high school students, in some countries and when verbal self-concept is considered (Fang et al., 2018). Student cognitive ability, age and gender will be used as moderators in the planned studies. Data for birth cohorts 1972, 1977, 1992, 1998 and 2004 will be used in the studies.

These cohorts are selected for two reasons. The first is that they all include measures of self-concept/ self-efficacy, although different scales were used in the 1972 1977 cohorts and the 1992-2004 cohorts, and that register data for school-level achievement in grade 9 is available in the form of grades. The other reason is that the 1972-1977 cohorts were essentially unaffected by the decentralization and deregulation reforms of the Swedish school system that took place in the early 1990s, while the 1992-2004 cohorts were heavily influenced by the dramatic increase of school segregation that was a consequence of the reforms. Thus, for the early birth cohorts school differences in level of achievement were almost non-existent (ICCs around .03) while for the later cohorts school differences in achievement have increased linearly up to around .14 (National Agency for Education, 2018; Yang Hansen & Gustafsson, 2019).

We therefore expect to see increasing BFLP effects as a function of increasing school differences in level of achievement. Given that several municipalities since some time back use GPA from 9th Grade as a basis for selection of students into popular schools and programs at upper secondary level this generates substantial segregation with respect to level of performance. It may be hypothesized that this generates strong BFLP effects. In the first step of the analyses, measurement models for self-concept/self-efficacy and motivational variables will be constructed, which will be done in close collaboration with Subproject 1. Models will then be constructed where school-level achievement relates to individual students´ self-concept and self-efficacy in the first step and in the next step is related to later academic achievement and educational attainment. Moderating variables such as gender, cognitive ability, age and family educational background will be included in the analyses. Comparisons will also be made between results for the 1972 - 1977 cohorts on the one hand and the 1992 – 2004 cohorts on the other hand, as well as between the 1992, 1998 and 2004 birth cohorts.

Subproject 3

The grading effect in the criterion-referenced grading system

In subproject 3 the starting point is the results from previously conducted studies in the COMP project. In the COMP project a negative grading effect was identified in the norm referenced grading system, low-ability students being negatively affected by being graded in Grade 6 for their subsequent achievement, compared to low-ability students who were not graded (Klapp, Cliffordson & Gustafsson, 2016, Klapp, 2015, Klapp, 2018). The present grading system has a fail-step (F) which makes the failing more severe for students´ possibility to continue within the educational system. Thus, the negative grading effect found in the previous norm-referenced grading system may be stronger for low-ability students in the criterion-referenced grading system. The previous results will be the foundation to investigate the grading effect on students´ subsequent achievement in the criterion referenced grading system, which has not been investigated before.

The birth cohort 1998 did not receive grades until Grade 8, while the birth cohort 2004 received grades and national tests already in Grade 6. These two cohorts thus differ with respect to several characteristics of the grading practice they have experienced. However, information on gender, cognitive ability, family educational background and subsequent grades is available for both cohorts. Even though it is not possible to create a perfect experimental design in retrospect, the 2004 birth cohort may be regarded an experimental group receiving grades and national tests as “treatment”, while the 1998 cohort is a quasicontrol group. Of course, the groups also differ in other respects: the 2004 cohort had national tests in Swedish and mathematics in Grade 3 while the 1998 cohort did not; the 1998 cohort had the Lpo94 curriculum, and the 2004 cohort the Lgr 11 curriculum; and during the six years that separate the two birth cohorts there have been demographic and societal changes.

Some of these differences can be controlled for with register data, while others can only be dealt with in the interpretation of results. Nevertheless, the fact that the two birth cohorts have data from identical questionnaires in Grade 6 and Grade 9 along with performance data towards the end of compulsory school allows for detailed comparisons of the two cohorts, using both descriptive approaches and multivariate modeling techniques with either a one-group or a two-group approach. Background variables such as gender, cognitive ability, and family educational background will be included in the analyses to investigate interaction effects.

Research Questions

The research program will investigate research questions regarding the following topics:

• The construct validity and predictive power of self-concept and self-efficacy for academic achievement, through mediation via motivation and moderation by gender, cognitive ability and family educational level. Reciprocal relations between self-perception of competence and academic achievement will also be investigated;
• The influence of social comparisons and reference groups on students´ self-concept, academic motivation and achievement;
• The consequences of grading for students´ subsequent academic achievement, and the role of mediating relations of self-perceived competence and academic motivation and moderating influences of gender and cognitive ability.
   

Project members

Alli Klapp, Project Leader and Supervisor, University of Gothenburg

Jan-Eric Gustafsson, Fellow Project Leader and Supervisor, University of Gothenburg

Stefan Johansson, Supervisor, University of Gothenburg

David Clarke, PhD student, University of Gothenburg

Thea Klapp, PhD student, University of Gothenburg