Relationships Between Global and Domain Measures of Self-Efficacy, Emotional Cost, and Achievement in Mathematics

• In Event: Identifying, Understanding, and Addressing Elementary Students' Negative STEM Emotions and Beliefs

Abstract

Objectives & Perspective
Control-Value Theory (Pekrun et al., 2007) postulates that control appraisals, such as feelings of competence, predict emotional states, which in turn predict academic performance. These control appraisals may appear as within a given task, a subject, or even a more general sense (as in school). There is evidence to suggest that feelings of competence form a hierarchical structure where domain-level competence, such as self-efficacy and outcomes expectancies, can be fed from more global perceptions (e.g., Bong,1997; Bong & Skaalvik, 2003; Choi, 2005). For example, Marsh and colleagues (2019) found evidence that general and cross domain self-efficacy facilitates beliefs in another domain. Combining these two areas (Control Value and perceived competence), we investigate indirect relationships between global self-efficacy and mathematics expectancy, emotions, and performance. Specifically, we investigate how differing levels of self-efficacy relate to mathematics emotions and performance.

Methods
Participants were those participating in a larger study of Spatial Temporal (ST) Math, a mathematics learning software (see Wendt et al., 2014). The sample included fifth grade students from one school district (N = 5,423). The analytic sample (N = 3,522; 65% of total sample) included those students who played ST Math and had all of our measures, including those collected outside of ST Math (see Table 1). For global self-efficacy, we used a four-item subscale from the CORE SEL survey (Meyer et al., 2018). For mathematics expectancy, we used two items from student surveys administered through ST Math. We measured mathematics emotions via student surveys administered through ST Math. Students selected up to three emotions for how they feel about mathematics. We conceptualize the total number of negative emotions (“bored,” “frustrated,” “nervous”) selected as a measure of emotional cost. We measured mathematics performance via end-of-year assessment scores from the California Department of Education–California Assessment of Student Progress and Performance (CAASPP). We examined direct and indirect effects via path models using R “lavaan,” bootstrapping the analyses over 5,000 samples.

Results
The initial relationship between global self-efficacy and performance was statistically significant (𝛽 = 0.16, p < .001). We then found support for direct effects for the relationship between global self-efficacy and expectancy (𝛽 = 0.41, p < .001), and emotions (𝛽 = -0.09, p < .001). Next, we found support for the indirect relationship between global self-efficacy and mathematics performance combined through both expectancy and emotions (𝛽 = 0.02, p <.001) reducing the initial relationship from .16 to .10. Additionally, we found indirect relationships separately for global self-efficacy to mathematics performance through expectancy (𝛽 = 0.04, p < .001) and emotions (𝛽 = 0.01, p < .001). See figures 1 and 2 for direct and indirect coefficients.

Significance
We found support for direct and indirect relationships among global and domain related efficacy and domain emotional cost and performance. Although there have been advances in comparison of global and domain self-efficacy outcomes with perceived values (e.g., generalized internal/external reference model; Köller et al., 2016), we extend this work by looking at these cross-level efficacy outcomes with academic emotions.  

Authors

• Kerry Duck, University of Delaware
  Presenting Author

• Teomara Rutherford, University of Delaware
  Non-Presenting Author