A structural equation model of principals communication patterns, funds management and school-community relationship

Abstract

Recent studies tend to explain the importance of communication in the organisation as well as prescribing the most commonly practised techniques adopted by school managers. Studies on financial management are quite limited with the available ones suggesting that poor financial management is a source of conflict between school leaders and host communities. Little seems to be known on the connection between principals' communication patterns and funds' management as predictors of school-community relationship. This study builds on existing studies and appears to be the first to assess the linkages between principals' communication patterns, fund management practices and schoolcommunity relationship based on empirical rather than subjective data. A structural modelling approach was adopted to examine the nexus using quantitative primary data obtained from a random sample of 2108 respondents. A questionnaire which was designed and validated by the researchers served as the data collection device. Collected data were subjected to Exploratory and Confirmatory Factor Analyses, as well as Multiple Regression Analysis. Findings revealed various significant communication, funds management and school-community relationship practices that are available for adoption. However, it was found that the extent to which principals utilised such practices were below expected minimum standards. It was also found that there were no significant partial and composite effects of principals' communication and funds management practices on school-community relationship (SCR). Based on these results, relevant theoretical, policy and practical implications are discussed.

Contributions of placement, retraining and motivation to teachers' job commitment: structural equation modelling of the linkages

“…Data were initially assessed using exploratory, descriptive analysis, including box plots, skewness, kurtosis (to check for outliers and normality) and inter-item correlation (to check for non-clustered items). These were done following instrument validation research instructions (e.g., Bassey et al., 2020 ; Field, 2005 ; Owan et al., 2021 ). The preliminary screening test found no non-clustered items across all instruments since they were associated with each other in the matrix.…”

“…Bartlett's test of sphericity yielded a significant result, χ 2 (153) = 9418.37, p < .05. "This means that there is no duplication among the variables and that the observed correlation matrix deviates from the identity matrix, indicating that it is acceptable for data reduction using ML estimation" ( Owan et al., 2021 , p.9). The three factors were preserved when examining the pattern matrix for loadings of the relevant items on their latent components.…”

“…CFA was done based on the Maximum Likelihood (ML) estimation approach. According to Owan et al. (2021) , the CFA "determines the accuracy with which variables measure their respective factors (constructs); assesses the acceptability or otherwise of hypothesised models based on theoretical models; and verifies the existence of relationships established by the EFA between observed variables (indicators) and their supposed constructs" (p. 11).…”

“…The Chi-square index measures the degree of difference between a sample and its fitted covariance matrix; it is commonly considered very sensitive to sample size ( Cornell Statistical Consulting Unit, CSCU, 2021 ). The presumption is that the Chi-Squared value of the model must be insignificant for it to be accepted ( Owan et al., 2021 ). The GFI is used to quantify "the proportion of variance explained by estimated population covariance," with cutoff values of 0.95 ( CSCU, 2021 , p.1).…”

“…The GFI is used to quantify "the proportion of variance explained by estimated population covariance," with cutoff values of 0.95 ( CSCU, 2021 , p.1). The NFI, RFI, IFI, CFI, and TLI measure how an estimated model improves the fit compared to the theoretical or null model ( Owan et al., 2021 ). These statistics have a range of values between .00 and 1.00, with values closer to 1.00 suggesting a better match ( Hooper et al., 2008 ).…”