Teachers' Predictions and Pupil Achievement in School Certificate

Abstract

This thesis examines the estimates of pupils' performances in School Certificate made by teachers at Hutt Valley High School over a six year period. It aims to assess the degree of accuracy of teacher predictions, in particular at the pass/fail boundary, and to identify and evaluate the relative importance of some correlates of accurate prediction. The data for this study is taken from school records showing for each class the name of the teacher and the estimated and achieved marks of each pupil in School Certificate for the period 1973 to 1978. The statistical analyses include the calculation of the Pearson Product-Moment Correlation Coefficient between the estimated and achieved marks of pupils in English, Mathematics, Science and French. Expectancy tables are provided to give greater detail of the accuracy of teacher's estimates at different points across the mark range. A multiple regression analysis is carried out in order to identify and evaluate the relative importance of some correlates of accurate prediction. The statistical section concludes with a summary analysis of the data from the subjects for the year following the major analysis. The results of the study show that the correlations between the estimated and achieved marks are greater in Mathematics and French than in English and Science, however, the estimated and achieved class means across subjects show greatest agreement in English and least agreement in Mathematics. Whilst variations across subjects and across years occur, in general teachers show a tendency to underestimate pupils at the upper end of the ability range and overestimate the marks of pupils at other points of the scale. A movement of approximately 16.5% across the pass/fail boundary was observed when comparing the estimated and achieved marks of pupils. The study also concludes that the only factor to reveal a substantial and systematic link with the quality of teachers' predictions was the ability level of the class under analysis - teachers were generally more able to predict the results of their pupils for higher ability than lower ability classes (English excepted).