Computer-based assessment of 7–12-year-old students’ civic knowledge and the connection between students’ performanceand test behaviour

Abstract

In line with international trends, civic competency is also measured using computer-based assessments. The aim of this paper is to investigate the computer-based performance of 7–12-year-old students. The research questions were the following: (1) how do students perform on the civic knowledge test; (2) are there any gender differences; (3) are there any connections between test results and the features of the testing procedures (e.g. time spent on test and returning to previous items); (4) do difficulty of items and students’ ability level correlate to test-taking time; and/or (5) do they correlate to background variables (e.g. perceived social position in the class and satisfaction withschool achievement)?The computer-based assessment was carried out using the eDia online platform in May 2014. A total of 926 students participated from Years 1 to 6. During data analysis, methods were applied from both classical and modern test theory. Students’ performance varied between 44.64 and 69.54 percentage points in the years under examination. Returning to previous items was not restricted during the assessment. A comparison of the answers students gave first and those they finally kept (after returning to the task) demonstrated significantly higher scores in every case. The difference is the highest in Year 1 (4.70 percentage points); at the same time, the difference decreased in higher years. Three-quarters of the students spent between 500 and 1500 seconds (i.e. between 8 min. 20 sec. and 25 min.) on the test. These students were labelled as a subgroup, and two further subgroups were formed: students spending less than 500 seconds on the test and those who spent more than 1500 seconds on it. Children spending the least time on the test are usually those in Years 3 to 6; in their case, weak and moderately significant connections were found to background variables. A significant correlation (r=0.49, p<0.01) was also identified between time spent on test and skill level. In the case of students spending the most time on the test, no relationship was ascertained with any of the background variables.