The main purpose of this study is to determine the effect of software education on elementary students' computational thinking. In this study, the meta-analysis method was employed which examined 24 studies conducted by year 2020 and have met the inclusion criteria. The types of software, group age, group size, and number of treatments were determined as moderating variables. An analysis based on a random effect model showed that the mean effect size value was calculated as 0.606 with an error of 0.08. It was also determined that there was no publication bias in the meta-analysis. According to the results of the moderator analysis, it was found that the effect of software on computational thinking of the students did not change according to the types of software education and group age, but it changed according to the group size and the number of software educational treatments. The effect size on computational thinking was high in small group(less than 6 students) and large group(more than 31 students who encouraged peer study). In addition, the number of treatments was 21 or more, the effect size was the highest.