Recently, the development of computational thinking (CT) skills has become widespread in the whole world. There is a consensus to develop students’ CT skills through unplugged and plugged-in programming. However, the choice of programming approaches taught in schools is arbitrary and few studies have compared the effectiveness differences between unplugged and plugged-in programming. This study aimed to explore the differences between unplugged and plugged-in programming in developing CT skills of students in Grades 1-2 students. It also explored the joint effects of students’ gender, programming experience, programming interest, and programming confidence with different programming approaches on CT skills. The quasi-experimental research approach was used to conduct 10 weeks of unplugged, plugged-in, and regular programming activities with 124 Grades 1-2 students. The results found that unplugged programming promoted CT skills more in the early primary grades than plugged-in programming. We found that boys and students with programming experience in unplugged group performed significantly better in CT skills than in plugged-in group. Moreover, the improvement of CT skills of students in unplugged group was positively correlated with the level of their programming interest and programming confidence. However, these were not observed in plugged-in group. The results of this study can provide precise guidance for teaching CT skills to early primary school students and provide references for CT education.