Self-Selections Bias in Surveys of MOOCs Students

Massive open online courses (MOOCs) allow collection of valuable data on student learning: they represent a huge number of users and open up new opportunities for educational research. MOOCs data consists of platform data and survey data. While platform data generates information about all MOOCs users, survey data has limited generalizability. Despite this fact, web–survey is the most frequently used method of data collection in research for MOOCs. However, the findings of web-surveys are often controversial. One of the possible explanations of this phenomenon is the low quality of survey data due to self–selection bias. Using data from the 26 MOOCs offered by National Research University Higher School of Economics on National Platform Open Education, this study investigates self–selection bias among MOOCs users. There are substantial differences between respondents and non–respondents of MOOCs surveys, which results in non–representativeness of survey data. Those who responded to the survey had higher academic results. Moreover, women tended to participate in the research more often than men did. More educated and older MOOCs users are also responded at a higher rate. The consequences of self–selection can be negative for the validity of research, which is based on survey data. Therefore, it is necessary to reduce this bias, using various correction techniques. This presents the most optimal weighting adjustment, which integrates a students behavior in MOOCs and socio–demographic measures.