ABSTRACT
With the rapid growth of the internet and digital technology in recent decades, public opinion and electoral research has undergone a transformation in how data are collected. As in-person surveys based on random probability samples of the population have become increasingly expensive, they have become increasingly rare. Traditional survey modes are often linked to particular sampling methods - face-to-face surveys are usually associated with area sampling, and phone surveys are most often conducted by random digit dialing (RDD). Statistical inference is the idea that one can make generalizations about an entire population from a sample that is drawn from it. This is based on the simple premise that everyone in the population has a non-zero and known chance of being included in the sample. The population distribution of political engagement is essentially unknowable and so if political engagement is a non-ignorable conditioning variable this might present a serious problem.
