I've never understood this argument. When performing scientific studies, there is a sample size of n = x hundred/thousand, and we then generalise the result across the entire population. Having 48% of the population participate in this "study" is likely to be very indicative of the likely voting choice for the remainder of the population, right? You really think that the proportion of votes for each party for those people that haven't voted would be any significant difference from those that did?

You're assuming the population is homogeneous

> Having 48% of the population participate in this "study" is likely to be very indicative of the likely voting choice for the remainder of the population, right?

That isn't how statistics work. Sample size reduces your error relative to the population you are randomly sampling from.

When you don't have a random sampling, then you sampling method is what determines how generalizable your findings are. A good sample size with a bad sampling method tells us little to nothing about the general population and only informs us about the specific sub population for which the sample can be considered a random selection from.

With significant differences in voting rates across many different demographics, votes are absolutely not a representative sample of the overall population.