Population - 50% random selection -- 75%

I am going to pull another population from the census this weekend. Whether I will use it or not I don't know. It is not going to change the low response rate necessarily, so I have to ponder this. If I decide to simply contact everyone, then it is not longer a random selection, and some of the probability theory I was going to rely on will disappears. However, because I have such a low response rate anyway, that probability theory disappears regardless. I mean, probability is still an issue, but the fact that I randomly selected a population isn't going to matter much when I have such a low response rate.

Response bias is a concern. Inferential statistics used in analyzing date from surveys assumes that 100% of the respondents returned the surveys. response rate is one guide to the overall representativeness of the survey participants. Response bias is less when a large number of individuals return the surveys. Response bias is more when few return the survey. Respondents are self-selecting anyway, and that result is pronounced when few return the survey. As Babbie (2004) points out, with a low response rate, there is usually something more going on than unwillingness to take the survey when you get a low response rate.

So far, I have contacted 50% of my entire population -- my census. And so, I have to navigate whether I am going to go to 75%, or whether if I go to 75% I might as well go to 100% and have a survey of the entire census rather than a random selection.

Interestingly, my biggest concern here is that I told the first two groups that they were randomly selected, and so if I go with 100%, that won't be true. So I worry about my researcher image and its credibility.