Econometrics inform us about the consequences of estimating effects (e.g., from regression) on nonrandomly selected samples or groups, namely bias in effects. The bias arises because we cannot, in the effect estimate, separate the “true” effect from the effect created by the nonrandom selection mechanisms leading people into the sample or group. In that sense, sample selection bias is about internal validity (and not generalization as such).
Now, in much qualitative research selection of respondents is (often intentionally) nonrandom, in two aspects. First, the group of interest (i.e., college students) is often a select group of the full population, which means that nonrandom processes lead individuals into the group. Second, the actual individuals interviewed are nonrandomly selected by the qualitative researcher. So, we have a “double” sample selection here.
Does such sample selection issues imply internal validity problems for qualitative research? I am not so much interested in the generalization issue (i.e., whether we can infer to the general population from our population), but on the internal validity issue. Perhaps I am way off the mark, because I use standards from quantitative research to judge qualitative research. Nevertheless, I have thought about it lately and would like other’s take on the problematique!