1. is an issue with external validity (i.e. does the findings generalize to other groups). But the comparison being made in the study is within a given group with similar socioeconomic status (all Harvard males). Assuming that they only vary in the quality of their personal relationships, any difference in happiness because of this could, assuming it is causal, be interpreted as a generalizable effect of relationship quality.