The literature on social capital and migration has not given adequate attention to two-way migration. This paper shows that by using the concepts of incomplete information and risk aversion to model social capital, we can get both two-way and return migration. The basic idea behind the model is that as networks grow, risk-averse individuals have better information about foreign wages. Better information mitigates risk, which makes individuals more likely to migrate. Risk is modeled by adjusting the distribution and confidence individuals have in their guesses about foreign wages. In a numerical analysis the model exhibits both two-way and return migration in every period.