Estimating the impact of peer group effects on socioeconomic outcomes: Does the distribution of peer group characteristics matter?