My hunch is that these surveys would suffer from

1. selection bias (unemployed or low-income graduates would feel embarrassed to respond)

2. the mean of the (biased) sample being much higher than the median (die to outliers)