You are getting a sampling bias.
You said "no matter who I randomly surveyed out on the streets". This illustrates where you are going wrong. When you "randomly surveying" people on the streets you aren't getting a representative sample. First of all, not everyone is equally likely to be on the streets, and the streets that you are walking down aren't going to be random either. And of course there are people who never are out on streets, and it is at quite possible that they are less likely to have a college degree.