The data is likely limited in scope by more than just income. For instance, its a web-based startup. This means that users of it are more likely to be tech savvy. It also means that the users are younger, which gives the baby boomers a weaker representation.
Mint is also a money management site, so by it's very nature it's users are interested in properly managing their money. People who care about proper money management are probably doing a better job of making/saving/investing it.
My life experience is limited (I'm only 24) but from it I can say with confidence that the majority of people in my community, family, school, etc are not tech savvy, don't worry about managing money, and frankly most still believe that if you can't buy the software at bestbuy or walmart it's not made by a reputable company.
Yes, this struck me as a classic example of how you can't just say "Look, I have a huge sample, so it's statistically valid!" (1% is a huge sample by statistical standards at this scale... or it would be if it were properly random.) The bias in the data is strong and pervasive.
Having such a huge sample means that the data does at least mean something. It does indeed tell us something about the habits of those likely to sign up for that service, after all. It's just that converting that to the universal data we're all really interested in is non-trivial, and if the bias is strong enough, basically impossible if the projection of the noise dominates the signal.
(In other news, this article worked. I'm seriously considering signing up. :) Informative marketing, the best kind.)
Mint is also a money management site, so by it's very nature it's users are interested in properly managing their money. People who care about proper money management are probably doing a better job of making/saving/investing it.
My life experience is limited (I'm only 24) but from it I can say with confidence that the majority of people in my community, family, school, etc are not tech savvy, don't worry about managing money, and frankly most still believe that if you can't buy the software at bestbuy or walmart it's not made by a reputable company.