Interesting. I'd imagine Comcast and BellSouth win because older, less tech-savvy people are more likely to use their ISP account, and credit scores correlate with age. AOL probably doesn't have the same benefit because you can sign up for a free @aol.com address, and kids are a large part of AOL's demographic.
A person's everyday use of a broadband ISP e-mail address is only logical if they have a somewhat permanent consumer relationship with their broadband ISP.
Possible criteria for such security:
1) The user will remain at their current place of residence for an extended (multi-year) period of time. This excludes renters and other non-homeowners. It also excludes people without stable or established careers, as they could be required to relocate.
2) The user will pay their bill on time. (Otherwise, their e-mail address will go away.)
Additionally, having a broadband ISP excludes the rural American demographic.
People fitting in both 1) and 2) will be more likely to have positive credit scores for obvious reasons.
While I agree there's likely a correlation to the two criteria you mention, there's a third which is think is more at the core.
That is that for a ton of ISP subscribers the concept that email doesn't need to be coupled with their internet service is not something they understand. For these people the idea that they'll need to change their email address when they switch ISPs is taken as a given. Just as if they were to move they would need to update their mailing address.
I doubt "kids" signing up for AOL are a large reason for the low score. The credit score is from people who wanted to check their score, and I would guess that the number of people <18 who tried to check it is almost negligible.
On the other hand, AOL, Yahoo, and MSN/hotmail all had low scores, which may have something to do with those emails being tied to an instant messaging account.
All we know is that from a sample of 20k scores those with Yahoo email addresses (which could well be 10k or 2k, we don't know) had a lower average than the rest. It says nothing about the whole population of Yahoo/Hotmail/etc. because we don't know how the email addresses were obtained.
It would be very easy to bias this sample by showing ads about finding your credit score to different demographics within each mail provider.
In addition, they don't show variability within the groups. The differences they find among the averages may not mean much in the context of the variations about the averages.
Furthermore, any statistically significant differences may be due to correlation with other factors. Perhaps domain name varies weakly with geographic location-- which the same source claims also predicts average credit scores.
Finally, their chart accentuates the differences by dropping the zero.
