Again, I'm not so sure I buy into this article because in my experience, those who are rating applicants have no idea how to weight IT skills. Because of this simple fact, I tend to believe that the study is BS.
I couldn't have said it better myself. If you look closely at the article it grades all of the certifications together and quotes "pay premiums" as increasing 3 times faster.
I could think of a quite a few ways to skew the findings.

1. lump all certs together, say you have 20 certs 18 rise 20% and 2 fall 150%

2. certs rise 2% and non-certs rise an amazingly three times faster 6%

3. non-certs <- what does that mean (kinda generic) allowing you to make it whatever you feel like
little or no parameters

4. 6 months is a generally small time frame for percentages rising and falling