There is some debate as to how exact defensive statistics are, but there is little question that they separate the good from the bad, especially with a largesamplesize.
First, the change of 3% (from 45% to 48%) is not a very significant change unless the samplesize of the survey, which is unclear, was sufficiently large.