In response to a question Do "confirm Email" Boxes In Sign-up Forms Work? on a marketing resources website, Steve Alker at UniMax Solutions writes as follows (my emboldening):
I’ve never come across any published data to say that it actually
improves the ratio of people who fill in a form in over the percentage
of people who decide to move on without completing the form.
There is however internal data from HP and Dell and informal feedback
from CRMGuru, IT Business Edge and Technology Evaluation Centres which
indicates that for sign-up pages which are a requirement for
newsletters, white papers or other user demanded free information,
that it does not diminish the sign-up rate to any appreciable extent.
You must understand that for at least the latter 3, the data is not
really scientific, because it is before and after changes to the
validation procedure on their websites were introduced, not as a
result of a single or double blind market test with a control group,
but it looks pretty convincing – you don’t lose people.
As a means of preventing unintentional incorrect email addresses being
entered, the improvements cited has been as high as 41% and as low as
27% but at either end of the range, that is impressive.
I'm inclined to think that the percentages above refer to the improvement in the number of bad addresses, rather than the actual percentage of bad addresses. I agree there is not enough information in the quote and that this should not be used for making evidence based decisions.
The only good evidence on which to make decisions would be the OP's own A/B testing on signups with and without the confirm email address field.