Most Worldcons in recent times grant voting rights only to supporting and full attending members. (This is required by the WSFS Constitution.) They rarely grant voting rights to single-day attendees. (They may do so if they choose.) I went back and looked up the numbers of at-door attending memberships relative to the total number of WSFS members (those eligible to vote), and got the following:
Year - City - WSFS Members - At Door Attending
2005 - Glasgow - Insufficient Data
2004 - Boston - 5984 - 284 (5%)
2003 - Toronto - No Data
2002 - San Jose - 4856 - 404 (8%)
2001 - Philadelphia - 5347 - 349 (7%)
2000 - Chicago - 5608 - 376 (7%)
Working off the assumption that relatively few people buy advance attending memberships between the Hugo Voting cut-off and the close of pre-registration, these figures represent the number of WSFS members who buy memberships but have no opportunity to vote for the Hugo Award in that year. (They do get nomination rights for the subsequent year, like all other WSFS members from a given Worldcon.)
Going back farther into the past, the at-door-attending percentage is about the same, although it was 10% in San Antonio in 1997 and 16% in The Hague in 1990.
And before you get all excited about the "disenfranchisement" of single-day members: The total at-door membership (full + single-day) for recent Worldcons represents between 10% and 20% of the total membership (plus a spike of 25% in 1993 in San Francisco).
The details behind this are available on the SMOFInfo site, in the file "Details of membership numbers and sizes for a several Worldcons."