See here:
http://uscountvotes.org/index.php?option=com_content&task=view&id=70&Itemid=43 Ballots that contained no selection for President are referred to here as "undervotes".
Undervotes may result when a voter intentionally omits to vote for president, but when undervotes are associated with a particular kind of voting technology, it suggests that the machine is either failing to record the voter?s intention, or failing to offer the voter feedback he needs to register his selection on the machine.
We find a strong association between election day undervotes and pushbutton DRE voting machines. The average undervote in precincts where pushbutton DRE voting machines were used was 3.5%, compared to less than 1% in precincts that used scanned paper ballots.
Precincts that used DRE voting machines were significantly more likely to record undervotes than precincts that used Optical scan machines, and of the two types of DRE machines, Push button machines were more likely to record undervotes than Touch screen machines.
We also found that undervotes were more prevalent in counties with large ethnic minority populations. This may indicate that the net loss of votes was for Kerry. Given the small margin in NM, the probability that Kerry lost NM to this problem would seem to be high.