I've lived in or next to NH for quite a while, and I heard this argument in 2004 too.

Couple of responses:

1. Though many moonbats (and though liberal myself, I use that term here) seem unable to realize, in NH the *rural* areas are often more "liberal", and the cities more conservative, at least when it comes to Democratic primaries. Actually, it's more upscale-intellectual vs downscale-traditional.

Why? Simple, Manchester and Nashua are old-school blue collar, union heavy, strongly Catholic towns. Many of the "rural" towns on the western (next to VT) and central part the state are strongly influenced by the belt of college towns (Keene, Hannover, Plymouth, plus the govt workers in Concord) or have been attractive to affluent people from NY/Boston who have retired, write/paint/invent/etc from a rural home, gave up Wall St to sell maple syrup and connect with nature, whatever. The Merrimack Valley cities have gone for Mondale, Gore, Kerry, Clinton; the bucolic Connecticut Valley towns for Hart, Bradley, Dean, Obama. The small towns are more likely to do hand counted voting.

2. Electons are done on a town-by-town basis, so tampering with machines requires a lot of effort -- there's no centralized point where machines are all programmed. Then again, it sounds like most of the towns contract with the same firm, so I'll grant that makes it a bit easier. But it's still totally easy to do an accurate recount, and someone would have to be damn bold to risk that kind of exposure. Heck, if just one town clerk got curious given these stories and did her own quick hand count -- not hard in a town of a few hundred -- that'd be enough right there.

3. Exit polls depend on turnout estimates. The way they work is they'll do a sample of X people to get the split between candidate A and B, then extrapolate that up to a total of Y people who usually vote at that precient over the course of the full day. It's more complicated than that as they take into account data on turnout by time of day, whether the election overall looks high or low turnout. But fundamentally it is *not* a Z% sample of a _known_ W total, it's a known Z number of samples of an _unknown_ total population.

If for whatever reason, turnout was unusually high relative to the statewide total in the pro-Clinton Manchester wards, that would cause the early raw exit poll numbers to be wrong, and THAT IS WHY EXIT POLLS MUST BE WEIGHTED. You have to get some of the final turnout / vote numbers to calibrate them.

4. There may be a Bradley effect -- whites overreporting their willingness to vote for a black candidate. That effect has been much smaller in recent races so it'd be odd for it to be so large in NH, but at this point it can't be ruled out (and is as plausible as massive vote hacking).