-
A problem with "X votes with Y Z% of the time" metrics is that not all votes are recorded. A lot of representatives vote differently on voice/division votes than they do on roll call votes, and that's by design. People are strategic about whether they call for roll calls or not.
-
Also, besides the roll call vs. otherwise distinction, there's also the distinction of where the votes are made. I haven't checked to be sure, but my feeling is that most of the people creating these figures are only counting floor votes, and ignoring committee votes.
-
Unlike floor votes, committee votes (at least in NH) *are* all done by roll call, and at least from what I've seen anecdotally, representatives are much more likely to split with other members of their party in committee than they are on the floor.
-
This is because by the time bills come to the floor, the party caucus has had time to come to a consensus about what position to take on the bill, whereas while bills are still in committee, that consensus is still being formed.
-
Then, at least in NH, besides floor votes and committee votes, there are also county delegation votes, which I'm betting aren't included in the figure either, since they're tracked completely separately.
-
And then there's also internal party caucus leadership votes, which I don't think get included either.
-
Also, going back to floor and committee votes, another factor to consider there is consent calendar votes vs. regular calendar votes. I bet the percentages would look extremely different based on whether the consent calendar is included or not.