I’d wager closer to the former than the latter. Political geography is pretty key here — 1 to 2 points in key places in a competitive race can produce results. But that’s derived from the general.
OTOH: In a caucus or primary where voters won’t be relying on party labels as cues — maybe factors like ground game are more important.
Seth Masket, John Sides and Lynn Vavreck did a study of this and found an effect for Obama but not Romney field offices in 2012.
