a) you are a hobbyist trying to do statistics
b) you have not considered at all demographics of RP support or lack there of.
c) you have not considered the demographics of each preinct or county.
d) you have cherry picked your results (at least on the short report).
Your primary error is in assuming that RP support is uniform across voting district size. Your secondary error is assuming that the demographics of the voters are also uniform across each district.
Further, in small districts just one additional vote can have a meaninful % effect (ie, 1/20=5%). This effect of course can be magnified if the candidate is likely to have more support in rural/smaller precincts.
Indeed, if you ran your totals in order largest to smallest, RP share would grow not decline. Are votes being 'flipped' in smaller districts to favor RP?
What about counties which RP won? If the effect does not show up there too than you have a serious issue with your theory
In fact, Buchanan is an adequate surrogate for RP as show in NH in 1996 and specifically your Nillsborough graph. Like RP, Buchanan's support was not uniform.
Finally - and the largest hole of them all in your theory. Who is doing the flipping, how and when? Are you accusing poll workers of altering results? Are you saying the poll workers have called in the correct totals but someone at the board of elections have modified them (even though there is evidence to the contrary lower down in the system?) Are you saying that some how all different types of voting equipment (paper ballots, scanned ballots, electronic, manual old style machine) have all been hacked to produce just this same result?
gedankenexperiment.dk views on finance, politics and science
Want DP delivered to your inbox daily? Subscribe here: