First look at the plot of total number of votes for a party president versus the number of voters registered to that party.

Here the red and blue are Republican and Democrat, respectively. The Solid symbols are for counties that Electroincally Vote and the Open Symbols are for counties that vote on paper. Because the range of precinct sizes is large both axes are show on a logarithmic scale to make detail visible.

The Green and Purple lines are eye-guides. The 45 degree purple line is what you would get if all members of a party voted for their own party only. The green line is same line with 70% turnout. Points that lie above the purple line arise from votes crossing party lines and from independent voters. When a point lies below the purple line it suggests that votes defecting significantly from the party.

It is fairly apparent that in the smaller precincts (lower left part of plot) the voting pattern curves away from the line indicating either a strong defection from democrat to republican votes.

All the e-voting precints are large (upper right corner). So we can zoom in on this region here:

Now we can compare the e-voting behavior with the optical scan voting. First note that wih e-voting both red and blue cluster tightly along a straight line and the expected vote by registration parallels the observed vote. The optical scans show much greater deviation from this tight line, but the bulk of the points follow the same line. Notably however there is more large defection from this line that hurts the democratic vote than the republican vote.

Now we look at these same effects in another style of plot. In the following we convert from raw votes to percentages. The bottom axis is the percentage of registered voters declaring a party (dem or rep). The left axis is the percentage of votes for a given party. For the Optical scan votes (open symbols) I have scaled their size by the size of the precinct.

Again both effects mentioned before are visible. The large precincts for both e-voting and optical scan lie on similar trend lines. The smaller precincts show better than expected (from party registration) for the republicans and the opposite for the democrats. The smaller the precinct the more this effect.

--charlie strauss