The Walker recall: A correlation analysis of voting machines
Sept. 25, 2012
The purpose of this Walker recall voting machine analysis is to determine the effect of paper ballots, touch screens (DRE) and optical scanners on county and municipal vote shares.
Note that this analysis is not as complete as it should be. There is no breakdown of votes in locations where there were several types of voting machines. Only the voting machine percentages are available. The analysis will be updated when and if votes in each location by machine type are released.
We need the data in the same form as used in this analysis of Winnnebago County vote counts in which probabilities of vote share differentials in the same unit/ward were calculated. Theoretically differences in the shares should have been minimal, say within 5%. But there were much larger discrepancies in a number of locations.
Using the municipal voting machine mix, there was a negative (-.24) correlation between Barrett’s county vote shares and corresponding percentage of total votes cast on DREs. Overall Barrett did better on paper ballots (.11) and optical scanners(.14). As the percentage of votes cast on DREs increased, so did Walker’s share.
The source of the data is the Wisconsin Government Accounting Board Form 190- Voting by Type of Equipment. I created this spreadsheet for the correlation analysis.
Of the 59 counties Walker won, 54 used touchscreens (DREs). But the majority of votes were cast on optical scanners.
In the 13 counties Barrett won, just five had DREs. These were the percentages of DRE votes: Iowa (76%), Eau Claire (21%), Kenosha (12%), Columbia (0.2%) and Milwaukee (0.5%). The total number of DREs was negligible in the counties.
Several correlations were calculated. The first set was to determine if there was a relationship between the municipal vote shares and the percentage of DRE votes cast in each municipality.
The correlation between votes cast on optical scanners and county vote size was 0.45. The larger counties used optical scanners almost exclusively. The correlations were -0.41 for DREs and -0.31 for paper ballots. DREs and paper ballots were mostly used in smaller counties.
In addition, correlation ratios measured the strength of the relationship between voting machines and county vote shares. Voters were encouraged to use DRE’s rather than paper ballots.
In the counties Walker won, Barrett’s vote shares were positively correlated to the percentage of paper ballots (.20) and to votes cast on DREs (0.17). His shares were negatively correlated to optical scanners (-0.21).
In the top ten Walker counties (highest vote shares), 85% of votes were cast on optical scanners, 10.7% on DREs. In the top ten Barrett counties, 96% of votes were cast on optical scanners, 1.2% on DREs.
In counties won by Walker, 76% of votes were cast on scanners, 18% on DREs.
In counties won by Barrett, 95% of votes were cast on scanners, 2.7% on DREs.