This report was compiled by the campaign's interns and volunteers. If you're interested in becoming an intern for the campaign, click here.
First, a quick summary of the overall results for Kevin. According to data released by Placer County, Kevin received 14,886 of the 86,621 votes cast during the election, giving him an overall percentage of 17.19% of the total vote.
Kevin did significantly better in precincts canvassed by volunteers (i.e., precincts in which volunteers contacted voters by knocking on doors). In fact, a majority of votes Kevin received came from canvassed precincts.
Canvassed? 
Yes 
No 
Vote Received 
8,144 
6,742 
Total Votes 
37,713 
48,908 
Percent 
21.59% 
13.79% 
To ensure the significance of the data, we performed a chisquared goodness of fit test. We assumed that if canvassing and number of votes received were independent, then Kevin would receive 17.19% of the votes in both canvassed and uncanvassed precincts. Using the data below, we calculated a chisquared value of 755.5 and a pvalue of less than 0.00001. The difference is clearly significant.
Canvassed? 
Yes 
No 
Votes Received 
8,144 
6,742 
Votes Expected 
6482.86 
8,407.29 
Comparing precinct by precinct data on number of votes and number of interactions with voters, we can make the following scatterplot.
The correlation value is 0.78, which indicates a moderately strong correlation between the two variables. The relationship is obviously affected by many factors. This is especially noticeable in the uncanvassed precincts, where the number of votes ranged from 0 to 277. A certain amount of each precinct’s votes can be attributed to Kevin’s other campaign operationswhich vary precinct by precinctand random chance, as some voters simply pick a random name on the ballot.
Next, we looked at the relationship within the canvassed precincts. The correlation is 0.8739, indicating a strong relationship. (Note that all graphs refer to canvassed precincts as walked precincts.)
Looking at the linear regression line, each additional voter contacted corresponds to an expected gain of 1.523 voters on average. This is probably a result of the networking that results from a successful voter interaction.
We then examined the effectiveness of Kevin’s canvassing versus our volunteers’ canvassing. The data is summarized in the chart.
Votes for Kevin 
Votes Cast 
Kevin’s Percentage 

Kevin Only 
3,372 
16,939 
19.91% 
Volunteer Only 
2,098 
11,849 
17.71% 
Both 
2,674 
8,925 
29.96% 
In the precincts canvassed by Kevin, the number of voters he contacted was strongly correlated (r=0.92).
The correlation (r=0.56) is somewhat weaker in volunteer canvassed precincts than in those canvassed by Kevin. The overall efficiency also favors Kevin himself. Kevin received 2.36 times as many votes as voters he interacted with, while the volunteers’ overall votes to interactions ratio was only 3.623.
We performed another chi squared test on the difference in results between the precincts canvassed by Kevin and the precincts canvassed by volunteers. If the effectiveness of both methods was the same, then we would expect both the campaign to receive 19% of the vote in both Kevin and volunteer canvassed precincts.
Who canvassed: 
Kevin 
Volunteers 
Number of Votes: 
3372 
2098 
Expected Votes: 
3218.41 
2251.31 
The chisquared value was 17.77, giving a pvalue of 0.000025. The difference is again significant.
Looking at the precincts canvassed by both, we see another relatively strong correlation of 0.86. The ratio of total votes to total interactions is 2.21, less than Kevin’s ratio. However, the method garnered high voter percentages perhaps because it combines Kevin’s allure as the candidate himself with the efficiency of the volunteers. The trendline for precincts canvassed by both also had the greatest slope.
Part 2 (Analysis Based on Percentage Share of Vote)
1 What is the correlation (if any) between our canvass operation and Kevin’s precinctbyprecinct vote share?
2 Did it matter whether a precinct was canvassed by volunteers or by Kevin himself?
3 Is the correlation (if any) between our canvass operation and Kevin’s vote share more pronounced if we consider only absentee ballots?
Conclusions:

Moderatestrong positive correlation (R=0.69) between % share of vote in that precinct and number of voters contacted in the precinct. (See Graph 1, Table 1)

Precincts where both Kevin and volunteers canvassed had a much higher share of votes than precincts where only Kevin or only volunteers canvassed. This may be due to the larger number of voters contacted. canvassed precincts had a higher share of votes than uncanvassed precincts, as expected. (See Table 1)

Kevin had a higher % vote share among absentee ballots compared to his overall % vote share. Moderatestrong positive correlation (R=0.678) of % absentee vote share with number of voters contacted on precinct canvasss. (See Graph 2, Table 2)
A Chisquare GOF test was conducted to see if there was a significant difference in the distribution of absentee vs. nonabsentee votes.
No. of Votes 
Observed 
Expected 
Absentee 
11467 
10848.41 
NonAbsentee 
3419 
4037.59 
Total 
14886 
14886 
This resulted in a Pvalue of <0.0001, which means that there was a significantly higher percentage of votes cast for Kevin in absentee ballots vs. votes cast for Kevin on election day.
Table 1: Share of Total Votes
r 
0.690362 
% vote share in volunteer canvassed precincts 
21.5946756 
% vote share in Kevin canvassed precincts 
19.90672413 
% vote share in noncanvassed precincts 
13.785065 
% vote share in combined precincts 
29.9607843 
% vote share (14,886/86,621) 
0.171852 
Table 2: Share of Absentee Votes
r 
0.6782329983 
% vote share among Absentee ballots 
18.1658 
% vote share among nonAbsentee 
14.5563 
Authors:
Ryan L.
Danny Y.
Researchers:
Taisiya T.
Michele V.
Rawan K.
Ryan J.
Sophia W.
Ellen C.
Brandon B.
Strong M.
Charle C.
Christina M.
Tyler T.
Allison W.