Upload
george-cook
View
221
Download
0
Embed Size (px)
Citation preview
Post Election Vote Auditing
Fritz Scheuren
University of Chicago
Murphy’s Corollary
If you did not check it, then it did go wrong!
Outline of Remarks
• Systems Thinking
• Sample Vote Verification
• Forensic Statistical Additions
• Exit Polls
• Better Together
Systems Thinking• Appreciation of Complexity• No Single System Owner• Political Party Roles• Media Roles• Voters’ Trust and
Participation
Proactive Response Needed
Benchmarking and Sharing What Works
Audited Votesand Voter
Surveys
Ishikawa (Fishbone) Diagram
TrustworthyVoting System
TestedBallot
Trained Pollworker
CertifiedEquipment
SecuredTabulationEducated
Voter
VerifiedIdentity
Sample Vote Verification
• Key to Accountability
• Transparency and Randomness
• Rules of Evidence (Florida?)
• Build A Body of Practice
Forensic Statistical Additions?
• Exploring Official Results for anomalies
• Confirming Outliers and Inliers• Linking Present to Past Patterns• Developing Lessons Learned
Data Bases, Persisting
Ohio Scatterplot of Kerry Difference Between Actual and Predicted Vs. The Total
(Trending 84 - 04)
-60,000
-40,000
-20,000
0
20,000
40,000
60,000
0 200000 400000 600000 800000
Total Vote of Two Parties
Dif
fere
nce Electronic
Punchcard
Scan
Cuyahoga
Franklin
Hamilton
Cuyahoga Scatterplot of Kerry Difference Between Actual and Predicted Vs. the Total
(Grouping Precincts 00 - 04)
-300
-200
-100
0
100
200
300
0 300 600 900 1200
Total Votes of Two Parties
Dif
fere
nce
Exit Polls• Warren Mitofsky• Not a Substitute for Sample
Audits• A Weak Fitness for Use Standard• Badly Misunderstood, Redirect
and Replace
0.00 0.20 0.40 0.60 0.80
Bush Vote Proportion In 2000
0.00
0.20
0.40
0.60
0.80
Bu
sh V
ote
Pro
po
rtio
n In
200
4
Whether the exit poll
overstated Bush or Kerry
Exit Poll Overstated Bush
Exit Poll Overstated Kerry
Bush Vote Proportion In 2000 and 2004 For the Ohio Sample Precincts
More on Refusal Versus Fraud Alternative – 2000 v. 2004
• Are Precincts with Gaps Different?
• Data Does not Support this!
• Actual Results Are Similar not Different
• Scatterplot Shows Rough Similarity
• Distributions Virtually Identical
• Mitofsky “Bias in Refusals” Hypothesis Supported Instead
Still More on Predictive Value of Exit Poll v. Actual Results
• Another Look at Gap over time
• 2004 Exit Poll v. 2004 Actual Gap
• Versus 2000-2004 Change
• Fraud Hypothesis would Predict
• Gap is Correlated to Change
• Correlation only 0.03 However
-0.30 -0.20 -0.10 0.00 0.10 0.20
Difference Between Actual and Exit Poll 2004 Bush Vote Proportion
-0.10
-0.05
0.00
0.05
0.10
0.15
Dif
fere
nce
Bet
wee
n 2
004
and
200
0 B
ush
Vo
te
Pro
po
rtio
nBush Vote Proportion Comparison For The Ohio Sample Precincts
Better Together
• Cooperation Already High Among Election Officials
• Bring in Skilled Outsiders, Statisticians. Computer Specialists, …, As You Have
• Include and Inform Critics• Make Accountability Evident
Media and Marketing
• Approach Media Ahead of Time
• Seize this Timely Moment
• Stress New Tools, Learning Style
• Conduct Demonstration Sample Audits and Get the Word Out
National Election Scorecard
• National Voter (Customer) Survey
• Build on 2006 Ohio Proof of Concept
• Put “Horror Stories” in Perspective
Fully Auditable Election
• Prepare prior data ahead of time, so analysis can be real-time
• Continue to use Exit Polls but adjusting for the bias in them, if possible.
More Examples
• Create and train election officials in new process recording and Sample Vote Verification Standards
• Make sure software is fully tested and as close to tamper proof as possible
Many [email protected]