Hand-Count Accuracy vs. Machine-Count Accuracy
Much of Europe, Canada, Israel, and many other nations have returned to hand-counted elections. Please give some careful thought to the question of why so many nations would return to hand-counts if they were as problematic as Americans have been led to believe.
Dis-information regarding hand-counts is prevalent, and being spread by many. But have you looked up their sources? Have you researched the studies? All things considered, hand-counts are more accurate than machine counts, they cost less, they provided America with results on election night for about 200 years, and they promote trust and transparency since they can be publicly-observed and verified by anyone.
For the truth about hand-count accuracy, see the M.I.T. studies that detractors often mis-characterize, which are referenced in Appendix A below. It’s important to know that machine-counts have error rates too. Have you compared error rates for machines vs. hand-counts? The difference is miniscule: around 0.0005.
The real difference is this: Due to potential software errors and digital manipulation, the SCALE for possible error in a machine-count is drastically greater than the scale for possible error in a hand-count. Every documented study of a ‘read-and-mark’ hand-count we’ve found has been 99%-100% accurate: we can’t find even one that was off by more than 1%. …But we HAVE found plenty of examples of machine-counts that have been off by 20% or more. The 2022 Primaries in Torrance Co. NM or DeKalb Co. GA come to mind, along with many others: see Appendix B for a list with some links.
APPENDIX A: HAND-COUNT ACCURACY STUDIES
First, we’ll mention the 2018 Wisconsin study (M.I.T.’s Stewart, et al), linked at bottom: The most recent race they studied was the 2016 Presidential race, where they found that hand-counted races that were recounted were off by 0.0018, while machine-counted races that were recounted were off by 0.0013 (the combination being 0.0017). As you can see, BOTH had errors, and the difference was very slight (0.0005).
This 2018 Wisconsin study, along with their 2004 New Hampshire studies (1946-2002) are often erroneously quoted to discredit hand-counts, but they fail to mention the following:
(a) These studies also expose known error rates for machine-counts, which no one ever mentions;
(b) These studies fail to address the much larger scale of machine-count error that is possible due to software errors or digital manipulation (see Appendix B);
(c) There are widely varying hand-count methods, and these studies fail to make distinctions about which hand count methods are the most accurate;
(d) These were not laboratory studies under controlled conditions, and objective true ballot counts were not ever known.
Second, the 2012 Univ. of California Berkeley study (Goggin, et al) is linked at bottom. It does not address items (a) and (b) above, but at least they controlled for item (d), and somewhat for (c). For candidate totals, they found their “read and-mark” hand-counting method to be more accurate than their “sort-and-stack” method, with a low hand-count error rate of less than half of one percent: 0.0048.
Third, we have evidence of far more accurate hand-counts which have been done much more recently across America. The study above from Univ of CA Berkeley is the least flattering, and presents the highest error rate we’ve seen at 0.0048: others are lower. (For example, you’re probably already familiar with Maricopa Co. AZ’s infamous 2021 recount: of the 2.1 million ballots they were ‘presented’, the hand re-count was nearly spot-on. Lesser-known examples include Osage Co. MO’s 2023 hand-count, where the only discrepancies were due to different adjudication opinions.)
The point is, neither hand-counts nor machine-counts are perfect, but either of them can be highly accurate in the right conditions. The “right conditions” can be controlled in a hand-count, but are far less controllable in a machine-count: see Appendix B for recent examples of machine-count disasters in over 70 counties.
--2012 Election Law Journal: Post-Election Auditing: Effects of Procedure and Ballot Type on Manual Counting Accuracy, Efficiency, and Auditor Satisfaction and Confidence | Election Law Journal: Rules, Politics, and Policy (liebertpub.com)
--2018 Election Law Journal: Learning from Recounts | Election Law Journal: Rules, Politics, and Policy (liebertpub.com)
APPENDIX B: DOCUMENTED PROBLEMS WITH MACHINE-COUNTS
Below are a few examples of counties having documented problems with their election machines.
First, know that it’s challenging to find cases for several reasons:
(1) very few counties have agreed to do hand-counts, so our sample size is miniscule;
(2) post-election audits in other states are just as useless as they are in Utah, with no ability to determine if the correct winner was chosen (and obviously, it’s impossible to find any count discrepancies if your procedures do nothing to check for them);
(3) counties get very little media coverage if discrepancies ARE found, so examples are hard to find.
But given all these limiting factors, below are a few of the recent problems we’re aware of. This includes over 70 counties, but even one would be problematic, since we're vulnerable either way:
--Torrance County New Mexico: a full hand-count audit revealed differences exceeding 20% amongst 2022 Republican primary candidates (and even higher amongst precincts). Strangely, the Democrat primary races were very accurate. Start around minute 0:41 of the audit presentation: County Manager Report On Torrance County Primary Election, New Mexico, October 20, 2022 (rumble.com) Or read a quick summary: https://estancia.news/hundreds-of-ballots-went missing-or-someone-altered-the-election-returns/
--Dekalb County, Georgia: hand-count revealed machine named the wrong winner in a 2022 Democrat Primary for a county commission race
--Monmouth County, New Jersey: hand-count revealed machine named wrong winner in a 2022 school board race https://www.theblaze.com/news/voting-machine-error-ocean-township
-- Cherokee County, Kansas: hand-count revealed machine named wrong winner in a 2022 Primary for a county commission race
https://www.deepcapture.com/2022/10/response-to-robert-reich-i-challenge/
--Williamson County Tennessee along with Alaska (and Iowa): EAC determined an “inconclusive root cause” after investigating errors in scanner source-code for Oct. 2021 municipal elections
https://www.eac.gov/news/2022/04/01/eac-issues-report-tennessee-voting-system-anomaly
--64 of the 66 counties that were checked in Georgia (including Chatham, Floyd, Gwinnett Counties and 61 more in Georgia) had the same problem as Tennessee (above) for their 2020 elections
https://kanekoa.substack.com/p/bombshell-dominion-error-code-uncovered
--Mesa County, Colorado massive missing data problems in 2020 general election which were repeated in 2021 municipal election:
--Antrim County Michigan multiple problems in 2020 election
All Expert Reports - DePerno Law Office, PLLC
--Fulton County Pennsylvania multiple problems in 2020 election
For a PDF of the information above, click here Accuracy Studies
(Note that the links are active above, but not in the PDF)