US Election Assistance Commission - Voluntary Voting System Guidelines Vote
EAC Home
Introduction
View Guidelines
View Comments
Glossary

View Comments

Section CommentsGeneral CommentsGlossary Comments
 
Name :   Ron Crane
Organization :   N/A
Post Date :   9/30/2005

Section Comments
Section :  3.4.3
Page no. :  
Line no.:  
Comment :  6a. The reliability standard in Vol. I, §3.4.3 permits the production and use of machines with unacceptably high failure rates.

The Guidelines thus create an opportunity for unscrupulous elections officials to manipulate elections by distributing failure-prone (but still compliant) machines to some precincts, while distributing much more reliable machines to others.

Finally, the Guidelines do not distinguish between failures that impact vote totals and those
that do not. If the former are not much rarer than the latter, elections lose all integrity.

a. Unacceptably high failure rates. Postulating that voting equipment will be operated for 45 hours during a typical election,

Vol. I, §3.4.3 mandates that
[t]he MTBF demonstrated during certification testing shall be at least 163 hours.
That's right: 163 hours, not 1,630 or 16,300 (further, Vol. II, §4.7.3 repeats "163 hours', and also uses the term "MBTF" for "MTBF".)5

Equipment failures usually are assumed to follow an exponential distribution, and the probability that a given system will fail at or before a specific time t is given by the
formula:
t e ė − − 1

where ė = 1/MTBF.6

Of course, this means that any given machine with a 163 hour MTBF has a 24% chance of failing during the "typical" 45 hour elections process.


The failure distribution of a collection of m such machines is usually described by a
Poisson distribution, where the probability of exactly n failures over a given time interval t is given by

(((mėt)^n)* e^mėt) / n!

*Note, ė represents lambda

The mean number of failures in a collection of m such machines is mėt , and the mean failure rate is simply ėt. Assuming failures are equally distributed among all machines, almost 28% of such machines will fail during a typical election. In a collection of 50 such machines, such as a very small jurisdiction might use, the mean number of failures over
the posited 45-hour election period is thus 13.8. This is plainly unacceptable.

By way of reference, typical panel computers, including motherboard, CPU, memory, LCD and touchscreen (such as might constitute a DRE’s core) have MTBFs of
approximately 80,000 hours (= 491 times the Guidelines’ proposed minimum MTBF).
See, e.g., http://www.ucs.co.uk/index.php?pid=1373.