Page 166 - General Election

device. By design the EVMs are safe and tamper-proof. However, to build trust and

confidence, various testing/checking/observation and mock poll protocols have
been introduced across various stages of EVM deployment. With the addition of
VVPATs to the BU & CUs the voter can instantly verify if his vote has been cast as

intended. All processes of EVM handling are in full view and participation of
stakeholders.

10. At the request of the Election Commission, a committee of eminent experts
in statistics provided a report describing how many EVMs should be cross-
checked and why. The report recommends the cross-checking of only 479
EVMs across the country, independent of how many total EVMs are used

(some reports mention that a total of 10.35 lakh EVMs were used in GE
2019). It says that, if a fraction of 2% or more of the EVMs are faulty, cross-
checking 479 chosen at random across the country will be sufficient to

detect this fact with near certainty. Therefore a) the faulty EVMin a random
sample will be detected only if the number of such EVMs in a parliamentary
election is more than 20,000; and the presumption that the entire quantity
of EVMs used in the country can be considered as the population is

incorrect. Can this be elaborated and explained?

The following two points are raised in the question above:

(i) The margin of error in the report of the Committee was put as 2%. Hence a
random sample of size 479 will detect a mismatch between EVM and VVPAT count

with ‘virtual certainty’ only if the number of such mismatches in EVMs in a
parliamentary election is more than 20,000.

(ii) The premise of using the whole country as the population is ‘profoundly

mistaken.’

Regarding the first point-- A general perception is that a small sample will not be

able to make any valid inference about a large population. There have been several
suggestions that ECI should sample a certain minimum percentage of EVM systems
and verify the electronic counts with the slip counts for ensuring that there are no
mismatches between EVM and VVPAT counts. The suggested percentage of EVMs

to be sampled varies from 10% to 50%.

This of course is completely erroneous as any of the suggested figures of 10% or

30% or 50% has no statistical basis whatsoever and the numbers have no sanctity.
The accuracy of the results obtained by a sampling procedure mainly depends on
the ‘absolute sample size’ and not on ‘the sample size as a percentage of the

population size.’ While non-intuitive this is the standard statistical result and is well
validated in all standard books on Sampling Theory.

164