Rectors Forum sets up parallel vote tabulation

Rectors Forum sets up parallel vote tabulation

By M.M.I. Ahyani

BANDUNG (JP): Are you interested in finding out early results
of the polls' ballot count?

Well then go to the website of the Rectors Forum at
http://forumrektor.itb.ac.id. or drop by the Rector Forum's
secretariat in Bandung or in Jakarta.

The location of the site for the public announcement of
details is as yet undecided. Wisma Antara or the Hotel Indonesia
have been proposed as possible forum locations.

If everything goes smoothly, two days after the voting day on
Monday, June 7, the Rectors Forum will be able to present the
estimated result of 70 percent of the entire ballot counting.

As one of the poll watchdogs -- besides KIPP, Unfrel and some
others -- the Rectors Forum will feature its own ballot counting
methods.

According to forum chairman Liliek Hendrajaya, who is also
rector of the Bandung Institute of Technology (ITB), the election
watchdog will estimate the vote with the help of a statistical
method known as Parallel Vote Tabulation (PVT).

"This method will be a complementary method and enables
election monitoring by, at the same time, conducting scientific
tests," he said in Bandung on Thursday.

The PVT is a common statistical sampling method. Its
superiority lies in its scientific analysis, for example with the
help of the cross correlation method.

Apart from quantitative calculations in the form of
compilation and tabulation of data, the forum's PVT method will
also feature a qualitative interpretation of data already
obtained.

Nevertheless, a statistician from ITB's mathematics
department, Ms. Dumaria Tampubolon, who will head the PVT program
for the Rectors Forum, said there were shortcomings in the method
to be used in the upcoming elections.

"As one of the statistical tests, PVT will not be entirely
perfect. However, as a scientific tool, this method will be very
useful to provide a track record and new data to be used in the
analysis of the next general election," she said.

The PVT method has been applied in more than 20 general
elections in countries such as Chile, the Philippines, some
African states, and Cambodia.

"The result obtained from PVT in the Philippines when Corazon
Aquino won the elections was quite significant with ballot
counting being undertaken by Comelec and Namfrel," Dumaria said.

PVT has stimulated the interest of academics joining the
Rectors Forum. It was introduced here by the National Democratic
Institute from the United States, which has extended technical
assistance to poll monitoring agencies.

"It seems that a scientific method like this is quite relevant
to us at the Rectors Forum," Hendrajaya said.

However, a dilemma caused by a limitation in the application
of PVT has arisen.

Data from local polling stations (TPS) for the elections is
meager, while their presence is important to determine the design
of the survey and the sample to be collected.

"Data from the 1997 elections is not relevant any more.
However, as the data we really need is not available, we have to
use various assumptions to be able to draw up a survey and
sampling design," Dumaria said.

Out of a total of 320,000 TPS already decided upon, the
Rectors Forum has agreed to collect random samples from 8,900
TPS.

"There are usually two methods for the design of the survey
and the sample. Under the first method, determine first the
margin of error and then the size of the sample. Under the second
method, it is a reverse process: determine first the size of the
sample and then the margin of error," Dumaria said.

After considering a variety of constraints, such as financing,
human resources, and available equipment, the Rectors Forum opted
for the second method.

"We determine first the size of the sample and then calculate
the margin of error. Besides it is not possible for us to collect
a sample from all TPS. We have found, finally, that the margin of
error is quite small, namely zero point such and such," Dumaria
said.

It is true that the number of TPS used as a sample is small.
However, the Rector Forum's PVT method uses individual voters as
sample units. With the assumption that one TPS will serve some
500 voters, some 400,000 samples will be obtained.

Isolated areas

According to Hendrajaya, the Rectors Forum will try to cover
areas that other monitoring agencies are unable to cover,
especially considering that the random sample for PVT has already
been obtained. Although a TPS is located in a remote place, it
must be visited to ensure that the result of PVT will be
accurate.

"Once a TPS has been included in the sample, it must be
monitored, no matter how difficult it is to get there," he said.

Provinces considered difficult to reach are Maluku, Aceh, East
Timor and some other provinces whose territories are separated by
the sea. However, Hendrajaya remained optimistic, saying the
forum had recruited at least 220,000 volunteers to monitor the
elections.

In addition, at least 270 universities are associated with the
forum. Data from them will be fed into computers and sent to
processing centers in Bandung and Jakarta.

"This data may be sent through facsimile or the Internet.
Currently, approximately 30 universities have been linked up to
the educational Internet managed by ITB with the assistance of a
satellite owned by Japan," said Basuki, who is responsible for
the computer network at the Rectors Forum.

"Universities not associated in the Internet will have to rely
on public Internet networks such as Wasantara Net or other
Internet Service Providers," he said.

To secure the data, 120 facsimiles at Wisma Antara will adopt
a closed user-group system so that other parties cannot
interfere. Data sent through the Internet channel will be
encrypted first to prevent decoding by other parties.

It is on the basis of these preparations that the Rectors
Forum predicts they can release estimated vote results eight days
after the polling date.

Data discrepancies

A problem may arise if the estimate made on the basis of PVT
methods differs markedly from the calculation made by the General
Elections Commission (KPU) or other poll watchdogs undertaking
ballot counting.

Hendrajaya emphasized that the results of the PVT calculations
will only be used only as a means of comparison.

"Certainly PVT-based calculations cannot be claimed or made a
political instrument by the losing parties. To avoid this
possibility, until the issuance of the official result by KPU,
PVT-based calculations will only be in the form of compilation
and tabulation data, along with some explanations but without a
conclusion."

Earlier, executive director of the forum Sudjana Sapi'ie said
the result of the calculations would be the same as those
obtained by KPU only if a computer network was allowed access to
KPU's computer network.

"It is always possible that calculation results may differ
from those issued by KPU. Therefore, an election monitoring
agency must acquire data-processing capability on a par with that
of KPU, so that a check may be run to find out whether the KPU
data is correct or otherwise," he said.

According to Sapi'ie, KPU would standby their result, while
monitoring agencies would have to verify their results.

He acknowledged a problem might arise if verification of KPU
results led to the discovery of discrepancies.

The question is whether an election monitoring agency has to
immediately expose this discrepancy to the public, at the risk of
sparking chaos.

"It is at this juncture that a poll watchdog should assume its
responsibility and act with great caution. We must weigh
carefully whether or not it is for the good of the society if we,
the intellectuals joining the election monitoring agency, divulge
the discrepancy to the public. Who will be responsible if this
exposure leads to social turmoil?

"So, the exposure of discrepancies in data must be weighed
very carefully because this may also enhance the community's
confidence in the credibility of poll watchdogs," Sapi'ie said.

If contradictory data stems from foul play, the situation will
be of great concern. Should this scenario play out, election
monitoring agencies must not stay silent, especially now that
there are 48 parties which must be ranked according to the
percentage of the vote won.