CHENNAI: The significant difference in the voter turnout data between the provisional figures released by the office of Chief Electoral Officer after the Lok Sabha elections at 7.30pm (72.09%) on Friday and at 12am (69.46%) on Saturday stirred considerable confusion among the public. While minor variations are expected in these figures in every poll, variations in the range of seven to 13.5 percentage points in several constituencies led to concerns over data quality.
Enquiries by TNIE revealed that a method of projection based on data directly collected by the office of the CEO from a sample number of polling stations led to these variations. According to officials, periodic updates provided by office of CEO on Friday and later at 7.30pm were based on this sample- based projection.
In contrast, the Saturday 12am figure was based on actual data from all assembly segments fed by the respective Returning Officers to the ECI’s online portal, Encore. Interestingly, even after 24 hours of polling, the final turnout figures were not released until Saturday night. Releasing the same 12am figures at 7.30pm on Saturday, CEO Satyabrata Sahoo said data entry for each polling station was still under way. “The respective ROs are doing it for all polling stations. We expect very minor variation (in the final figures),” he said.
The statewide turnout remained at 69.46% as on Saturday night. Sahoo, however, added that turnout for Thoothukudi will be sharply revised to 66.88% from 59.96% as there were delays in entering data to the Encore portal on Friday. Except for this constituency, only very little changes are likely in other seats, he added.
The revision in Thoothukudi will marginally increase the statewide turnout data when the final figures are released. The variations reported earlier due to sample-based projections were sharp in the three seats in Chennai district, and Sriperumbudur, Coimbatore, Sivaganga, and Madurai (refer infographic).
It can be noted that while these variations were noted in the data released by the office of CEO, the periodic data released throughout Friday by Returning Officers of constituencies in their respective districts ended up being closer to the figures released at 12am since their data was based on figures aggregated from all polling stations and not just samples.
Sources said that CEO’s office was releasing figures based on data reported to them directly by the Presiding Officers (POs) of a sample of around 13,000 polling stations from the total of 68,321 polling stations across the state.
This reporting mechanism was working parallelly with the standard reporting mechanism of POs of each polling station reporting hourly data based on the register of voters (Form 17A) to their zonal officers, who in turn report to assistant returning officers (AROs). The AROs report the numbers to ROs, who feed them to the Encore portal.
While district officials acknowledged that there is room for variations in the system as these numbers are orally relayed, the margin of error could not have as large as happened on Friday due to sample-based projections.
“I was alerted by political parties at 5.30pm that there were two sets of data (CEO and ROs) that were being released and both were at odds with each other. By 7pm, we had started to receive calls from reporters and others asking us how we managed to sharply increase the turnout. We knew that there had been a mistake,” said an election official.
“When a sample projection is made, there could be a huge difference in the actual turnout of voters because different polling stations in the same constituency don’t always register similar numbers,” said another official. For instance, in Chennai North seat, Jutkapuram (Periamet) recorded nearly 71% turnout while Elephant Gate had merely around 40%.
“When an extrapolation is done, these factors should all be taken into account. However, there was no need for an extrapolation since returning officers were promptly recording the data in the Encore portal. We are not certain as to why the projection was done when we have the infrastructure for compiling the actual numbers,” said another election official.
When contacted, the CEO acknowledged that a sample projection was done but did not divulge more information on why it was done or the number and criteria of the polling stations that were chosen as samples.
(With inputs by T Muruganandham)