In the Lok Sabha elections held in 2019, Andhra Pradesh went to polls in the first phase on April 11.

Provisional voter turnout data for the first phase, compiled on April 12 and released by the Election Commission of India on April 13, showed that 3,07,43,680 or 3.07 crore votes had been cast in Andhra Pradesh. This came to 78.14% of the total registered voters eligible to vote. The disclosure even cited the gender-wise breakup of the voters who had cast their votes.

Eleven days later, on April 24, 2019, the Election Commission of India released revised voter turnout data for the first phase of polling, both in absolute numbers and in percentages. For Andhra Pradesh, this came to 3,13,50,159 (3.13 crore) votes cast, a turnout of 79.88% of the eligible electorate.

On April 29, 2019, it released even more granular data showing the absolute number of votes cast in each constituency that had voted in the first phase, including in Andhra Pradesh.

Compare this with the Election Commission’s conduct in the 2024 general elections. Andhra Pradesh voted in the fourth phase on May 13 this year. That night, the Election Commission released data showing the approximate voter turnout as of 11.45 pm in each of the states where votes had been cast that day. For Andhra Pradesh, this was 76.5%. The absolute numbers for the votes cast were missing.

Four days later, when the Election Commission released a new set of turnout figures, it adjusted Andhra Pradesh’s turnout to 80.66%. No absolute numbers were given.

The missing data is fuelling concerns. A citizen watchdog group, Association for Democratic Reforms, filed a petition in the Supreme Court seeking directions to the Election Commission to release detailed data on the absolute number of votes cast in every constituency. While the Election Commission of India may not have done so, six states have already put out this data, within 48 hours of polling, as Scroll found out.

On Friday, the court adjourned the matter, refusing to pass an interim order. During the hearing, the Election Commission’s lawyer argued that the gap between the provisional and final turnouts will be just 1%-2%.

This is incorrect. In Andhra Pradesh, a comparison of provisional and revised voter turnouts released by the Election Commission shows that the difference between the two metrics in Andhra Pradesh is 4.16%. Significantly, this is more than double the gap between provisional and final data released in 2019.

For states for which comparable official data from both 2019 and 2024 is available – Telangana, and four constituencies in Odisha – the same trend can be seen.

A widening gap

In 2019, the revision of turnout numbers in Telangana had led to a drop of 2,800 votes per constituency on average. In the four constituencies of Odisha, the votes cast per constituency went up by 900 on average.

The 2024 election has seen a much higher jump in numbers between the provisional and the final turnout declarations.

Unlike 2019, the absolute number of votes polled is not available for 2014. We derived an approximate number based on the turnout percentage and the total number of eligible voters.

This shows the provisional votes increased by approximately 3.1 lakh in Telangana, and by 1.1 lakh in four seats in Odisha – Kalahandi, Nabarangpur, Berhampur and Koraput. That comes to about 18,000 votes per constituency in Telangana and 27,000 votes per constituency in Odisha.

It is in Andhra Pradesh that gaps grow bigger. This year, the voter turnout increased by 4.16% after revision. Deriving an approximate number based on the turnout percentage shows 17.2 lakh votes were added between provisional and final counts – nearly 69,000 votes in every constituency in the state.

In 2019, the increase in voter turnout in Andhra Pradesh was a third of this – 6.06 lakh votes or about 24,000 votes per constituency.

However, in 2019, the Election Commission had an explanation for the increase. When it released Andhra Pradesh’s provisional voter turnout, the poll body noted that the numbers were only from 172 of the 175 Assembly seats.

“Data for 3 ACs [assembly constituencies] shall be available tomorrow,” it added in the remarks section.

But in the 2024 general elections, explanations from the Election Commission have been missing.

Did the Election Commission’s provisional figures for Andhra Pradesh this year also exclude data from some Assembly seats? If yes, then how many? We do not know. Scroll called and messaged the chief electoral officer of the Election Commission in Andhra Pradesh asking for these details. The story will be updated if he responds.

So far, the Election Commission of India has not released any nation-wide voter turnout data in absolute numbers. “The tradition has been to release the absolute number of electors and the absolute number of voters, with percentages of male and female voters and the overall percentage of voters,” said former chief election commissioner Om Prakash Rawat. “The Election Commission has a set format to collect this data.”

On May 23, former chief election commissioner SY Quraishi told India Today that he was surprised by the increase in revised voter turnouts and held the Election Commission responsible for the resulting confusion. “Why did they not release it [the voter turnout] that evening or the next morning?” he asked. “In our time, there used to be a press conference at the end of the day, which has been discontinued for what good reason I don’t know.”

Former election commissioner Ashok Lavasa told India Today that increases in voter turnout may also have been recorded in previous elections. “The point is that the more open you are about your information and the explanation that you can provide for any inconsistency or discrepancy is what will satisfy the people,” he said.

Rawat, too, said that the Election Commission should be forthright on these matters. “If a state has not submitted their data, they should say so,” he said. “Say that the data from these many states is still tentative and say why. Or that data from certain states in certain phases is missing. The ECI does not have to hide anything.”

The delay was needlessly stoking doubts about the integrity of the press, Quraishi told News24, even as he asserted that the Election Commission’s system of securing the data was rigorous. “Once the number of final votes in an EVM machine has been recorded in what is called the Form 17C, it is signed off by the presiding officer, copies given to booth agents of the candidates, the machine is sealed and kept under high security. It is impossible to change that data.”

Five states

Though the Election Commission of India in Delhi has not released the voter turnout in absolute numbers, chief electoral officers in at least five states have released the same figures on their websites.

The CEOs in Karnataka, Tamil Nadu, Goa, and Rajasthan uploaded the data on their websites within 48 hours of polling. The CEO in Madhya Pradesh, released the absolute numbers only for the first phase, relegating the turnout in the next three phases to percentages.

The turnout data released by these five states include constituency-wise number of electors and the number of votes polled, with a gender-wise breakup of both these metrics. The percentage of voter turnout has also been mentioned for every constituency.

Tamil Nadu and parts of Rajasthan and Madhya Pradesh, which voted on April 19, put out their voter turnout in absolute numbers on April 20 and April 21.

The voter turnout percentage in these disclosures by state units of the Election Commission matches the revised turnout which the national poll panel released on April 30. This means that the Election Commission had the absolute numbers for these states, but it did not publish them.

Rawat told Scroll that it is likely that while few states can collate turnout data accurately and on time, most do not. Or, that the data submitted by several states has problems. “The ECI giving out incomplete information like percentages is itself indicative that primary officers who are responsible for coming up with [the voter turnout] data have not zeroed in on the final figures,” he said. “So to cover that up, they give whatever is manageable.”

Part of the fear, Rawat added, might be the scrutiny that the poll body might face if it puts out erroneous data.

“It becomes very embarrassing for ECI to have such information with glaring gaps and incongruities,” said the former chief election commissioner.

In 2019, the Quint had reported that there was mismatch between votes polled and votes counted in at least 373 constituencies.

The revised voter turnout released by Rajasthan.

‘EC needs quality manpower’

The significant increase in the provisional and revised voter turnout has continued till May 20, the fifth phase of the elections.

Odisha is an example. According to the Election Commission, the turnout in five parliamentary constituencies in the state – Bargarh, Sundargarh, Bolangir, Kandhamal and Aska – till 11.30 pm on May 20 was 67.59%.

On May 23, when the poll body revised the turnout, the figure climbed to 73.5%.

The five seats in Odisha have a total 79.7 lakh electors, according to Election Commission data.

When we derive an approximate number based on the turnout percentage and the total registered electors, it appears that the Election Commission accounted for 4.7 lakh more votes in Odisha between May 20 and May 23. That’s an increase of nearly 94,000 votes per constituency.

The Election Commission press release said that this might increase further since the data from two polling stations in Kandhamal constituency was yet to be added.

Similarly, in West Bengal, the increase in turnout in seven constituencies – Bongaon, Barrackpore, Howrah, Uluberia, Serampore, Hooghly and Arambagh – is nearly 68,000 votes per constituency.

The seven constituencies in the state have 1.25 crore electors, according to Election Commission data.

Is the difference in provisional and revised voter turnouts significant? Rawat told Scroll that he does not believe that there is any manipulation in the voting process. “There is no hanky-panky as is being alleged,” he said. “Nothing can be done to change any numbers. There are so many checks and balances that anything can be caught very easily.”

Instead, the former chief election commissioner believes that the Election Commission has been affected by an overall deterioration in state capacity.

“There is little recruitment in Class 3 and Class 4 levels of the government because of a financial crunch,” he explained. “We might have trained IAS officers. But at the level of clerks and inspectors, there are hardly any experienced people in the right places in state governments.”

Rawat added that he has seen this erosion first hand. “When I was at the ECI [in 2018], one state took 15 days to give us the data on voter turnout. And in those 15 days, this state kept sharing data that had errors,” he said.

The Election Commission of India is represented at the polling station by a presiding officer. When voting finishes at a polling station, the presiding officer fills Form 17C, which has details like the number of electors registered at the polling station and the number of votes recorded in the electronic voting machine and the gender-wise break-up of the voters who cast their ballot.

For Rawat, the problems in data occur at this level. “Many form 17Cs have empty spaces, wrong information and sometimes the poll officials forget to delete the data from the mock poll,” said Rawat. A mock poll is carried out ahead of the polling to test EVMs for errors, in the presence of all candidates or their representatives.

The former chief election commissioner added: “To remove those errors, a verification process is conducted at the level of the returning officer using documents submitted by the presiding officers. Every piece of data is cross-checked and confirmed.”

Getting accurate voter turnout usually takes three to four days in a constituency. But with poor capacity, this process can drag on. “There is a big variation in the quality of manpower available with the chief electoral officers and the returning officers,” said Rawat. “The Election Commission needs quality manpower.”