With just a week to go for the elections, a sample survey of Delhi’s electoral rolls has raised the possibility that 20 per cent of registered voters should have actually had their names deleted.

Extrapolating the numbers from their sample to the whole State, the researchers say that 23 lakh names needed to be deleted from the rolls.

The Bangalore–based urban advocacy organisation, Janaagraha, conducted a survey of 2,847 respondents through the surveying firm TNS Global in October. They selected seven Assembly constituencies (AC) that reflected the urban-rural mix through semi-purposive stratified random sampling.

They then selected 34 polling booths in each constituency and ten respondents from the rolls at the level of each polling booth. The findings are significant at the State-level with a 95% confidence level, Ebony Bertorelli, manager of Janaagraha’s Applied Research Program, told The Hindu.

Janaagraha found that 20 per cent of respondents should not have been on the rolls. The main reason for the names to be deleted was because they had shifted (ten per cent) or could not be found (eight per cent). The errors were highest in Burari and Najafgargh (24 per cent), followed by Uttam Nagar (20 per cent), Deoli, Shahadra (19 per cent), Jangpura (10 per cent) and Ambedkar Nagar (8 per cent).

“The Election Commission needs to update its methodology. It continue to operate as it would in rural areas, while there is high intra-city migration in urban areas,” Maj. Gen. (Retd.) K.R. Prasad, coordinator of the organisation’s Janaa Roots programme, told The Hindu.

If the margin of error had hypothetically been the same in 2008, it would have meant that 20 lakh voters’ names should have been deleted, while the entire margin of victory in the polls was 7 lakh, Santosh More, programme manager of the organisation’s Jaagte Raho campaign, told The Hindu. If these names had been deleted, the actual voter turnout would have been 71 per cent in 2008 as against the 58 per cent that it officially was, he said.

However, the organisation did not calculate exclusion errors, or the number of people left out of the electoral rolls, which should have been included.

“We found that neither socio-economic status, gender, religion or status of home ownership had an impact on the likelihood of there being an error in the voter’s name,” Ms. Bertorelli said.

However, Scheduled Caste voters were twice as likely to have errors in their names as voters from other backward classes or forward castes, she added.

Delhi’s Special Chief Electoral Officer Shurbir Singh said the Commission was constantly updating its rolls.

“Many people make representations to us about extra names. We cannot just delete them; we have to follow the full process. We have made a huge effort to keep the voter roll pure,” Mr. Singh told The Hindu. He added that he would take a look at Janaagraha’s findings and methodology.

While the numbers may come from a limited sample within one city-state, they are significant because India does not yet have a robust tradition of independent organisations trying to confirm government numbers by carrying out their own sample surveys. But given the scale at which the government operates, many in government are sceptical of small sample surveys.

“If someone goes back and checks the Census numbers in one village they will claim that so many people can’t be found. The authority that the Government of India commands gets a level of public response that cannot be replicated,” said a senior EC official, requesting anonymity. NGOs did not have the manpower the GoI had at its disposal, nor did they put in as much effort, the official said. Ms. Bertorelli said that Janaagraha had made three attempts to locate each person, and canvassed the local community for its response.