No evidence of disproportionate Muslim deletions in Bihar SIR

An analysis of electors deleted during the Special Intensive Revision (SIR) in Bihar this year suggests that is no clear evidence that Muslims were disproportionately impacted. Data show that 18.4% of the names deleted in the exercise were Muslims, close to their population share of 17.7% in the State as recorded in the 2022 Bihar Caste Census . The chart below shows the share of Muslims among elector deletions and among the population.

The conclusion is based on an analysis of elector names in the deleted lists by The Hindu’s data team. To identify Muslim electors, we implemented character-based machine learning models developed by Rachana and Sugat Chaturvedi in their 2023 paper, “It’s All in the Name: A Character-Based Approach to Infer Religion”. The results of the machine learning model were randomly checked manually to ensure accuracy.

About 52.8 lakh of the 65 lakh names deleted were considered for this analysis. In 10 constituencies, the ECI had released the data only as image PDFs, which could not be accurately scraped, while in a few others, technical reasons prevented the analysis.

Of the 52.8 lakh names analysed, the model was able to identify with very high probability that about 9.7 lakh (18.4%) were Muslims. Moreover, a reason-wise breakdown of Muslims and non-Muslims deleted from the Bihar electoral roll showed no significant variation. For instance, 44% of deleted Muslim electors were removed because they had “permanently shifted,” a share that was nearly identical among non-Muslims too. A similar analysis of deletions under other categories — such as “deceased,” “untraceable,” and “duplicate enrolments” — also showed that the shares among Muslim and non-Muslim voters were largely similar. The chart below shows the reason-wise share of Muslims and non-Muslims.

A gender-wise breakdown of Muslims and non-Muslims deleted from the rolls also showed no notable variation. Of all deleted Muslim electors, 56% were female, compared with 55% among non-Muslims. The chart below shows the gender-wise share of deletions among Muslims and non-Muslims.

A booth-wise analysis of deletions showed that in about 490 of the more than 82,400 booths for which data were analysed, all the deleted electors were Muslims. However, a random check of 10 such booths showed that almost all electors in these booths were Muslims to begin with — meaning that all deletions being of Muslims was not an anomaly.

Among districts, nearly 65% of deleted electors were Muslims in Kishanganj — the highest share across the State. Notably, the 2011 Census recorded Muslims as 68% of Kishanganj’s population, a figure closely aligned with the deletion share. The share of Muslims in deleted electors was higher than 30% in Purnia, Araria and Katihar districts. And these are the next three districts for Muslim population share in State in the Census. Taken together, the district-wise analysis shows that Muslim deletions were broadly in line with their population share. The data indicates that the share of Muslims deleted is not disproportionate to their population size — no matter how the data are sliced.

In Kochadhaman (Kishanganj district) and Amour (Purnia district), more than 70% of electors deleted were Muslims. In the Assembly constituencies of Bahadurganj, Jokihat, Baisi, and Balrampur, the share was between 60% and 70%. In Kishanganj, Thakurganj, Pranpur, and Araria, over 50% to 60% of deleted electors were Muslims. The map below shows the constituency-wise share of Muslims among those deleted in the SIR.

The analysis also highlighted booths with a high concentration of Muslims and mapped their distribution across Assembly constituencies. In the Araria seat, all deletions in 25 of the 394 booths were Muslims— showing that in these booths, most electors were Muslims.

Source: Election Commission of India, Bihar Caste Census

areena.arora@thehindu.co.in

vignesh.r@thehindu.co.in

srinivasan.vr@thehindu.co.in