The decision of the BJP not to field any Muslim candidate in the recent Uttar Pradesh election has been criticised by some as biased and “communal”, and hailed by others as a courageous rejection of identity politics in favour of merit based selection. We carry out a statistical test to discriminate between these two hypotheses.
The BJP fought the 2017 Uttar Pradesh assembly elections as part of the National Democratic Alliance, putting up candidates in 384 seats, while the Apna Dal (Sonelal) contested 12 seats, and the Suheldev Bharatiya Samaj Party contested the remaining 7 seats. Below we consider the candidate selection process for the BJP alone, rather than the entire NDA coalition.
None of the 384 BJP candidates hailed from the Muslim community. Some critics have held this up as an example of bias and communal polarisation. For instance, journalist Shekhar Gupta has observed:
Unfair to say Modi/BJP hadn't given us a clue. @yogi_adityanath is only a logical next step to politics of not fielding one Muslim in 403
— Shekhar Gupta (@ShekharGupta) March 18, 2017
Other observers have countered this allegation, saying that the idea that only Muslims can represent a Muslim electorate is a discredited relic from an earlier age, and a form of unsavoury identity politics. For instance, economist Rupa Subramanya has written:
That only a Muslim can represent Muslim interests is an extension of religion based voting in colonial times.Worst form of identity politics https://t.co/UN2WgeTpkW
— Rupa Subramanya (@rupasubramanya) March 19, 2017
For our statistical test we take as our null hypothesis that the NDA does not practice identity politics in its selection of candidates. If this hypothesis is true then the probability that a Muslim candidate will be selected to contest any one of the 384 seats is equal to their proportion in the applicant pool who are seeking tickets. This proportion, the percentage of Muslims in the pool of qualified applicants, is a product of three factors: the percentage of Muslims among the general population, the percentage of Muslims that openly support the BJP, and the relative distribution of “merit” among the population. We know that Muslims are about 20% of the population, and we will assume that merit is uniformly distributed among different communities, since there is no data that implies otherwise. The percentage of Muslims who are avowed BJP supporters is unknown, but anecdotal evidence suggests that it is likely to be quite small, probably no more than 10%, or even 5%. Of course, the lower the BJP support among Muslims, the higher the chance that a merit based selection will not turn up any Muslim candidates.
Let us say that the percentage of BJP supporters among Muslims is x. Then, of every 1000 people, 200 are Muslims and 2x are BJP supporters. Since BJP polled 40% of the votes, 400 out of 1000 overall are BJP supporters. Therefore the Muslim proportion among the applicant pool is x/200. The probability under the null hypothesis that in 384 trials no Muslim was chosen equals (1-x/200)^384.
If the percentage of BJP supporters among Muslims is 5%, then the probability that a religion neutral, purely merit based selection ended up choosing no Muslim candidate in 384 attempts is only 0.006%. If Muslims among BJP supporters are as much as 10%, then the probability of religion neutrality is about 0.0000003%! Even if BJP supporters are only 1% of the Muslim population, the probability that the selection was purely merit based turns out to be only about 13%. In order to believe that the selection was purely merit based, we would have to assume that BJP commands only about 0.1% support among Muslims.
These calculations show that either the BJP discriminated against its Muslim supporters in their selection of candidates, or it has so little support among Muslims, about 0.1%, that it was unable to find any viable Muslim candidates. The first conclusion seems more likely.