The lack of a timely Census has muddied the SIR debate | Number Theory

.

The majority of deletions in electoral rolls under the ongoing Special Intensive Revision (SIR) exercise can be explained by de-duplication of multiple entries, this newspaper argued on Thursday. The argument was based on a strong correlation between districts by their share in state-wise growth in electors and share in deletions as seen in the draft roll released immediately after the enumeration phase. To be sure, as that analysis pointed out, this circumstantial evidence needs to be backed by more data to provide conclusive proof that this is indeed the case. While some of the data can be released by the Election Commission of India (ECI) and respective Chief Electoral Officers (CEO) if they wanted to, there is a bigger data hole which cannot be plugged even if the agencies wanted to. This is on account of the inordinate delay in conducting a census after 2011 ; it was delayed from 2021 onwards on the excuse of the pandemic, even as equally intensive exercises, such as elections, were conducted on schedule. Here are three charts which explain the argument in detail.

An official collects data for Census. (HT Photo)

The lack of a timely Census has muddied the SIR debate

• 
  There is a large PC-wise variation in growth in electors between 2009 and 2024 in India

  The number of registered voters in India increased from 717.9 million in 2009 – it is the earliest period for which we have comparable parliamentary constituency (PC) boundaries – to 980.8 million in 2024, an increase of 36.7%. As is to be expected, there is a large state-wise variation in this trend with Uttar Pradesh leading the ranking and Delhi at the bottom among major states and union territories which have at least seven PCs. What is even more interesting, however, is the variation at the PC-level where the 10th and 90th percentile value differ by a multiple of 2.5. Even a cursory look at some of the really high voter growth PCs suggests that there are two clear drivers of voter growth in this period. There are PCs such as Faridabad, Gurgaon, Sriprerumbudur and Gautam Buddha Nagar (Noida), that have seen a large economic growth and perhaps in-migration during this period. But then there are also PCs such as Maldah Uttar and Dakshin in West Bengal which are more likely to have high population growth than migration for economic reasons.
• 
  Most migrant importing states show a larger variation in voter growth than in TFR

  Had the 2021 census been conducted on time, we would have known the exact number of voting age people in every district of India and could have generated a reasonably accurate mapping of these numbers with the number of electors. The 15-year long gap between census population figures means that this cannot be done and we are in the dark as far as the net impact of migration and births and deaths on the number of voters is concerned. One way to approximate the dynamics of this flow can be to compare the variation in TFR (total fertility rate) in the 2015-16 National Family and Health Survey (NFHS) -- the earliest NFHS round after the 2011 census -- across districts and elector growth across PCs between 2009 and 2024. States where migration played a bigger role in driving up voter counts should see a bigger variation in voter growth than TFR numbers. Migrant importing states such as Gujarat, Telgangana, Maharashtra, Kerala, and Delhi indeed show much higher values of coefficient of variation -- it is the standard devaition divided by the mean of a distribution -- for PC-wise voter growth than respective values for district-wise TFR. To be sure, a state need not be a inter-state migrant importing one for this to hold true. Even states with high intra-state migration can show higher variation in voter growth than in TFR.
• 
  And electors are less likely to overshoot voting age population in migrant importing districts

  Let us take the state of Uttar Pradesh as an example. A district-wise matching of voting age population from the 2011 census and voters in the 2012 assembly election shows that the count of electors overshot the voting age population by a smaller degree in the intuitively migrant importing districts (not just inter- but also intra-state). The opposite was true for migrant exporting districts. Once again, this confirms the argument we made yesterday that even though voting age adults were recorded as living in some other place in the census, they were still registered as voters in their native places. All of these claims would be significantly stronger if we had census data of a more recent vintage. Hopefully, we will be able to revisit some of these arguments when the census data becomes avialable in 2027.