When India’s Delimitation Commission next convenes, it will inherit a task that no commission has faced in fifty years: the freedom to actually change how many parliamentary seats each state sends to the Lok Sabha. The 84th Constitutional Amendment of 2001 froze that allocation until after the first census taken after 2026. The 2027 Census will lift that freeze, and the redistribution of roughly 281 new seats across 36 states and Union Territories will follow.

The political debate has focused almost entirely on the North-South dimension, on whether states such as Kerala and Tamil Nadu, which succeeded in controlling population growth, will now be punished for that success by a reallocation that rewards higher-growth Hindi-belt states. That concern is legitimate, and it deserves the attention it has received. But it has obscured an equally consequential question that will determine whether any new boundaries actually improve Indian democracy: which constituencies within states should be split, into how many parts, and on what evidence? An EAC-PM working paper provides the most systematic answer yet attempted.

Drawing on a panel of 2,171 constituency-elections across four general elections from 2009 to 2024 and linking each constituency to its demographic and linguistic profile, it overturns one of the oldest assumptions in Indian electoral studies. For decades, conventional wisdom held that large constituencies suppress voter turnout. The logic seemed self-evident: queues at metropolitan polling booths can run for hours, per-capita polling-station access in dense urban areas is lower, and the voter in a constituency of two million feels more remote from power than the voter in one of 500,000. Smaller constituencies, on this reading, should vote more. And they do, in the raw data. The smallest decile of constituencies turned out at a remarkable 22.9 percentage points above the largest decile in 2009. But that gap has been halving steadily. By 2024, it had fallen to 12 percentage points.

More importantly, once you account for the demographic and linguistic composition of constituencies, how urban they are, how many residents belong to Scheduled Castes or Scheduled Tribes, how linguistically polarised or diverse the local population is, the size penalty largely disappears and by 2024 reverses. A constituency with two million electors at a typical demographic profile now turns out approximately six percentage points higher than a comparable one with one million electors. The residual........

© The Statesman