The break in the series

Every headline labour indicator moved the government's way in the year the survey design changed. What that does and does not let you conclude.

A comparison of India's PLFS labour market indicators across the January 2025 survey redesign, testing whether the design break moved every headline number in a government-flattering direction.

India's Periodic Labour Force Survey documented a reversal of structural transformation: agriculture's share of employment fell to 42.5% in 2018-19, the lowest ever recorded, then climbed to a PLFS-era high of 46.5% in 2020-21 and was still at 46.1% in 2023-24. From January 2025 that survey was retired and replaced. The sampling frame, the stratification, the selection method, the panel structure, the reference year and the questionnaire all changed at once. There was no parallel run.

This panel does one thing: it lines up every headline indicator on either side of that break and asks whether the redesign moved them in a consistent direction. It is not a manipulation detector. Without a parallel run, nothing here can separate a design effect from a real improvement in a year of strong growth. What it can establish is how much of the good news rests on a single discontinuity, and how large the coincidence is.

The reversal is not a straight line. The share spiked during the pandemic years, when reverse migration pushed workers back to villages, then eased in 2021-22 before climbing again. What matters is that it never came back down: every year from 2019-20 onward sits above 45%, and the 2023-24 reading of 46.1% is within half a point of the pandemic peak. Three years after the disruption ended, the workforce had not returned to where it was.

Eighteen of the twenty-eight percentage-point indicators moved in the flattering direction. Five moved against it, all of them small. Five did not move or are ambiguous. The five that went the wrong way are the ones nobody puts in a press release: casual labour's share of employment, female casual labour's share, female youth unemployment, rural participation and the rural worker-population ratio.

This chart is the strongest argument against reading the break as manipulation. Look at rural female participation between 2022 and 2023: LFPR jumped 9.8 points and WPR jumped 9.7, entirely within the old design, with no methodology change to blame. Against a series that routinely moves that much on its own, a 1.8-point fall in agriculture's share is unremarkable. "Large move at the break" is not a reliable tell here.

It also shows the participation numbers barely moved across the break at all. If a redesign had been built to flatter, LFPR and WPR are the headlines you would expect it to lift, and they are flat.

Even taken entirely at face value, the new series does not rescue the structural transformation story. Agriculture at 43.0% in 2025 is still above the 42.5% trough recorded in 2018-19. Seven years on, and after an instrument change that moved the number down, the share of Indians working on farms has not returned to where it was.

The casual male daily wage went from ₹456 to ₹455. That is the only earnings figure in the release that moved backwards in nominal terms, and it belongs to the largest low-end category in the Indian labour market.

The loss of independent sub-samples is the change least discussed and most consequential for verification. Independent sub-samples were the NSS's own internal consistency check: two parallel estimates from the same round that could be compared against each other. That check no longer exists.

Historical precedents cited in the second reading are from public record and should be independently sourced before republication. The female youth unemployment figures are SBI Research's reading of PLFS rather than the MoSPI press note. Percentage-point changes are computed from published rounded figures and inherit their rounding.