Skip to article content

The month rural India got a 12.7 per cent raise

Between June and July 2025 the average daily wage across all rural occupations rose from ₹454 to ₹511. In one month. Reported year-on-year growth, which had run at a stable 5 to 6 per cent for years, stepped up to 17 per cent and stayed there; by March 2026 the average stood at ₹522 a day.

Some occupations moved further than others. Coastal and deep-sea fishing wages rose 62 per cent in a single month. Beedi-making rose 29 per cent. No drought, no scheme, no migration shock and no harvest failure occurred in July 2025 that would explain any of this.

Set the two periods side by side and the discontinuity is the whole story. In the ten months to June 2025, wages grew 6.23 per cent year on year overall, 6.79 per cent in agriculture and 5.75 per cent outside it. By March 2026 the same three series read 18.12, 20.16 and 16.40 per cent.

What actually changed

Not the wages. The villages.

The Labour Bureau revised the base year of its Consumer Price Index for Agricultural and Rural Labourers — CPI-AL and CPI-RL — from 1986-87 to 2019 = 100, and adopted that new series’ sampling frame for its wage collection at the same time. The revision is real and officially documented. Its scale is the point:

Every one of those changes is defensible on its own terms. A 1986-87 base and a 1983 consumption basket were four decades stale; the geometric mean damps price volatility; COICOP-2018 classification is the international standard. This was an overdue modernisation, not a manipulation.

The problem is what it did to the composition of the sample, and the fact that the Bureau published no explanation and issued no press release when the wage series broke. Ten months on, the discontinuity still is not flagged in the publication itself.

Why a bigger sample made wages look bigger

The expansion from 20 states to 34 brought several north-eastern states and union territories into the count for the first time, and split four large states for sampling purposes — Bihar from Jharkhand, Uttar Pradesh from Uttarakhand, Madhya Pradesh from Chhattisgarh, Andhra Pradesh from Telangana. The combined weight of the north-eastern states, the National Capital Region and Goa rose by roughly 11 percentage points, with Maharashtra down 2.6, West Bengal down 1.9, and Madhya Pradesh-Chhattisgarh and Odisha down 1.7 each.

Now the arithmetic. The newly added regions have smaller agricultural workforces, more concentrated skills, and average wages roughly 50 to 55 per cent above the states in the original sample. Those states held 1.15 per cent of India’s rural population and 1.22 per cent of its workforce at the 2011 Census. They now supply about 11 per cent of sample villages — close to ten times their demographic weight.

A national average built that way rises whether or not anyone is paid more. That is the entire mechanism.

The trend underneath

Splice the two series to strip the break out and a different picture appears. Momentum before July 2025 implied annualised growth of about 5.5 to 6.8 per cent. Under the new series, month-on-month growth has actually decelerated to 0.28 per cent, and to 0.18 per cent in February and March 2026 — an annualised pace near 2.2 per cent. Adjusted for the break, year-on-year growth in March 2026 comes out around 4.3 per cent: the weakest in four years.

With headline CPI inflation running in the 4 to 6 per cent band, that is a real wage somewhere between flat and shrinking — at the exact moment the published series says rural India is enjoying its best pay rise in a generation.

This splice is an analyst’s reconstruction, not an official figure, and it should be read as an estimate. But the sample change it corrects for is a matter of public record, and no official back-series has been published that would let anyone do better.

Why a sampling note becomes a pay cut

This would be a technical curiosity if the index were only used for measurement. It is not. CPI-AL is the index used to set wages under the central government’s rural employment guarantee, and it is the reference both the Centre and states use for minimum wages for agricultural and rural labour. It also feeds dearness allowance for large numbers of workers in government, public sector undertakings and private establishments.

So the same village sample now does two jobs at once. On the wage side it pushes the reported average up. On the price side it appears to push measured inflation down: preliminary state-level data show inflation in the newly added regions running near 2.9 per cent against a 3.9 per cent national average in May 2026.

And the sample is not the only thing pushing the price side down. Buried in the list of methodological improvements is a line with more force than it looks: the revised series uses the geometric mean in place of the arithmetic mean, on the stated ground that it “moderates the volatility in prices”.

Moderates in one direction. For any set of positive numbers that are not all identical, the arithmetic mean is always greater than the geometric mean — that is a theorem, not a tendency. Swap one for the other in the averaging step of a price index and the index comes out lower, before a single price has actually changed. This is well understood in index construction and is precisely why several statistical agencies adopted geometric averaging: it produces a lower measured inflation rate. The change is defensible, standard, and downward.

The government does not dispute the direction. Its own release states plainly that average annual inflation under the new base “have been moderate as compared to the old series”. The first prints bear that out: for June 2025, CPI-AL and CPI-RL both stood at 134 points, with year-on-year inflation of 1.42 per cent and 1.73 per cent respectively.

Consider what 1.42 per cent means as an input to a wage formula. Whatever agricultural labourers were experiencing in mid-2025, a 1.42 per cent annual rise in their cost of living is a number that, applied mechanically, justifies almost no wage revision at all.

There is a real argument on the other side, and it was the expectation going in. When the revision was announced, the case made for it — including in policy commentary at the time — was that a basket built on 2011-12 consumption rather than 1983 would capture what rural households actually buy, and would therefore produce more realistic and probably higher wage rates. That is a coherent argument and it may yet prove right over a full cycle. It is simply not what the first year of data shows.

Read the two halves together. A sample that overstates the wage level, and a methodology that by construction understates the price level, will make real rural wages look healthier than they are twice over — and because the indexation formula runs off the price side, it will also produce smaller statutory wage revisions than the actual cost of living warrants. The measurement error does not stay in the spreadsheet. It arrives in a bank account.

The scheme that pays the wage has just been replaced

All of this lands in the middle of the largest change to rural employment policy in twenty years. The Viksit Bharat — Guarantee for Rozgar and Ajeevika Mission (Gramin) Act, 2025, universally shortened to VB-G RAM G, received presidential assent on 21 December 2025, repealing MGNREGA. It came into force on 1 July 2026.

The reserved no-work window is the most interesting provision on that table, and it points straight back at the wage question. Up to 60 days are deliberately kept free during sowing and harvest so that labour stays available for farms. That is an explicit admission of what the economics literature has argued for two decades: the public works wage sets a floor under the private farm wage, and when the scheme runs, farms compete with it. An ICAR-published study using ARIMA-intervention modelling on gender-disaggregated wage data from 1995 to 2022 found precisely that — the rollout of MGNREGA between 2006 and 2008 significantly raised real farm wages, with the effect growing over time. The same study found that the Covid lockdown, by contrast, produced no lasting wage effect at all.

Two figures on that table do not reconcile with the ones the government published a year earlier, and are worth flagging rather than smoothing. In March 2025 the Ministry of Rural Development notified an MGNREGA national average of ₹370 a day, up from ₹349, with Haryana at ₹400 and Nagaland and Arunachal Pradesh at ₹241. The VB-G RAM G announcement describes the outgoing MGNREGA average as ₹298.8. Both are official. The most likely explanation is that one is a simple average across states and the other is weighted — but no published note says which, so the “10 per cent rise” should not be compared against the ₹370 figure.

The funding change is the part states are contesting. Under MGNREGA the Centre paid the entire unskilled wage bill. Under VB-G RAM G most states pay 40 per cent of wages, materials and administration. State outlay is projected to rise from about ₹8,690 crore in FY26 to ₹35,300 crore in FY27 — more than fourfold. Punjab and Telangana have formally asked for more central support; the Centre has retained the 60:40 formula. In the Union Budget for 2026-27, MGNREGA’s own line fell from ₹86,000 crore to ₹30,000 crore while VB-G RAM G received ₹95,692 crore.

The other half of the transfer: cash

Rural India receives money from the state through two quite different instruments, and they are usually discussed separately. One pays for work. The other simply pays.

PM-KISAN transfers ₹6,000 a year to landholding farmers in three instalments, straight to a bank account, with no work requirement and no targeting beyond landholding. It has moved ₹4.47 lakh crore across 23 instalments; the 21st alone, in November 2025, sent ₹18,680.63 crore to 9.34 crore farmers. In August 2026 the Cabinet extended it from 2026-27 through 2030-31 with an outlay of ₹3.15 lakh crore.

Behind it sits the wider Direct Benefit Transfer architecture: 314 schemes across 53 central ministries, beneficiary coverage up sixteen-fold from 11 crore to 176 crore, and claimed cumulative savings from plugged leakage of ₹3.48 lakh crore. Subsidy spending has fallen from about 16 per cent of total government expenditure to 9 per cent since DBT began in 2013.

The contrast between the two instruments is sharper than the shared plumbing suggests. A rural employment wage is self-targeting — only someone willing to do manual work at ₹327 a day turns up, so it reaches the landless, and it puts a floor under the private wage. PM-KISAN is targeted by landholding, which means it reaches farmers and by construction excludes the agricultural labourers who own no land, and it exerts no pressure on the wage at all. They are not substitutes, and a rupee moved from one to the other changes who is helped.

Which returns to the measurement problem. The cash transfer is a fixed nominal amount: ₹6,000 has not changed since 2019, so inflation has already eroded roughly a third of its real value. The work wage is indexed — but indexed to CPI-AL, the series whose sample just changed in a direction that appears to understate rural inflation. One transfer is silently shrinking because nobody adjusts it. The other risks shrinking because the thing that adjusts it may now be measuring the wrong villages.

What this piece could not establish

The break-adjusted growth figure of about 4.3 per cent is a splice by an outside analyst, not an official statistic. The Labour Bureau has not published a back-cast series on the new sampling frame, which is what would settle the question properly; until it does, any estimate of the underlying trend is a reconstruction. The direction is well supported — the sample change is documented and its composition effect is arithmetic — but the precise magnitude is not.

The two national average wage figures for MGNREGA, ₹370 and ₹298.8, are both government numbers and this piece could not determine the weighting behind either. The 62 per cent single-month jump in fishing wages and the 29 per cent in beedi-making are reported from the Bureau’s own tables; no occupation-level explanation has been offered for them.

Finally, VB-G RAM G is nine weeks old at the time of writing. Person-day figures in its first months are not yet a reliable guide to how it will run: reports of a sharp year-on-year fall sit alongside a month-on-month rise from 30 crore person-days approved in April 2026 to 43 crore in May, and the two are not necessarily in conflict. Judgement on the scheme should wait for a full season.

Related on this blog: CPI Inflation Heatmap — India, June 2026 · Inflation Trends Globally: What Five Years Actually Cost · Fourteen Years of Indian Wholesale Prices: Diesel Up 79%, Urea Up 13.8% — and Three Corrections to Our Own Earlier Posts — the CPI/WPI series this post's real-wage math is built on.