EconomyGrowth
What will the next GDP figure be? An experimental nowcast
Our experimental nowcast puts India's GDP growth for July to September 2026 between 6.9% and 11.0%, a 70% range, with 9.0% at its centre. The RBI projected 7.2%. The nowcast has not yet beaten simple benchmarks, so read the range before the point.
Written for anyone waiting for India's next GDP figure who wants to know what the monthly data already say. Issued 25 October 2026.
Experimental. It has not yet beaten simple benchmarks. Both models were written down before they were run, and neither will change before MoSPI's estimate on 30 November 2026.
Q2 FY27, July to September 2026
These are the bridge model's figures. It reads each sector's growth from the monthly indicators the statisticians themselves use, from the index of industrial production to cement output, air passengers, e-way bills and bank credit, using how the two have moved together in the past, and weights the sectors by their share of output. The range is wide on purpose: it is set by the model's own errors in testing, and on the old base those were large. Mining, manufacturing, utilities and finance and real estate have two months of the quarter's data so far, the other indicator sectors three; September's industrial production is due on 28 October 2026, and the nowcast is re-issued then.
The challenger, also written down in advance, assumes each sector grows at its indicator's rate plus the gap between the two over the last year, with nothing fitted. It puts GDP growth at 9.6% (7.1% to 12.1%) and output at 9.8%. In testing on the old base it did worse than the bridge.
Against the benchmarks
A nowcast is worth something only if it beats simpler guesses. The plainest is that growth stays where it was: 7.8% in Q1 FY27. An AR(1) model, which pulls growth toward its average, gives 7.5%. The RBI's projection, from its policy statement of 7 October 2026, is 7.2%. Both our models sit above all three.
GDP growth in Q2 FY27: our nowcasts and the benchmarks
Real GDP growth on a year earlier. The vertical line is Q1 FY27's growth, 7.8%.
How to read it. Dots with lines are our two models and their 70% ranges; plain dots are benchmarks. A range drawn hollow spans last quarter's growth: on its own errors, the model cannot yet say growth has changed.
Loading chart…
RBI projection: RBI, Resolution of the Monetary Policy Committee, 5-7 October 2026 (press release 63742). The RBI's survey of professional forecasters joins the benchmarks once its median for the quarter is in our data.
The sectors behind it
Where the nowcast differs from last quarter, and why. The biggest move is in manufacturing: the bridge expects 14.2% growth against 9.2% in Q1 FY27. Its indicator grew 8.6% on average in the months so far, and the bridge, fitted on past quarters, multiplies indicator growth by 1.44 and adds 1.9 points; the challenger, with no fitting, expects 12.2%. The bridge's figure adds 2.3 points to output growth. Agriculture, forestry and fishing and public administration, defence and other services have no monthly indicator in our data; both models carry forward the average of their last four quarters.
Sector growth in Q2 FY27: the two models and the last quarter
Real growth on a year earlier, %. Q1 FY27 as now published.
How to read it. Where the models' bars are far from last quarter's, the monthly data say that sector has changed pace. Where the two models disagree, the past relationship between sector and indicator is doing the work.
Loading chart…
Every issue
The nowcast is re-issued whenever new monthly data or revisions change it. Every issue is kept here, with what moved it. The one scored is the last before MoSPI's release.
| Issued | Bridge: GDP | 70% range | Output | Challenger: GDP | What moved it |
|---|---|---|---|---|---|
| 25 October 2026 | 9.0% | 6.9% to 11.0% | 9.2% | 9.6% | Up 0.04 points on output. New data: construction (1 more month), finance, real estate and professional services (-1 more months). Biggest changes in contribution (points): finance, real estate and professional services +0.05; construction −0.01. |
| 24 October 2026 | 8.9% | 6.9% to 11.0% | 9.1% | 9.5% | The first issue for this quarter. |
How it has done
We tested both models on past quarters, each estimated only from the quarters before it, with two months of indicators. On the new base, over 9 quarters, the bridge's typical error (root mean square) was 0.66 points, against 0.79 for "growth stays where it was"; a test of the difference cannot tell it apart from luck. On the old base, over 29 quarters, it was 1.70 against 1.57: worse than the naive guess.
Backtest on the new base: the estimates against what was published
Real GVA growth, quarterly. Each estimate uses two months of indicators and only earlier quarters. Pseudo-real-time: tested on today's revised figures, not the first estimates.
How to read it. The closer a line follows the bold one, the better. Nine quarters is too few to rank the lines with confidence.
Loading chart…
Backtest on the old base: how far off each estimate was
Estimate minus published output growth, quarterly, old 2011-12 base; lockdown quarters left out. Two months of indicators.
How to read it. Bars above zero are estimates that were too high. From Q4 FY17 to Q4 FY20 the bridge was too high 13 quarters running, when output grew well below what its indicators implied.
Loading chart…
| New base (9 quarters) | Old base (29 quarters) | |
|---|---|---|
| Bridge | 0.66 | 1.70 |
| Challenger | 0.67 | 2.43 |
| Last quarter's growth | 0.79 | 1.57 |
| AR(1) | 0.79 | 1.42 |
| Average of the last four quarters | 0.72 | 1.65 |
How it will be scored
The rules were registered before any model was run. The nowcast scored is the last one issued before MoSPI's release. It is judged against MoSPI's first estimate, not later revisions, by its absolute error against each benchmark (last quarter's growth, the AR(1) and, for GDP, the RBI's projection), and by whether the first estimate falls inside the 70% and 90% ranges.
One quarter decides nothing, and will be reported as a single result, never as a pass or a fail. After eight scored quarters we will publish each model's typical error against each benchmark, a test of whether the difference is more than luck, and how often the first estimate fell inside each range against the 70% and 90% promised. What we expect for Q2 FY27, written down in advance: the first estimate of output growth falls inside the bridge's 70% range, and the bridge's GDP error is smaller than the error of assuming growth stays where it was. The result will be logged here whatever it is.
The workings
The question
Method
Data
Every chart has a table view, a CSV download and a link to its published JSON. Bundles are listed in data status.