EconomyStates
Do the monthly readings say which states are growing?
Not reliably. Across India's 19 larger states, a state's yearly growth in GST collections, e-way bills or vehicle registrations lines up only weakly with its output growth (average rank correlations from −0.04 to 0.18), and a state's lead on any of them one year says little about the next (from 0.02 to 0.15). Both tests were written down before they were run; neither passed.
Written for readers who use state data to judge state economies. Data checked on 27 Oct 2026.
Do the readings move with state output?
A rank correlation of one would mean the reading put the states in exactly the order of their output growth; zero, no relation. Vehicle registrations, with the longest record (17 years, FY06 to FY25), average 0.16. That is more than chance would give (permutation p 0.003) but too weak to rank states: in FY25 it was −0.54. GST averages 0.18 over 4 years and e-way bills −0.04 over 3.
Two tests of the monthly state readings
Average Spearman rank correlation across the 19 larger states. Tracks output: a reading's yearly growth against nominal GSDP growth. Lasts a year: a state's three-month growth against its own a year earlier, ranked among states.
How to read it. One would be a perfect ranking, zero none. Below about 0.4, a reading puts states in an order little better than shuffled.
Loading chart…
Vehicle registrations against state output, year by year
Spearman rank correlation of each state's growth in registrations with its nominal GSDP growth; financial years, marked by the calendar year they start in. FY12 (a change of GSDP base), FY21 and FY22 are left out.
How to read it. The link comes and goes: strong in some years, reversed in others. A reading that tracked output would sit above the line year after year.
Loading chart…
Does a state's lead last a year?
Over 209 months from Mar 2005, the states growing fastest in vehicle registrations were no more likely than chance to be the fastest a year later (average 0.02, p 0.32); in the latest month the correlation was −0.39. GST (0.13) and e-way bills (0.15) do little better. UPI payments, only 16 months long, average 0.20: too short to judge.
States' growth against their own a year earlier, ranked
Each month, the Spearman rank correlation across the larger states of three-month growth on a year earlier with the same measure twelve months before. Blank where the comparison touches April 2020 to March 2022.
How to read it. Above zero, the states ahead a year ago tend to still be ahead. The lines cross zero again and again.
Loading chart…
The tests, as written down beforehand
Both tests (C110 and C111 in our research log) fixed the sample, the statistic and the pass line before they were run. Each p is the share of 10,000 random reshufflings of states that matched or beat the observed correlation.
| Reading | Tracks outputmean ρ (years) | p | Lasts a yearmean ρ (months) | p |
|---|---|---|---|---|
| GST collected | 0.18 (4) | 0.062 | 0.13 (34) | 0.049 |
| E-way bills | −0.04 (3) | 0.62 | 0.15 (27) | 0.11 |
| Vehicle registrations | 0.16 (17) | 0.003 | 0.02 (209) | 0.32 |
| UPI payments | 0.07 (1) | 0.39 | 0.20 (16) | 0.098 |
| Electricity demand | – | −0.27 (12) | 1.00 |
Pass lines: tracks output, mean ρ at least 0.4 with p below 0.05; lasts a year, at least 0.3 with p below 0.05. Graded on GST, e-way bills and vehicles; electricity and UPI are too short and are shown only. Result: no reading passed either test, as predicted. Method.
What the numbers can't tell you
- What the readings measure
- GST is collected where the supplier is registered, e-way bills follow goods rather than value added, vehicles count the state of registration and UPI the payer's state. None is designed to measure a state's output.
- State GDP is itself uncertain
- State output is estimated by state statistics offices, late and with revisions. Part of the weak link may lie on that side: a reading cannot track a figure that is itself noisy.
- Years left out
- The tests leave out FY21 and FY22 and any month whose comparison touches April 2020 to March 2022, when lockdowns and their base effects moved every reading at once.
- What this does not say
- The readings can still describe the month well: where GST was collected and how many vehicles were sold is true. They do not, on this evidence, rank states by how fast their economies grow.
The workings
The question
Data
Every chart has a table view, a CSV download and a link to its published JSON. Bundles are listed in data status.