Before you start
When the big numbers come out, how to join two Indian datasets without losing rows, and the questions India's data cannot answer yet.
When the numbers come out
The statistics ministry fixes its release dates a year ahead. Knowing them lets you prepare before a release instead of reacting after it.
- 15th of every month
- Labour force survey, monthly bulletin
- 12th of every month
- Consumer Price Index
- 28th of every month
- Index of Industrial Production
- 30 November 2026
- GDP estimates for July to September 2026
- By December 2026
- GDP back series on the 2022-23 base, expected
- 7 January 2027
- First Advance Estimates of GDP for 2026-27
- 1 February 2027
- Union Budget and Economic Survey, by convention
- February 2027
- Census 2027 population enumeration, including caste
- 26 February 2027
- Second Advance Estimates of GDP and revised estimates for earlier years
- 25 March 2027
- Labour force survey annual report for 2026, with unit-level data
- Around mid-2027
- First results of the National Household Income Survey, expected
Dates can move around holidays. Check the official release calendar before planning around one. Surveys run by other bodies, such as the family health survey or the crime report, follow no fixed date.
Joining datasets without losing rows
Most errors in Indian policy analysis happen during the merge, not the maths.
Use codes, never names
The Local Government Directory gives stable codes down to village and ward, and records every split, merger and rename. The Census uses its own 2011 codes and some ministries keep their own. Build one table that links them, keep it, and never join two datasets on a name.
Decide what a district is before you begin
India had 640 districts at the 2011 Census and has well over 750 today. New districts are carved from parts of old ones, so no clean match exists either way. For a time series you can keep 2011 boundaries and add newer districts back to their parents, or use only the years that share boundaries. Choose one, note it, and keep to it.
Watch for states that changed shape
Telangana was formed from Andhra Pradesh in 2014. Jammu and Kashmir became two union territories in 2019, with Ladakh separated. Dadra and Nagar Haveli merged with Daman and Diu in 2020. Any series that crosses these dates needs a clear rule.
Check units and kinds of year
Indian tables mix lakh, crore, million and billion, sometimes in one sheet. They mix financial years (April to March), calendar years and farm years (July to June). Money may be in current or constant prices, and since 2026 the base for constant prices has changed too.
Read the notes before the table
The notes say which states are left out, when a series was revised, what a definition now covers and which figures are provisional. NFHS-6, for example, does not cover Manipur. None of this travels with the spreadsheet you download.
When the data is not published
Every ministry has a Chief Data Officer, listed on the open government data platform, and a request can go to them. A Right to Information application is the formal route. If official data does not exist, careful counts built from news reports and court records have filled gaps before, as long as the method is written down.
What India does not measure well
Where the evidence is thin, confident claims deserve doubt, ours included.
Income
Until the first official household income survey reports, expected around mid-2027, claims about how income is spread rest on consumption data, tax tables or models.
Migration
There is no regular count of people who move. The main sources are old and miss seasonal and short stays, because they ask where people usually live.
Small informal businesses
Enterprise surveys sample the tens of millions of household businesses but do not list them. Voluntary registration mostly captures those seeking credit or government contracts.
Disability
The 2011 Census counted about 2.2 per cent of people as disabled, well below global estimates. Narrow categories and reluctance to report both hold the count down.
The very rich
Surveys miss the wealthiest households, who refuse more often and report less. With no wealth data and no public tax records, estimates of the top vary widely.
City wards
Little official data exists at ward level. Municipal budgets, services and local pollution are mostly unavailable in usable form for most cities.
Private health care
Half or more of treatment happens in private facilities that report little. The main health information system covers public facilities.
Caste
No full caste count has been published since 1931. Census 2027 is set to record caste, which will be the largest addition to Indian social data in nearly a century.
Citing a dataset
Give the publisher, the exact name of the series or round, the period it covers, the link and the date you opened it. The date matters because Indian tables are revised quietly and portals are rebuilt without redirects. Each Sarani entry gives a ready citation with a copy button.