Investor participation: methods
This page uses NSE’s public Market Pulse archive. The first issue is September 2026, with participant and profile observations through August 2026. Report month, observation dates, source captions and PDF page numbers travel with the downloadable data. No client-level records are collected.
Populations and periods
The monthly participation diagram includes all investor categories. Its registered base is NSE’s unique PAN-based registration measure. Active means traded at least once during the specified month, across cash and equity F&O. Not trading in a month does not imply an investor sold their holdings or is dormant. Counts are source-rounded; the diagram’s branch counts need not sum exactly to the rounded active total. We display the source values without forcing an identity or estimating missing participants.
The annual table covers domestic individuals, NRIs, sole proprietorships and HUFs who traded at least once during each trailing twelve-month period. There is one observation per year, with the same ending month. It is not a monthly time series. An investor active in both segments appears in each segment total. Cash-only, F&O-only and both are mutually exclusive groups. Cash total plus F&O total must not be treated as a unique-investor count. Year-on-year changes compare adjacent twelve-month windows in the same source table. Source values are rounded to 0.1 lakh.
Age tables cover individuals and sole proprietorships only. The registered base is measured at month end. New registrations cover April through the observation month of the current Indian financial year. Each percentage uses its own cohort denominator. Historical source columns remain in the JSON. These age shares do not describe the age mix of active traders.
State counts are unique registered investors, reported in thousands. State shares are NSE’s source-rounded percentages, with every category retained in the national denominator. The chart compares March 2020 with the latest observation for the fifteen largest named current bases. Share change is latest share minus March 2020 share, in percentage points. The full table retains Others, a source grouping of smaller states/territories, but excludes it from named-state rankings. Source labels Keralam and Orissa are displayed as Kerala and Odisha; original labels remain in downloads. Geographic boundaries and historical attribution follow the publication, including its historical Telangana rows. No population penetration, invested-money flow or trading intensity is inferred from registrations.
Acquisition and validation
The collection phase checks the official listing daily using the
existing proxy-first exchange HTTP adapter. It discovers dated Market Pulse
PDF links only on NSE’s archive host. A new issue is downloaded and parsed
with pypdf. The same issue is rechecked weekly for replacements. No OCR or
model call is required. The investors step reads these sealed inputs during
the offline build and never opens a network connection. A persisted acquisition
health record keeps failed-refresh warnings visible in that build. Layout extraction handles tables; reading-order text
and explicit node labels handle the monthly diagram, so its visual column
order cannot swap F&O-only and overlapping investors.
Parsing uses caption text rather than fixed table numbers or PDF pages. Publication and observation months must agree, annual windows must be twelve months and form consecutive annual observations, and population definitions must be present. Both cash/derivatives cohort identities must agree within independent rounding bounds (0.15 lakh). State totals reconcile within the number of independently rounded thousand-count cells; their shares and counts are cross-checked. Age cohorts must be complete and sum to 100% within rounding. The registered PAN total is also checked across the diagram and state table. Unknown layouts are rejected as a whole. No failed extraction becomes zero.
PDFs and compressed extracted-page records are cached by issue and content
hash in .cache/publishers/nse-investors/, included in the existing R2 publishers
cache. A checked-in four-page text seed, carrying the full PDF’s SHA-256,
allows offline builds. Failed refreshes retain the last validated issue with
a visible warning. Observations over 90 days old receive a staleness notice.
The first PDF is 9.37 MB. At that size, twelve monthly PDFs would occupy about
113 MB before any same-issue revisions; this is a measured-file extrapolation,
not a guarantee about all historical issues. Extracted data is much smaller.
Research references
The page links the Factbook and selected entries in NSE’s working-paper catalogue. These are reading references, separate from the automatically extracted statistics. A published paper’s finding is not automatically a live indicator, and its underlying investor-level data may not be publicly available. No paper results or causal claims are reproduced here. Source revisions are retained; this latest-issue dataset is unsuitable as a point-in-time backtest input without original publication/availability dates.