Data to 9 October 2026
About
Menu

Markets

Trading

Flows

Funds

Economy

World

Tools

The questions

Pre-registered research

NSDL and CDSL: additions worth making to Tipsheet

Inspected 9 October 2026. This is a bounded source-discovery audit, not a completed historical ingestion or a publication. Source captures live in docs/reference/depositories-2026-10-09/; reproduce the numerical checks with python3 audits/depository_discovery.py.

Recommendation

Start with the two depositories’ monthly BO Aggregate reports. They offer genuinely new material beyond Tipsheet’s existing demat counts, NSDL sector flows and foreign bond holdings. Add FPI country and investor-type holdings next. The issuer archive could then enrich the bond section, after cleaning its reference fields. State, age and gender data would make an excellent editorial project, but bulk access remains unverified.

1. Who holds the assets behind the demat boom?

Downloads verified: CDSL BO Aggregate, monthly links from 2015, and NSDL Depository Statistics. CDSL’s current link points to August 2026 even though other tables on the page are September. NSDL’s initially empty HTML is a client-rendering issue: its public API returns download metadata.

The NSDL website’s own JavaScript uses these public GET endpoints:

The August NSDL workbook downloaded successfully from the field_file path in that listing. It has 13 investor categories, account counts and equity/debt/funds/other custody value. The category values sum to 46,700,882 accounts and ₹552.696 lakh crore of custody, matching NSDL’s published August statistics to their displayed precision. CDSL’s August category sums match all six printed totals, with currency gaps below ₹0.01 crore.

Observed August 2026 findings:

Tipsheet treatments: paired account/value shares; individual equity and fund holdings over time; total holdings versus holdings per account; a monthly chart of investor-category asset mix. These could sit alongside /activity/who-trades/ and ownership.

Definitions that matter: accounts are not unique people. NSDL’s workbook explicitly says both count columns include suspended accounts. Do not equate its “Active” column with recent trading activity. Its Funds column includes AIFs, InvITs and REITs as well as mutual funds; CDSL’s MF column is therefore not directly equivalent. NSDL separately identifies residents, HUFs, NRIs and FPIs; CDSL’s published categories differ. Harmonise only categories whose definitions agree, and preserve unmatched groups. Equity custody includes unlisted/promoter/corporate holdings. Custody value is a stock influenced by prices, issuance and transfers, not net investment. A mean per account is neither a median portfolio nor the experience of a fixed investor cohort.

2. What is inside the foreign-investor label?

Verified sources: September 2026 CDSL country AUC and CDSL category AUC. The archive pages embed current links in JavaScript window.open handlers, so collecting only anchor hrefs misses them. Both archives show year selectors back to 2012, but a complete backfill has not been tested.

NSDL production FPI URLs disconnected from this environment. The pilot pages were reachable; their September country US row matches CDSL exactly, and the country total matches the category Grand Total. This validates the sampled current values, not every pilot endpoint. Prefer the confirmed CDSL production publications for new extraction.

Observed September findings: US-domiciled FPIs held 44.44% of foreign equity AUC and 41.31% of all FPI AUC. Registered pension funds held ₹6.84 lakh crore of equity and registered sovereign wealth funds ₹4.55 lakh crore.

Tipsheet treatments: domicile shares through time; debt versus equity mix by domicile; foreign pension, sovereign, central-bank and fund holdings. Keep “Other” because top-ten membership changes. Domicile does not identify ultimate investor nationality. A change in AUC is not a flow. Category labels do not establish investment horizon or passive/active strategy. Include registered and expired tables when reconciling to all-FPI totals. The separately reported FAR, MF and AIF fields introduce a layout break from August 2024.

NSDL also publishes category-wise holdings for clients of custodians (banks, insurers, domestic pension funds, MFs, PMS, AIFs). It is useful for institutional asset mix, but its denominator is custodial assets, not all Indian financial wealth. It overlaps intermediary and underlying holdings, so do not add the categories into household wealth.

3. A bond maturity and custody atlas

Verified download: CDSL August 2026 Issuer Details CSV, linked from Depository Statistics, 11,510,833 bytes. Its archive contains 136 file links in the inspected page, from 2015 through August 2026. These are listed files, not proven gap-free coverage.

The current CSV has 40,198 rows, 40,197 distinct ISINs and 29 columns. There are 7,445 debt rows: all have a populated maturity field, 5,997 a rating field, and 1,608 a positive monthly settlement value. Other columns include coupon, payment frequency, listed status, issue size, demat quantity/value and monthly settlement quantity/value.

Tipsheet treatments: debt maturity buckets by issuer and credit rating; CDSL custody concentration; how much of the registered bond universe records settlement activity; ISIN reference enrichment for the bond-trade feed. A populated date is not proof of a valid date. Show missing ratings as missing and use rating dates.

Quality blockers: one duplicate ISIN; 6,015 rows have demat quantity above the reported issued quantity; observed legacy classification inconsistencies; date fields lose leading zeroes; settlement blanks have no verified zero meaning. Do not publish demat/issued ratios before investigating these failures. Settlement value is depository activity, not necessarily exchange turnover or a price from which to infer yield. CDSL alone is not the full debt market. These are discovery checks, not a completed reference-data audit.

4. India’s investor geography, age and gender

NSDL’s published research data-sharing list identifies state-wise account counts with age/gender distributions as public/shareable, and PIN-code counts with age/gender distributions as voluminous data available on request. This is an access lead, not confirmation that a downloadable dataset is currently exposed.

Potential treatments: states gaining share of accounts; female account share; age composition; accounts per adult using aligned population denominators. Request history, definitions and aggregate fields only. Multi-account ownership prevents interpreting account/population ratios as the percentage of people investing.

CDSL’s geographical PDF was downloaded and rendered: it contains counts of service centres and PIN codes by participant type, not account-holder geography. Its DP CSV likewise maps service centres and registered offices. Do not use either as investor location.

Leads to defer

Retrieval and publication notes

Keep immutable files, URL, retrieval date and SHA-256; detect dates from content; reconcile category totals before publishing; preserve unknowns and source categories; backfill from discovered links, with bounded retries and incremental monthly refreshes. No paid model calls are needed for the CSV, workbook and HTML extraction.

The archived NSDL FPI disclaimer prohibits reproduction/redistribution, while NSDL’s separate research-sharing policy lists public/shareable datasets. Their implications for a public Tipsheet product remain unresolved; technical download success is not a reuse licence. Verify applicable publication terms before a live rollout. No external request was sent and no site change or deployment was made.

This is docs/research/depository_portals_2026-10-09.md. The specification was committed before any result was computed; changes after that are logged in it with dates and reasons.