Active funds, new launches and fund flows: methods

Code: pipeline/tipsheet/compute/fund_scorecard.py, pipeline/tipsheet/compute/fund_flows_lab.py, pipeline/tipsheet/publish/funds_insights.py (step funds_insights). Tests: pipeline/tests/test_fund_scorecard.py. Pre-registrations and results: C50 to C53 in docs/research/site_calculations_spec.md.

Used on /funds/passive/ (did active funds beat the index; do winners keep winning) and /funds/ (flows against assets, return-chasing, SIPs through falls, NFOs and the market, new funds).

Which funds

Every open-ended active equity scheme in AMFI’s daily NAV history, all eras from 2006, including schemes that have since closed or merged. A scheme enters through its growth plans: one direct and one regular (mf_realised_cost.plan_pairs). Two adjustments, both general rules:

• Dividend-yield funds are kept. The option test reads the word “dividend” as an IDCW option, so the phrase “dividend yield” is masked while pairing.

• Duplicate codes. The pairing needs exactly one growth code per plan and name. AMFI’s file holds codes that stopped years ago under names that normalise to the same words (a UTI Large Cap code that ended in 2007; “UTI Growth & Value Fund”, which reduces to “uti value”). Before this rule those schemes dropped out entirely. Codes that stopped before direct plans existed (1 January 2013) are removed, and where two growth codes still share a name and plan, the one alive latest is the scheme’s. This added seven schemes and moved the headline shares by 0.1 to 0.2 points.

• Re-coded schemes. Codes sharing an ISIN across a fund-house hand-over are joined into one fund (next section).

• IDCW spelled out. “Income Distribution cum Capital Withdrawal” plans are excluded even when the scheme’s own name contains “Growth”.

Direct plans are the headline: they carry only the fund’s own costs, as an index fund’s direct plan does. Regular plans, which also pay the distributor, are shown beside them.

Data integrity: renames, takeovers, mergers, category changes

Checked on 2026-10-22 against the Data bank snapshot of that day.

• Renames. AMFI usually keeps a scheme’s code when it is renamed, and its NAV file prints the current name, fund house and category on every row: none of the 1,400 plan codes in the sample (including old codes of re-coded schemes) shows a second name or fund house. So the file cannot show what a scheme was called earlier.
• Takeovers that re-code. Some takeovers did not keep codes. On 28 November 2022 AMFI closed the L&T schemes’ codes and opened new HSBC codes for the same schemes (HSBC Value Fund, code 151110, was L&T India Value, code 118102, which stopped on 25 November). The same happened with IDBI to LIC (2023), Principal to Sundaram (2021–22), BNP Paribas to Baroda BNP Paribas (2022) and JPMorgan to Edelweiss (2016). Read naively, each old scheme looks closed and each new code looks like a launch. The depository ISIN carries over, so two codes that printed the same ISIN, the old one stopping at most 31 days before the new one starts, are treated as one fund: the old NAV history is joined on, scaled so the two meet at the hand-over (the few days between count as no change), and the old codes’ assets count as the new code’s. Across AMFI’s file this links 614 code pairs; 27 schemes in the sample are re-coded, which removed 29 false closures and 21 false launches. Joined series show no monthly move above 31% (the March 2020 crash). A takeover in which the acquirer merged a scheme into one of its own (L&T India Large Cap into HSBC Large Cap) is a real merger: the absorbed scheme counts as closed.
• Mergers and closures. When schemes merge, the absorbed scheme’s code stops; the survivor’s continues. The absorbed scheme is kept, with its NAVs to the last day. In a window it did not finish, it counts as not beating its benchmark (S&P SPIVA’s convention). Its investors were moved into another fund, so this is strict; the “to last NAV” figures judge it on its return up to its final day instead, and the page reports both (five years to September 2026: 67% against 71% of diversified direct plans).
• NAV discontinuities. A merger, bonus or face-value change that broke a series would show as a jump. Month-end moves above 35% in the sample: seven, five of them March 2020, one a dead ING scheme in 2014 (outside every window) and one Nippon India Taiwan Equity in April 2026, a genuine rally checked day by day.
• Coverage. In June 2018, schemes in the sample held 98.7% of open-ended active equity assets (IDCW plans count with their scheme). The 1.3% outside are schemes whose plans cannot be paired by name (Nippon India Vision’s direct plan has a different name; UTI Master Equity has no growth plan). Old codes of re-coded schemes count with their scheme. Schemes outside the sample that later closed held ₹694 crore, 0.1% of assets: the remaining survivorship leak.
• Category changes: the real limit. Because the file prints today’s label, a scheme’s category at an earlier date cannot be read from it. Each fund is held to the benchmark of its last category. This matters most for multi-cap funds: in January 2021 most moved to the new flexi-cap category, and today’s multi caps (benchmarked to the Nifty 500 Multicap 50:25:25) may have been something else before. Funds that closed before SEBI’s 2018 recategorisation keep the pre-2018 label (“Growth”) and cannot be placed in a category, which is why category figures use only windows opening after June 2018, and why C50’s 2015 to 2017 cohorts are flagged (spec amendment 6).

Benchmarks

Tier-1 total-return indices (dividends in), from NSE Indices, repaired against price indices (compute/indices.py): large cap → Nifty 100; mid cap → Midcap 150; small cap → Smallcap 250; large and mid cap → LargeMidcap 250; multi cap → Nifty 500 Multicap 50:25:25 (launched December 2020; earlier values are back-calculated and used only as a hurdle); flexi cap, ELSS, focused, value, contra, dividend yield, sectoral and thematic → Nifty 500.

Check: index funds tracking each benchmark trail it by what their costs predict over the year to September 2026 (median direct plan: Nifty 100 −0.23 points, Midcap 150 −0.40, Smallcap 250 −0.53, LargeMidcap 250 −0.33, Nifty 500 −0.34), so NAVs and indices line up in date and level.

The scorecard

For each window end from June 2021 (three years) or June 2023 (five years) to the last full month: every fund alive at the window’s start; its annualised return less its benchmark’s; the share beating. Also: survival, the share beating when closed funds are judged to their last NAV, the share weighted by assets at the start (AMFI quarterly average AUM, both plans), and the median lead of surviving funds.

The index-fund hurdle. An index fund is the real alternative and costs something. Over each window, the median Nifty 50 index fund’s shortfall against its index (direct or regular, matching the active plan) is the hurdle; an active fund “beats the index fund” if its lead over its own benchmark exceeds that shortfall.

Neighbouring window ends share most of their months: 64 three-year windows hold only about two independent periods.

C50: persistence

Pre-registered. Formation each December 2015 to 2022: rank funds within category (at least eight funds) on three years’ excess return; evaluate the next three years, a closed fund ranking bottom. Headline: the non-overlapping cohorts 2015, 2018 and 2021, with 10,000 within-category shuffles of the evaluation ranks.

C51: new launches

Pre-registered. Active equity schemes whose direct and regular plans both start between January 2019 and September 2025, within seven days of each other (excludes conversions and plans opened later). Return from the first NAV (allotment) to 12, 24 and 36 months later, against the Nifty 500 TRI over the same dates; cumulative. Intervals resample launch quarters (2,000 draws). Exploratory, added after the pre-registered result: against the category’s own index, and against the median older fund (launched a year or more earlier) in the same category over the same month-ends.

Flows (fund_flows_lab.py)

• Flows against assets: active equity net flow ÷ equity fund AUM a month earlier; twelve months’ flow ÷ AUM twelve months earlier. AMFI monthly report from May 2019.
• C53, return-chasing: nine equity categories (flexi and multi cap as one). Each category’s return is implied by its own assets and flows, R = (A₁ − A₀ − F) / (A₀ + F/2); flow intensity F / A₀. Monthly Spearman correlation across categories; mean over months, Newey-West t with 12 lags; a 1,000-shuffle placebo, which ignores the overlap of twelve-month windows and so overstates significance for them (the forward test’s placebo p is 0.003 while its t is 0.9; the registered statistic is the t).
• SIPs and the rest: equity funds’ net flow less all SIP contributions (AMFI, from April 2016). SIPs also go to hybrid and index funds, so this understates lump-sum buying of equity funds. A fall month: the Nifty 500 TRI month-end at least 10% below its earlier peak.
• C52, NFOs: the trailing three months’ active equity NFO money ÷ equity fund AUM three months earlier (from August 2019, when the AUM series allows). Forward returns start at the end of the month after the signal month, when AMFI’s report is out. Windows touching the 2022 pause on launches (May and June 2022) are blank.

What this does not show

• Category is today’s label, not the one at the time (above).
• Gross of tax and exit loads; the regular plan’s commission is in the NAV.
• The era: every window and launch here falls in 2013 to 2026, mostly a rising market led for long stretches by mid and small caps. Active funds, which hold more of those than large-cap indices, look better in such years.