Research
What we tested in the disclosures — including what did not hold up
Every study below was written down before its result was computed: the question, the measure, and the bar a result had to clear. Results that missed the bar are reported as misses. Data: monthly portfolio disclosures of Indian mutual funds, Jan 2022 – Aug 2026, valued at official NSE closing prices. Studies run on 10 October 2026.
Findings that held up
Each now has its own view on the site
| Finding | Evidence | See it |
|---|---|---|
| Most active large-cap funds are close to the index. 22 of 23 differ from the Nifty 100 in less than 60% of their equity (median 40%). | Active share vs a Nifty 100 built from the index funds’ own disclosures; the index funds themselves come out within 0.1%. | Large-cap table |
| A fund house’s schemes trade as a bloc. Two schemes of the same house trade a stock the same way 71% of the time; two schemes of different houses, 55%. | +15 percentage points after a placebo correction, every month 2022–2026. | Fund houses buying / selling |
| The bigger a small-cap fund, the longer it would take to sell. The largest would need months to sell half their holdings at normal trading volumes; a typical large-cap fund, under a day. | Rank correlation of size and exit time 0.85–0.95 in every year. | Small-cap exit time |
| Among flexi and multi-cap funds, the bigger ones sit closer to the index. | Rank correlation −0.42 (55 funds). But following each fund over time, growing did not reliably make it more index-like — this is a difference between funds, not proof that size causes it. | Flexi-cap table |
| Bigger funds trade less — but growing does not make a fund trade less. Among small-cap funds, the larger ones turn over less of their portfolio every year (typical turnover 2–4% a month). Large-cap funds show the same, so it is not the small-cap liquidity squeeze. | Rank correlation of size and turnover −0.61 to −0.71 in each year 2022–2026 (11–18 funds); −0.47 on the turnover figure shown on fund pages, which leaves out buying from new money. Following each fund over time: about zero. A difference in how funds are run, not something size forces. | — |
| The size of a new position tells you how long the fund will keep it. When an active fund opens a position big, it is still there a year later 58% of the time; opened small, 44%. Small openings are more often added to soon after — a starter position — but also more often dropped. Only about half of all new positions last a year. | 21,360 new positions in 343 active funds, sized against each fund’s own openings. +14 points in 2022–24 and again in 2024–25 (t 11.5). Says nothing about whether the big openings do better. | On each active fund page (“How long new positions last”) |
| How long a fund keeps what it buys is mostly decided by its fund house. The house a fund belongs to accounts for about half the difference between funds, and a house’s other funds predict how a fund will behave later. The range is wide: a typical quant fund keeps under 10% of new positions for a year, a typical DSP or UTI fund about 80%. | 306 active funds in 23 houses; 48% of the spread explained by house, against 7% for houses assigned at random. The other funds’ rate in 2022–23 predicted each fund’s 2024–25 rate (rank correlation 0.48). Describes style, not which style earns more. | On each active fund page (“How long new positions last”) |
| SEBI’s 10% limit on a single company works as a hard edge. Active equity funds hold five times fewer positions just above 10% of the fund than just below it, and a position above the line is cut the next month 58.3% of the time, against 18.5% for an 8–9% position. Funds do not crowd up against the limit; they stay clear of it. | 401,033 holdings in 6,758 scheme-months, disclosed weights. We expected positions to pile up just under 10%; they do not. About 0.5% of scheme-months still show a holding at 10.5% or more; checked one by one, only 2 were funds buying above the line. Corrected for summary lines some disclosures mislabel as shares. | — |
| New equity funds take about two months to invest their launch money. At their first month-end, new active equity funds hold a median 6.4 points more cash than established funds of the same kind; by the third month the gap is gone. Funds launched since mid-2024 started with far more uninvested money (about 19 points). | 108 funds launched 2022–2026. Our first attempt counted older funds of fund houses that joined our data late as “new”; the result shown uses the corrected definition, which we chose after seeing that problem. | — |
| Thematic funds are launched after their theme has already run. The first portfolio of a new sectoral or thematic fund had beaten the Nifty by a median 19.9% over the year before launch. The 2024 wave of launches came after a 41% run; in the following year those portfolios lagged the Nifty by about 11%. Funds launched in 2022–23 did better afterwards. | 41 funds launched 2022–2026 (32 ahead before launch, sign test p 0.0002). The after-launch figures follow each fund’s first portfolio unchanged and overlap heavily in stocks and timing — one market episode seen through many funds, not a rule that new thematic funds underperform. | — |
| Mutual fund money has spread across more companies — mostly through index funds. The 20 biggest holdings fell from 46% to 38% of the equity held by a fixed group of fund houses between January 2023 and August 2026, and the number of companies held rose from 572 to 972. Active funds’ top 20 did not shrink; the broadening came from passive funds tracking broader indices. | 8 fund houses present every month (the largest joined our data later); all houses together show the same direction. Effective number of stocks 58 → 90; active funds’ top-20 trend about zero (t −0.1). | — |
| Nifty 50 index funds hold the same portfolio — to a few hundredths of a percent. Across 49 Nifty 50 index funds and ETFs, a typical fund’s portfolio differs from the others’ by 0.004% of its value; the largest gap in any month was about 1%. Index funds on the Sensex, Nifty Next 50, Bank, IT, 100, 500, Midcap 150 and Smallcap 250 agree just as closely — small-cap index funds included. | 773 fund-months, 2022–2026. Also a check on this site’s data: share counts from 40+ fund houses’ files, valued at NSE closes, line up almost exactly. | Used as the index benchmark |
| Funds quickly sell the shares they receive in a demerger. Within three months, active funds had sold out of 52.3% of the spun-off holdings they were handed, against 22.5% for small positions they chose themselves. The big spin-offs went fastest: the number of funds holding ITC Hotels fell from 184 to 61 in its first month. | 15 demergers 2022–2026 traced from the holdings themselves (shares received in proportion to the parent), 773 positions; 95% interval for the gap 6–43 points. Small spin-offs were mostly kept. Says nothing about the spin-offs’ share prices. | — |
| Debt funds stop at 12% of the fund in any one issuer. SEBI allows 10%, or 12% with the trustees’ approval. Many debt funds go past 10% — but holdings thin out sharply right at 12%: 214 issuer exposures between 12% and 12.5%, then 47 between 12.5% and 13%. | 176,516 issuer exposures in 9,507 debt-fund months (issuer read from the bond’s ISIN). About 4% of fund-months still show one issuer above 12%; not examined case by case. | — |
| Money follows individual funds’ recent performance — and piles into winners more than it leaves losers. Funds that beat their category over the past year take in about 3 points more of their assets a quarter than middling funds; the worst performers lose only a little more than the middle. | 217 active equity funds, AMFI quarterly assets 2022–2026, flows estimated net of each fund’s own return. Rank correlation 0.155 after excluding the most recent quarter (which would otherwise flatter the result), positive in both halves. Unlike whole categories (above), single funds’ flows do follow returns. And the winners did not keep winning: over the following year, last year’s top fifth of funds did no better than the bottom fifth (in fact 2.6 points worse, not statistically significant). | — |
| New money goes partly into new companies; money leaving does not close positions. Funds taking in the most money add new holdings equal to about 14% of their existing list each quarter, against about 11% for funds taking in the least. When money leaves, funds trim across the board instead of selling out of companies. | 2,559 fund-quarters, 2022–2026; rank correlation of inflows with new positions 0.13 (t 4.2), positive in both halves; with positions closed, about zero. | — |
| A fund house’s schemes raise and cut cash together. Month to month, the cash levels of two schemes from the same house move together (correlation 0.13); two schemes from different houses barely do (0.03). The link remains after removing market-wide moves — a house-level call on how much to keep in cash. | 132 active equity schemes in 21 houses, 2023–2026; no reshuffling of house labels in 1,000 tries produced a gap that large. Modest in size; says nothing about whether the call is right. | — |
| Fund houses value listed shares alike — except at events with no market price. 99.7% of holdings sit within 1% of the NSE close; for the Vedanta demerger entities awaiting listing, houses’ prices differed up to about 3×. | Implied price = disclosed value ÷ shares, per house. | — |
| Value and contra funds mostly do buy recent losers. Some do not: a few funds with “contra” or “dividend yield” in the name lean toward recent winners. | Buying vs the previous six months’ return: value/contra −0.10, other active funds about 0, momentum-named funds +0.23; checked against momentum index funds. | On each fund page (“How it trades”) |
| Balanced advantage funds really do move their equity. As a group they swing their equity after futures hedges about four times as much as fixed-mix hybrid funds. | Typical swing 8.9 vs 2.3 percentage points (13 and 5 funds). Counted only for fund houses whose futures positions our data captures: for 14 houses the hedges are missing from the parsed disclosures, and an earlier version of this page wrongly read several of their funds as barely moving. Reading those houses’ futures straight from their disclosure files, all six we could measure — SBI, Nippon, Parag Parikh, Aditya Birla, LIC and Helios — move their equity like the rest (typical swing 7.4 points); Franklin’s, read from separate columns in its files, swings 6.4 points. The direction of the moves alone could not be told apart from ordinary rebalancing. | On hybrid and dynamic fund pages (“How much is in equity”) |
| Fund houses report the same rating for the same bond — almost always. They differ on 0.1% of AAA and government bond-months, and on 2.9% below AAA, always by one notch. One fund house more often shows the higher of two agencies’ ratings. | 55,680 AAA and 5,477 lower-rated bond-months held by two or more houses. | — |
| Same bond, same yield. When two or more fund houses hold a bond in the same month, 99.9% of their reported yields agree within 0.25 percentage points. The exceptions are floating-rate government bonds and perpetual bank bonds, where “yield” has no single convention. | 124,110 house-bond-months, 2022–2026; median gap 0.00 points. A fund’s headline yield that includes those two kinds of bond is not directly comparable with another house’s. | — |
| Investors do not chase returns. Monthly inflows into equity categories do not rise after strong months; in large-cap funds they rise after falls. | AMFI category flows 2019–2026 against each category’s previous three months; large cap t −3.3 after removing a mechanical SIP effect. | Flows |
| Fund houses owned by a bank or insurer avoid their sponsor’s shares; conglomerate-owned houses lean toward their group. Over the last year, nine houses held none of their sponsor group’s stock in active funds. The Aditya Birla, Tata and Bajaj houses gave their own group’s companies about 2.2×, 1.3× and 1.2× the weight other houses did. | Weight in group stocks vs all other houses, active funds only, 16 houses. We tested for a general tilt toward the sponsor group and it failed; the split between the two kinds of owner was seen afterwards. A fresh test on debt funds confirmed the avoidance (7 of 7 bank- and insurer-owned houses lend little or nothing to their sponsor) but not the lean: none of the 6 group-owned houses overweight their group’s bonds. Being re-tested on the next 12 months. Within SEBI’s limits; not a claim of wrongdoing. | — |
| When fund houses sell out of a stock, others tend to follow the next month. After two or more houses exit a stock in one month, about a third more of the remaining houses exit the month after than in comparable stocks. | +1.6 points (4.1% vs a 3.1% base), matched on month, company size, number of holders and the stock’s own price move; held in 2024–2026 data kept aside. It may be one piece of news reaching houses at different speeds rather than houses copying each other. The share price was tested separately and showed no effect (see below). | Fund houses buying / selling |
What did not hold up
Reported because a test that fails tells you as much as one that passes
- Fund buying as a share-price signal. Stocks funds were buying did not reliably beat stocks they were selling over the month after the holdings were published; an early result faded on later data. The same was true of each manager’s biggest overweight (“best ideas”).
- A fund’s biggest new positions as a share-price signal. Stocks a fund opened big did no better than the ones it opened small over the three months after the holdings were published (−0.6%); an early one-month edge reversed in later data. Opening size tells you how long a fund will keep a stock, not how the stock will do.
- Fund buying of an industry as a signal. Industries that funds were buying, as a group, did no better than other industries in the months after the holdings were published.
- Widely held stocks as a crowded trade. Stocks held by many fund houses did not fall harder in market sell-offs; compared with similar-sized stocks they fell less in falling months (14 of them) and rose less in rising months — steadier, not more fragile.
- Fire sales by funds losing money. Stocks held mostly by funds with large outflows did not fall behind in that quarter (they did slightly better) and showed no reliable rebound afterwards.
- Exit waves as a price signal. Stocks that several fund houses sold out of did no worse than comparable stocks in the months after the holdings were published (+0.3% a month, sign flipping between 2022–24 and 2024–26).
- Cash as market timing. How much cash active equity funds hold (typically 3–5%) said nothing about how the market did the next month; funds hold slightly more cash after falls, not before them.
- Liquid funds stockpiling cash for quarter-end. Money does leave liquid funds at quarter-ends — about ₹94,000 crore more than in the months around them — but the funds do not hold extra cash at those month-ends; they hold less (see open leads).
- Tax-saver (ELSS) funds using their 3-year lock-in. With no redemptions to meet, ELSS funds could hold smaller, less liquid companies and trade less than flexi-cap funds. They do not, measurably: similar small-cap shares (22% vs 19%) and no significant difference in turnover.
- Big funds earning less because they are big. Larger funds did not go on to do worse than smaller funds of the same category over the next year; among small-cap funds the bigger ones did slightly better (holdings-based returns, 2022–2026).
- New money sitting idle as cash. Funds that attract more money do hold more cash — but a fund’s cash does not rise in the quarters money comes in; inflows are invested about as fast as they arrive.
- Window dressing at quarter ends. Funds did not buy the quarter’s winners any more at quarter ends than in other months.
- Selling winners and holding losers. An apparent pattern disappeared once position size was accounted for: managers trim positions that have grown large, not winners as such.
- Active funds converging on the same portfolios. Over 2022–2026 they drifted apart instead: average overlap with peers fell from 22.5% to 20%.
- A fund house’s schemes converging on one portfolio. In the three houses with a full history, sibling schemes became less alike: their average overlap fell from 36% to 24% between 2023 and 2026, for the same schemes throughout.
- Portfolio pumping each month-end. Stocks funds hold heavily did not reliably jump on the last trading day of the month.
- Flipping IPO allotments. Funds sold out of new listings only about 2 points more often within three months than out of other new positions — fewer in the first month, when anchor lock-ins apply.
Open leads, not findings
Too few observations, or found after looking; to be re-tested on new data
- Since 2024, much more money has gone into new equity fund launches after strong markets — but not in 2021–23, so it does not yet count as a pattern. Fund houses did not launch noticeably more funds after rallies; investors put more into the ones launched.
- Liquid funds’ cash share is lower at the quarter-end disclosure than in the months either side in 10 of 11 quarter-ends — as if redemptions are paid out of cash. Seen after testing the opposite; re-tested at each quarter-end.
- Active funds appear to sell stocks just as they join the Nifty 50, when index funds must buy — 9 inclusions so far. The same did not show for 75 promotions into the Nifty Next 50 and Midcap 150, so if real it is specific to the Nifty 50. Re-tested at each March and September rebalance.
- In March, the fiscal year-end, funds leaned toward the month’s losers rather than its winners — 5 Marches so far.
- At quarter-ends, stocks that funds hold most heavily relative to their trading fell back about 0.9% on the first day of the next quarter — the pattern of “window pumping”, seen after looking and over 18 quarter-ends only. Being re-tested on each new quarter-end.
Nothing on this page is investment advice or a forecast of returns. Each study’s full write-up, code and numbers are kept with the project (docs/research).