What is the Sensex?

The Sensex is quoted as though it were a place. The Sensex fell today; the Sensex is at a record. The grammar smuggles in a claim that is not true: that the number is the market, observed. It is not. It is the output of a calculation performed on a deliberately chosen sample, and someone chose the sample, chose the weighting, and wrote down what happens when a company is swapped out. Nothing about the Indian equity market obliges a benchmark to hold exactly 30 names, or to weight them by the shares that happen to be tradeable. Those are decisions. The number is downstream of the decisions, and if you do not know the decisions you cannot read the number.

Start with who makes them, because this is where most published descriptions are now simply out of date. The index is compiled and maintained by BSE Index Services Private Limited, which was formerly named Asia Index Private Limited. That company spent years as a joint venture between the BSE and an international index provider, which is why an entire generation of articles, data feeds and textbooks call the index the "S&P BSE Sensex". The BSE bought out its partner's entire stake in 2024, the joint venture ended, and the index provider became a wholly owned BSE subsidiary. BSE Index Services' own current methodology documents call it the BSE SENSEX throughout, with no S&P prefix anywhere in them. The co-branded name is a historical artefact, and a page still using it is telling you when it was written.

The rest of the specification is short, and worth having in one place before any of the mechanics make sense. Note two entries in particular. The base period is 1978 to 1979 but the index did not launch until 2 January 1986, and its first value date is 3 April 1979. Read those three dates together and they say something the headline never does: the first seven years of the famous long-run chart were computed backwards after the index existed, not published as the market traded. That is not a scandal, it is ordinary index practice, but it is the first hint that the chart is a construction rather than a recording.

The name itself is a small piece of the same story. "Sensex" is a contraction of Sensitive Index, coined in 1989, three years after the thing it names began publishing. Even the word is a retrofit.

Every constituent enters the calculation at its free float market capitalisation: its price, multiplied by its shares outstanding, multiplied by an investable weight factor. That third term is the whole idea. BSE Index Services defines the investable weight factor as available float shares divided by total shares outstanding, and defines available float as the total less the shares held by strategic holders. In practice the excluded blocks are the ones you would expect and a couple you would not: promoter and promoter group holdings, shares held by government agencies, shares held by other listed companies, strategic stakes sitting under the public category of the shareholding pattern, promoter-held depository receipts, and shares under a lock-in. What survives that subtraction is the float, and only the float counts. If you want the underlying idea of market capitalisation itself before going further, it is set out in the market capitalisation guide.

The principle is one line: shares that cannot be bought should not vote. A promoter family's controlling block is not for sale on any given Tuesday. A government's holding in a state-owned major is not there for economic reasons and is not going to be sold to you. Counting those shares in the index would hand a company influence over the number in proportion to a size the market can never actually trade, and would set every tracking fund an impossible task: replicate a weight built out of stock that is not available. Free float weighting makes the index describe investable value rather than nominal size, which is why the Sensex moved to it on 1 September 2003 and why every serious benchmark now works the same way.

The consequence is larger than it sounds, and it is the reason the figure below is worth reading slowly rather than skimming. Free float weighting does not merely shave a few percentage points off tightly held companies. It re-orders them. The arithmetic is done here on six companies rather than thirty, purely so that every step stays legible, but the rule is identical to the one the index applies across its full list.

One practical note follows from this, and it explains an event that otherwise looks arbitrary. Because the investable weight factor is an input, a company's index weight can move sharply without its share price moving at all. If a promoter sells a slice of a controlling block, or a government pares its stake, the float rises, the investable weight factor rises, and the company's weight in the index rises with it. BSE Index Services reviews these factors quarterly and applies changes at the open on the Monday after the third Friday of March, June, September and December, with an ad hoc route when a change is worth at least five percent of shares outstanding. Index funds then have to buy more of a company whose business did nothing that quarter. The demand was manufactured by a rule, and knowing that the rule exists is the difference between watching the flow and being surprised by it.

Add up the free float market capitalisation of 30 companies and you get a number in the region of tens of lakh crore of rupees. That is not an index, it is a total, and it is useless for comparison because it lurches whenever the basket's contents change rather than its prices. The Sensex divides that total by an index divisor, and the divisor is the single most important and least discussed object in the whole machine. BSE Index Services describes it as the only link back to the original base period value. Everything the index claims about 1979 rests on it.

The rule governing it is precise, and once you have it you can answer every corporate-action question yourself without looking anything up. BSE Index Services states the test as follows: any change to the stocks in the index that alters the total market value of the index while holding stock prices constant requires a divisor adjustment. Read that clause carefully, because the phrase "while holding stock prices constant" is doing all the work. It is a filter that separates two kinds of event. A genuine price move changes the total through prices, and the index is supposed to register it. A structural event changes the total without prices, and the index must not register it, so the divisor is re-based by the same proportion and the ratio is left exactly where it was.

Apply that filter to a share split and you get an answer that surprises almost everyone, including a great many published explanations of this index. A split doubles the share count and halves the price. The company's market capitalisation is unchanged, so the basket's total is unchanged, so there is nothing for the divisor to cancel and no adjustment is made. BSE Index Services' corporate action policy for market capitalisation indices says so directly: a split adjusts shares and price by the same ratio, producing no change to the index market capitalisation and no divisor adjustment. The same policy records that a bonus issue is the same action quoted differently, and so gets the same treatment. If you have read that the Sensex divisor is re-based on every bonus and split, you have read something that is true of a price-weighted index, such as the old American averages, and false of this one. It is an easy error because both statements sound like index plumbing, and the distinction only appears when you ask what the weighting scheme actually is.

The full list falls out of the same test. Divisor adjustment: a constituent added or deleted, a change in shares outstanding, a change in the investable weight factor, an in-the-money rights offering, a special dividend. No divisor adjustment: a split or reverse split, a bonus issue, an ordinary cash dividend. That last one is quietly important and we will return to it, because an ordinary dividend gets no price adjustment and no divisor adjustment, which means the headline Sensex simply does not see it. The dividend leaves the company, the share price adjusts in the market, and the index records that as a fall. There is a separate total return version of the index that reinvests dividends; the number on the news is not it.

The divisor is the index's memory. It carries the 1979 scale forward through every structural change, so that the only thing left able to move the number is the thing the number is for.

Here is where the rule stops being neutral plumbing and starts shaping what the number can honestly say. Weight is proportional to free float market capitalisation. Nothing in the method distributes influence evenly, and nothing caps it. So the index is not thirty companies each with a thirtieth of a say. It is thirty companies with wildly unequal says, and the inequality is not a defect that crept in; it is the direct output of the weighting rule working as designed.

The figure below computes the weights from a set of free float market capitalisations rather than asserting them, so the shape of the ladder is an output. The distribution used is illustrative, but it is deliberately built to match how a free float weighted index of thirty large names actually distributes, and the pattern it produces is the point.

Two things follow immediately. The first is that the Sensex is a large cap instrument and only that. It has no mid caps, no small caps, and no exposure to the parts of the market where most listed companies actually live. The second is subtler and more often missed: because a few names dominate, the Sensex carries real single-name risk. A benchmark that people describe as "the market", and treat as the safest possible thing to own, has a material fraction of itself sitting in a small number of companies and in a small number of sectors. Diversification is not the same as being a benchmark, and the word "index" does not by itself confer it.

Concentration is an abstraction until you decompose a single session, at which point it becomes the most practical thing on this page. The move in the index is the sum, across the thirty, of each constituent's weight multiplied by its own return. That is the whole calculation. And a sum of thirty terms can come out positive while twenty-four of the terms are negative, provided the positive ones are attached to large enough weights. This is not a hypothetical pathology. It is an ordinary Tuesday, and it is what a narrow rally is.

The figure below works one such session all the way through, using the weights computed in the previous figure and applying a set of one-day returns to them. The bars are the contributions, and they sum to the index move exactly, because that is what the index move is.

This is the single most useful habit to build around any benchmark, and it costs nothing: when the index moves, ask how many of its constituents moved with it. That ratio is called breadth, and it is the difference between an advance that a market is participating in and an advance that three companies are carrying. The two look identical on the headline and are completely different facts about the world. The number in the news is a summary statistic, and a summary statistic loses information by construction. Knowing precisely which information it threw away is the entire skill.

If the index is a rule, the selection criteria are where the rule is most explicit and most revealing. They are also more demanding, and considerably more specific, than the usual description of "large, well-known, blue-chip companies" suggests. The eligible universe is not the exchange; it is the BSE 100. A candidate must have at least six months of listing history at the BSE, must have traded on every single trading day of the six-month reference period, and must have a derivative contract. That last criterion is rarely quoted and quietly consequential: a company without a listed derivative is not eligible for the Sensex no matter how large it is, which ties membership of the country's oldest index to the existence of a futures and options market on the name.

The step worth pausing on is the incumbency buffer. The top 21 seats go on rank alone. The remaining nine are offered first to companies already in the index that rank anywhere from 22nd to 39th, and only if seats remain do outsiders ranked 22nd to 30th get considered. The effect is a deliberate hysteresis band: a sitting constituent can drift as low as 39th and keep its seat, while a challenger sitting at 22nd waits outside. This is not favouritism, it is engineering. Without a buffer the index would swap names every time two companies crossed on a noisy six-month average, and every fund tracking it would have to trade the churn at its own cost. The rule is written to be stable rather than maximally accurate, which is a real trade-off, honestly made and worth knowing about. It also means the constituent list lags the market slightly by design, which is not a bug you should expect the index to fix.

Notice what is nowhere in the cascade: valuation. Nothing in it asks whether a company is cheap, whether its earnings are growing, or whether its price makes any sense. The screens are size, liquidity, tradeability and continuity of trading. A company can be egregiously overvalued and sail through every one of them, and by construction a company enters the index after it has already grown large, which is to say after the growth. Membership certifies that a business is big and liquid. It certifies nothing whatsoever about the price. Those are separate questions and conflating them is the most expensive mistake available here. Where the Sensex sits in the wider size spectrum, and what the mid and small cap segments do differently, is covered in the guide to large, mid and small cap stocks.

The Sensex is reconstituted semi-annually. Changes take effect at the open on the Monday following the third Friday of June and December, decided from data as at the last trading day of April and October. Twice a year the cascade in the previous section runs again, and the list it produces is the list. Companies that have faded out of the top ranks are removed. Companies that have grown into them are added. This has been happening, twice a year, for four decades.

Now put that fact next to the divisor and see what it does to the chart. Every one of those swaps re-based the divisor so the published level would not jump. That is the correct treatment, and it has an implication almost nobody states out loud. The line is continuous across a set of changes that were not continuous at all. The index in 1986 and the index today are, in the ordinary sense, not the same object: only a handful of the original thirty are still in it. The number is comparable across that gulf only because the divisor was engineered to make it comparable. The right mental model is not thirty companies with a long history. It is thirty slots, maintained by a rule, with companies flowing through them.

This is survivorship, and it is worth naming precisely because the innocent reading is so natural. When someone says the Sensex has risen roughly six hundredfold from its base, that is a true statement about the index. It is not a statement about any portfolio a human being could have held. A person who bought the original thirty in 1986 and never traded again did not get that line; they got a portfolio containing textile mills, and vacuum flask makers, and companies that were merged out from under them, with no mechanism to sell the faders and buy the risers. The index had that mechanism. The index is that mechanism. Its long-run return includes, and cannot be separated from, the effect of a committee removing losers and adding winners twice a year for forty years, by rule.

None of that makes the index dishonest, and it is important not to overcorrect into cynicism. Pruning is what makes a benchmark stay representative; an index that never changed its list would after four decades be measuring an economy that no longer exists. The rule is doing its job. But the rule is doing a job, and the job includes systematic replacement. So the correct sentence is not "the top 30 Indian companies returned X since 1986". It is "a rule that holds 30 large, liquid, derivative-linked BSE 100 names, and refreshes them twice a year, returned X since 1986". The second sentence is longer, and it is the one that is true.

The two are close cousins and are constantly confused, so it is worth fixing the structural differences in one place and then leaving them alone. They are both free float market capitalisation weighted large cap benchmarks, both run a divisor for exactly the reasons set out above, and both are maintained by an exchange-owned index company. They move together because they share most of their heavyweight names, which is another way of saying that the concentration described earlier is common to both. The genuine differences are the exchange, the count, the base, and the age.

Holding 30 names rather than 50 is the difference that matters most, and it cuts in a direction people do not expect. Fewer names means more concentration, so everything in the two figures above bites slightly harder on the Sensex than on its larger sibling: a given heavyweight is a bigger share of 30 than of 50, and a narrow rally shows up more strongly. That is the structural trade, and it is the only one worth carrying away. Why the derivatives liquidity gathered on the other benchmark, and how a stricter impact-cost gate shapes its list, is properly the subject of the Nifty 50 guide rather than this one.

Everything above converges on one practical question, because it is the question almost every reader actually arrived with: the Sensex is up, so how am I doing? The honest answer is that the index has no opinion about you, and the reasons are now all on the table rather than being a matter of taste. It holds 30 large BSE 100 names, weighted so that five of them outweigh twenty. It excludes mid caps and small caps entirely. It refreshes its list twice a year by a rule that systematically drops faders. And the headline version is a price index, which means ordinary cash dividends get no price adjustment and no divisor adjustment and are therefore invisible to it: the dividend leaves the company, the share price falls, and the index records the fall and not the payment. There is a total return version that reinvests them. It is not the number on the news.

So what is it good for? A great deal, once it is read as what it is. It is a fast, reliable, four-decade-consistent read on how the largest and most liquid slice of the Indian market is being priced. Its continuity is genuinely hard engineering and genuinely valuable: very few economic series survive forty years of structural change with their scale intact. It is the reference against which large cap exposure is priced and sold, and if you hold that exposure through a fund, the tracking of this index is the product you have bought; the mechanics and costs of that route are set out in the guide to ETF investing in India. It is a serious instrument. It is simply an instrument with a specification, and the specification is the fine print.

1. Read it as a rule, not a place. "The Sensex rose" means a weighted sum of 30 chosen numbers went up. It does not mean the market rose, and the two come apart more often than the phrasing admits.
2. Ask for breadth before you believe a move. How many of the 30 actually went the same way as the number? A move carried by three heavyweights and a move the whole list participated in are different facts wearing the same headline.
3. Do not read a weight as a verdict. A company's weight is a statement about how much of it is available to trade, not about how good it is or what it is worth. The selection cascade never once asks about valuation.
4. Treat a corporate action as a divisor question. Does this change the basket's value with prices held constant? If yes, the divisor moves and the level does not. If no, as with a split or a bonus, nothing happens at all.
5. Never quote the long-run chart as a portfolio. It is the track record of a rule that replaced its losers twice a year for forty years. No holder of the 1986 list ever earned that line.
6. Match the benchmark to what you hold, or drop the comparison. A large cap price index is the wrong yardstick for most portfolios, and the wrong yardstick produces confident, wrong conclusions.

That habit of opening up a number before trusting it, of asking what rule produced it and what the rule threw away, is not a trick specific to indices. It is the whole of the method we teach, applied here to the most quoted figure in the country. The Sensex is an unusually good place to practise it, because it is a rule that is fully published, entirely knowable, and almost universally misread.

The Sensex answers one question precisely: how the free float value of thirty large BSE companies is moving against a 1978 base. It answers nothing else, and it never claimed to.

• BSE Indices Methodology (BSE Index Services Pvt. Ltd., June 2026). The primary specification. Defines the BSE SENSEX as 30 large, well established and financially sound companies across key sectors, drawn from a BSE 100 eligible universe, weighted by float adjusted market capitalisation. Sets out the eligibility factors (six months of listing history, traded on every trading day of the six month reference period, a derivative contract) and the full selection cascade including the top 21 rule and the rank 22 to 39 buffer for existing constituents. Its Base Dates table records the launch date as 2 January 1986, the first value date as 3 April 1979, the base date as 1978-79 and the base value as 100. Its Index Maintenance table sets the SENSEX rebalancing frequency as semi-annual, effective the Monday following the third Friday of June and December, with reference dates on the last trading day of April and October. bseindices.com
• Index Mathematics Methodology (BSE Index Services Pvt. Ltd., August 2025). Establishes that the index level is the index market value divided by the divisor, that the divisor is the only link to the original base period value, and the governing test quoted in this guide: any change to the stocks in the index that alters the total market value of the index while holding stock prices constant will require a divisor adjustment. Gives both the multiplicative and the additive form of the adjustment, the latter as new divisor equals old divisor plus the change in market value divided by the index level. bseindices.com
• Equity Indices Policies & Practices Methodology (BSE Index Services Pvt. Ltd., August 2025). Carries the corporate action treatment table for market capitalisation indices used in this guide. States that a split or reverse split adjusts shares outstanding and price by the split ratio, producing no change to the index market capitalisation and no divisor adjustment, and that company additions and deletions, changes in shares outstanding, changes in the investable weight factor, special dividends and in-the-money rights offerings each cause a divisor adjustment. Also defines a stock split as leaving the market capitalisation of the company unchanged before and after the event, and records that stock splits, stock dividends and bonus issues imply the same action. bseindices.com
• Float Adjustment Methodology (BSE Index Services Pvt. Ltd., August 2025). Defines the investable weight factor as available float shares divided by total shares outstanding, and float adjustment as excluding shares held by other publicly traded companies, government agencies and certain strategic shareholders. Lists the non-free-float categories: promoter and promoter group, public locked-in shares, promoter GDR holdings, and strategic and other holdings appearing under the public categories of the shareholding pattern. Identifies the publisher as BSE Index Services Pvt. Ltd., formerly Asia Index Pvt. Ltd., and sets the quarterly investable weight factor review schedule. bseindices.com
• BSE completes acquisition of S&P Dow Jones Indices' entire stake in Asia Index (Business Standard, 31 May 2024). Records the end of the BSE and S&P Dow Jones Indices joint venture and the index provider becoming a wholly owned BSE subsidiary, which is the corporate event behind the retirement of the S&P BSE prefix. business-standard.com
• BSE SENSEX, index overview and history. Supporting reference for the migration from full market capitalisation to free float weighting with effect from 1 September 2003, for the naming of the index as a contraction of Sensitive Index, and for the turnover of the original 1986 constituent list. en.wikipedia.org