🧺 Diversification and correlation: how many eggs, how many baskets

🎯 In 30 seconds: Diversification means spreading money across things that do not all move together, so one bad outcome cannot sink you. The key measure is correlation: how closely two prices move together, from -1 to +1. Owning five lenders that move as one is not diversified. In India that trap is easy to fall into, because financial services make up more than a third of the Nifty 50 and many listed companies belong to the same business groups. Adding positions cuts risk quickly at first, then hardly at all, and in a sell-off correlations tend to rise just when you need them low. Traders manage this by capping the total risk of all open trades, called portfolio heat.

The farmer's lesson

A farmer who plants only one crop loses the season if a pest attacks it. A farmer who plants three crops loses one field to the pest and harvests the other two. That is diversification.

But if the monsoon fails, all three fields dry up together. Spreading across crops does nothing against a risk that hits every crop at once.

Shares work the same way. Diversification reduces the risk that comes from any single company, called company-specific or unsystematic risk: a fraud, a plant shutdown, a bad set of results, a regulatory order. It cannot remove market risk, also called systematic risk: the chance that most stocks fall together because of a slowdown, rising interest rates, heavy foreign selling or a global panic.

That second kind is why diversification helps a lot but never makes you safe.

Correlation in plain words

Correlation is a number from -1 to +1 that describes how two things move together.

• +1: they always move in the same direction, in proportion. Owning both is like owning more of one.
• 0: no consistent relationship. One tells you nothing about the other.
• -1: they always move in opposite directions. Gains in one offset losses in the other.

Most pairs of Indian stocks have a positive correlation, because they all respond to the same economy, the same RBI decisions and the same flows of foreign and domestic money. Two companies in the same industry usually have a higher correlation than two in unrelated industries.

A quick hypothetical example. Here are six months of returns for two made-up private banks:

They go up and down in the same months. The correlation works out to about 0.87, which is high. Owning both gives you very little protection if lenders have a bad month, for example after a rise in bad loans or a surprise rate decision.

Correlation is calculated from past data and it changes over time. Six months is far too short to rely on. Real estimates use years of data, and even those shift.

How much does adding positions help?

Here is the core math, using round hypothetical numbers. Suppose each stock has a volatility of 30% a year (a measure of how much it typically swings) and you split money equally.

Two stocks, 50/50:

Many stocks. If every pair has the same correlation, portfolio volatility is:

portfolio volatility = stock volatility x square root of (1/n + (1 - 1/n) x correlation)

where n is the number of equal-sized positions.

Two lessons. First, going from 1 stock to 10 cuts volatility a lot. Going from 20 to 50 barely changes it. Second, correlation sets a floor you cannot diversify below. Stocks that move together leave you with most of the risk no matter how many you own.

India's built-in concentration

Even the main index is not as spread out as "50 companies" sounds. In the Nifty 50 factsheet dated 30 September 2026, the Financial Services sector (banks, NBFCs, insurers and similar) had a weight of 37.45%. The next largest, Oil, Gas and Consumable Fuels, was 9.39%, and Information Technology was 7.52%. The ten largest companies together made up a little over half the index.

These weights change with prices and at each six-monthly rebalance, so check the latest factsheet on the NSE Indices website. The shape has been similar for years: Indian large caps lean heavily on lenders. (Why the index is built this way is in Market indexes.)

So if you hold a Nifty 50 index fund and trade a few bank stocks and own a banking sector fund, a large share of your money is riding on one sector, however many line items your demat statement shows.

Concentration risk and hidden overlap

Concentration risk is having too much riding on one thing. It is obvious when one stock is 40% of your account. It is less obvious when several positions are really the same bet.

• Sector overlap. Two private banks, a housing finance company and a gold-loan NBFC are four tickers, but largely one bet on credit growth, bad loans and interest rates. Three IT services exporters are largely one bet on overseas tech spending and the rupee. See Sectors and the business cycle.
• Business-group overlap. Many Indian business groups have several listed companies under the same promoter family. They can share lenders, cross-holdings, management and reputation. Bad news about the group, from a governance question to a debt problem, often hits every listed company in it at once. A hypothetical "Malhotra group" cement company, power company and finance company are three sectors on paper and one promoter in practice.
• Theme overlap. Railway, defence and PSU stocks, or a basket of recent SME listings, can rise and fall together on the same story, even across sectors. A factor, such as company size or momentum, can link stocks in the same way.
• Fund overlap. A Nifty 50 index fund plus a large-cap fund holds many of the same companies twice. Check each fund's monthly portfolio disclosure. See Mutual funds, index funds and ETFs.
• Life overlap. If you work at an IT services company and also hold its shares through ESOPs, plus two other IT stocks, your salary and your savings ride on the same industry.

Business-group risk is not theoretical. After a US short seller published a report on the Adani group on 24 January 2023, the group's listed companies together lost more than $100 billion of market value by 2 February 2023, CNBC reported. The group denied the allegations, and Reuters later reported that the shares had recovered much of the losses. The point here is not who was right. It is that many listed companies across different sectors fell together on news about one group.

Correlations rise in a sell-off

Diversification works least well exactly when it is needed most. Ang and Chen, in the Journal of Financial Economics (2002), found that correlations between US stocks and the overall market are much higher during market declines, especially extreme ones, than during rises.

That study used US data. The mechanism is not US-specific, though. In a panic, large investors who need cash sell what they can, which is usually the liquid large caps across every sector. Traders facing margin calls sell their best holdings to cover their worst. Both push otherwise unrelated stocks down on the same days.

Run the formula above with ten stocks: at correlation 0.3, volatility is 18.2%. If correlation jumps to 0.9 in a panic, it is 28.6%, almost as risky as a single stock.

So plan for sell-off correlation, not calm-day correlation. Keep total risk modest. Money that is not in shares, such as cash and bank deposits, does not fall when share prices fall, and government securities are driven mostly by interest rates rather than company news. See Bonds, yields and stocks.

Portfolio heat: total open risk

Portfolio heat is the sum of the risk on all open trades: what you would lose if every stop were hit at once. Each trade's risk is shares x (entry - stop), or zero once the stop is at or above your entry, plus the charges to get out.

A hypothetical ₹2,00,000 account with five open swing trades, each sized for about 1%:

The positions use about ₹1,13,000 of the account, but the total planned risk is under 5%. That is a common ceiling: many traders cap heat somewhere around 5% to 6% and do not open new trades above it.

Add a second rule for correlated groups, for example "no more than 2% total risk in one sector, theme or business group." In the table, the three lenders carry ₹5,508 of price risk plus charges, close to 3%, so this trader would skip the next bank or NBFC setup until one of these closes or its stop moves up.

How many positions?

There is no magic number. For long-term investors, a broad index fund holds 50 to 500 companies in one purchase, which is why many people use one as the core of their money (see A plan for a young person). Broader indices spread money across more companies, but check their factsheets too: sector tilts do not disappear just because the company count is higher.

For an active trading account, the limit is usually attention, not math. Each position needs a stop that you re-enter or maintain, a plan and some monitoring. Many beginners do better with three to six open trades than with fifteen they cannot follow. Use the heat and sector caps to control risk, and the table above to see that a handful of truly different positions gets you a large share of the benefit.

Try it

Download the latest Nifty 50 factsheet from the NSE Indices website and write down the top three sector weights. Then list every position you hold or would like to hold, including mutual funds. Next to each, write its sector, its business group if it belongs to one, and its open risk in rupees. Add up total heat and the largest sector total. Finally, open the portfolio disclosure of any equity fund you own and count how many of its top ten holdings are also in your own list.

Sources

• NSE Indices: Nifty 50 factsheet (30 September 2026)
• CNBC (2 February 2023): Adani losses top $100 billion in wake of Hindenburg Research report
• ThePrint, from Reuters (August 2024): Adani group shares lose up to $13.4 billion after Hindenburg accusations against regulator
• Ang and Chen, Journal of Financial Economics (2002): Asymmetric correlations of equity portfolios
• NSE: SEBI turnover fees, STT and other levies