How Many Stocks Do You Need to Diversify? The Research and a Practical Number

Most of the benefit arrives fast. The classic finding from Evans and Archer (1968) is that a randomly chosen portfolio of roughly 10 to 15 stocks removes the large majority of company-specific risk. Later work, notably Statman (1987), argued the real number is higher once you account for borrowing and costs, at least 30 to 40 names. In practice, 20 to 30 holdings is the range where a self-built index stops being a bet on a few companies, and every name past that adds less than the one before.

The question sounds like it has a single number as an answer, and the reason it does not is that "diversified" is not one thing. You can diversify away company-specific risk with a surprisingly short list. You cannot diversify away market risk at all, no matter how many stocks you own. Everything below is educational content, not investment advice.

How many stocks do you need to diversify?

Around 20 to 30 for a portfolio you build yourself, assuming those names are genuinely spread across sectors. Below about 15, one bad earnings report is a portfolio event. Above about 30, you are mostly adding administration rather than protection, and each additional holding dilutes your best ideas without removing much remaining risk.

The academic range exists because different studies asked slightly different questions.

The drift upward over time is not academics disagreeing for its own sake. Individual stock outcomes have become more dispersed, and a portfolio of 12 names today carries more idiosyncratic risk than a portfolio of 12 names did in 1968. If you want one number to work from, use 25 and adjust for how correlated your holdings actually are.

Why the number is not the whole story

Thirty stocks that are all regional banks is not a diversified portfolio, it is a bet on regional banks with thirty tickers attached. The count only measures diversification if the holdings are genuinely doing different things. Two questions matter more than the number:

How correlated are they? If everything in the list rises and falls together, you have one position expressed thirty ways. The whole mechanism of diversification is that some holdings zig while others zag, and correlated names cannot do that.

How is it weighted? This is the part people miss most often. A 30 stock portfolio where one name is 40% of the money is not a 30 stock portfolio in any risk sense. It is a concentrated position with 29 rounding errors attached, and it will behave like the big holding. The count is meaningless without the weights, which is why equal weighting is the only scheme that makes the number mean what people assume it means.

What you cannot diversify away

Total risk splits into two parts. Unsystematic risk is company-specific: a failed drug trial, a fraud, a lost contract. That is the part more holdings remove, and it disappears quickly, which is the Evans and Archer result. Systematic risk is the market itself: recessions, rate shocks, liquidity crises. Adding a five hundredth stock does nothing about it, because in a real crisis correlations converge toward one and the diversification you thought you had temporarily stops working.

This matters practically because it caps what more names can buy you. Past 30 or so, additional holdings are mostly removing risk you already removed. If you want to reduce risk further at that point, the lever is asset classes rather than tickers.

Can you over-diversify?

Yes, though it is a milder problem than under-diversifying. Holding 100 hand-picked stocks means you have effectively built a slightly worse index fund with far more work, and you can buy an actual index fund for a few basis points. The more useful framing is that every position past the point of diminishing returns dilutes your conviction without buying safety. If you did the work to identify twelve companies you genuinely believe in, padding to sixty with names you have not researched is not risk management.

The practical constraint is attention. Every holding is something you are implicitly claiming to understand, and it is worth being able to pull a structured research view on each ticker before it goes in, rather than adding names to hit a count. A list you cannot explain is not diversified, it is just long.

How many holdings should a custom index have?

If you are building your own index rather than picking stocks, the answer shifts slightly upward, because an index is meant to represent something rather than express a handful of specific views. A sector or thematic index usually needs 15 to 30 constituents to actually represent the theme instead of tracking its two largest members. A broad market replacement needs considerably more.

Two design decisions do more work than the count itself. The first is the weighting scheme: cap weighting a 25 stock theme index will often leave one incumbent at 30% or more, which quietly undoes the breadth you built. The second is the rebalancing rule, because whatever spread you set at the start erodes as prices move. Both are covered in index construction, and the reset side specifically in portfolio rebalancing.

How to check whether your portfolio is actually diversified

Three checks, in order of how much they tell you.

Look at the top ten weight. If your ten largest positions are more than half the portfolio, the count is not describing your risk. For reference, in July 2026 the ten largest holdings were 36.33% of a cap weighted S&P 500 fund and 2.50% of the equal weighted version of the same 500 companies, which is a useful illustration of how far apart two portfolios with identical holdings can be.

Look at sector concentration. If more than about a third of the money sits in one sector, that sector is your portfolio. Sometimes that is the intent, and it should at least be deliberate.

Then test it. Build the list, run it against real market history, and look at the worst drawdown and at how much of the result came from the top two or three names. A backtest where removing one holding changes the outcome dramatically has told you something important about concentration that no headline count would have. You can backtest the portfolio under different holding counts and weighting rules to see where your own curve flattens out.

Indexes is educational and informational software. It never places trades, connects to a brokerage or holds assets, and nothing here is investment advice. Backtests are hypothetical and past results do not predict future returns.