Effective Number of Holdings: A Concentration Check

A trader opens a brokerage statement and counts twenty tickers. Twenty names feels diversified. Then two of those positions have run so far that together they carry roughly half the account, and the other eighteen split what’s left. The statement still reads twenty. The risk reads closer to a handful.

That distance between how many positions you hold and how many actually drive the outcome is what the effective number of holdings is built to measure. It’s a single figure that turns a set of unequal weights into the number of equally weighted positions your portfolio behaves like. Before I trust a headline count, this is the number I reach for.

What a holdings count hides

A holdings count answers one question: how many separate securities are in the account. It says nothing about how the money’s split among them. Twenty equally sized positions and twenty positions where one name owns 40 percent of the capital score the same on a count. They aren’t the same portfolio.

The intuition is worth sitting with before any arithmetic. Diversification isn’t a headcount. It’s a statement about how much any single position can move the whole. When one or two names dominate the weight, the account effectively rides on those names, and the long tail of small positions is closer to rounding error than real exposure. A count treats a 40 percent position and a 2 percent position as one unit each. Your equity curve doesn’t.

Calculating the Effective Number of Holdings

The common weight-based version is simple to compute by hand. Take each position’s weight as a decimal fraction of the portfolio, square it, add the squares together, then divide one by that sum. In shorthand, the effective number of holdings equals one divided by the sum of the squared weights. Readers who know their concentration metrics will recognise this as the inverse of the Herfindahl index applied to portfolio weights.

A worked comparison makes it concrete. Start with an equally weighted ten-stock portfolio. Each position holds 0.10 of the capital. Each squared weight is 0.01, and ten of them sum to 0.10. One divided by 0.10 is 10. The effective number of holdings is exactly ten, which matches the count, because the weights are perfectly even.

Now tilt it. Keep ten names, but let one position grow to 0.55 of the book while the remaining nine hold about 0.05 each. The squared weights are 0.3025 for the large position and 0.0025 for each of the nine small ones, which adds to 0.325. One divided by 0.325 is about 3.08. Same ten tickers on the statement, but the portfolio now behaves like roughly three equally weighted positions.

When I ran this on a book I’d have described as ten positions, one name near 55 percent pulled the effective figure down to about three. The label said ten. The behaviour was closer to three. That single calculation changed how I read the account, and it’s why I check the number whenever a winner runs and quietly takes over the weight. The weights themselves come out of whatever position sizing methods you use, so a concentrated effective count is often a downstream symptom of how the entries were sized, not a separate problem to solve on its own.

Why a 500-name index can behave like a few dozen

The same arithmetic explains a fact that surprises people about broad indices. An index can list hundreds of constituents and still carry a low effective number of holdings, because a small group at the top holds much of the weight. Constituent count and effective count are different measurements, and a capitalisation-weighted index pulls them far apart.

Take a simplified fifty-name index where the top five names each hold 0.10 of the weight, so half the index sits in five positions, and the other forty-five names split the remaining half at about 0.011 each. The five large squared weights contribute 0.05, the forty-five small ones contribute roughly 0.006, and the sum is about 0.056. One divided by 0.056 is close to 18. Fifty listed names, but the index behaves like eighteen equally weighted holdings. Scale that logic to a real large-cap benchmark where the top handful of companies carry an outsized share, and a list of five hundred names can behave like a few dozen.

That’s where the metric earns its keep against a naive reading of breadth. A trader comparing an equally weighted version of an index to its cap-weighted version is really comparing two very different effective counts, even though the two versions hold the exact same list of companies. The way the index weighting methods assign weight decides how concentrated the exposure actually is, long before any single stock moves.

What the measure quietly ignores

Here’s the trap I want you to sidestep. A comfortable effective number of holdings can still sit on top of a badly concentrated set of risks, because the metric sees weights and nothing else. It doesn’t know what the securities are, only how much of the book each one represents.

Concretely, the effective number of holdings is silent on correlation. Fifteen names that all move together during a selloff score exactly the same as fifteen genuinely independent bets, as long as the weights match. It’s silent on shared factor exposure, so a portfolio of eighteen names that are all high-beta growth stocks looks as spread out as eighteen names across different beta and idiosyncratic risk profiles. It ignores overlapping business models, so three suppliers to the same end customer count as three, even when a single demand shock hits all of them. And it ignores liquidity, treating a thin small-cap position and a mega-cap position as interchangeable units of weight.

The most expensive blind spot shows up under stress. Correlations that look tame in calm markets tend to converge toward one when everything sells off at once, so nominally separate securities move as a block precisely when the diversification is supposed to help. A weight-based count can’t see that convergence coming. If you want the exposure the metric misses, you’ve got to look at how positions co-move in the tail, which is the domain of measures like tail dependence rather than a simple weight sum.

A higher number is not automatically better

It’s tempting to treat the effective number of holdings as a score to maximise, where a bigger figure means a safer portfolio. That reading goes too far. The metric describes weight dispersion and stops there. It says nothing about expected returns, nothing about the correlations just discussed, nothing about transaction costs, and nothing about whether any of it suits a particular trader’s plan.

Push the effective count higher by spreading capital across many small positions, and you’ll raise the number while adding names that all rise and fall together, which buys the appearance of diversification without the substance. You also pay for it. More positions mean more commissions, more spread paid, and more monitoring, and that drag’s real whether or not the added names reduce risk. A trend follower running a deliberately concentrated book of a few strong names is making a defensible choice, not failing a test, and the Nassim Taleb trading lessons on fragility and hidden correlation are a useful counterweight to the assumption that more names always means more safety.

Using it as an audit, not a scorecard

The honest way to use this metric is as a portfolio-audit question, not a grade. I compute the effective number of holdings, set it next to the raw count, and read the gap. When the two are close, the weights are reasonably even and the headline count means what it says. When the effective figure sits far below the count, a small group of positions owns the outcome, and I’d rather know that on purpose than discover it in a drawdown. In my own review I keep the two numbers side by side, and any time the effective figure drops below half the name count, I treat it as a flag to check what’s really carrying the book.

The gap’s a prompt to ask better questions, not an instruction to act. Is the concentration intentional, the result of letting a winner run, or accidental drift from an entry that was sized too large? Do the dominant names share a sector, a factor, or a customer, so the real exposure’s even narrower than the number suggests? Those answers live outside the formula. The effective number of holdings just makes sure the question gets asked, by turning a vague sense that a position has gotten big into a figure you can track over time.

Keep it in its lane and it stays useful. It tests whether your stated number of positions reflects your actual weight concentration, and it does that one job well. Learn the pattern. Ride the trend. Keep the gains.

Educational content only. Not investment advice. Trading involves risk. You are responsible for your decisions.