Calculate Portfolio Diversification via Herfindahl Index

That is the practical use case behind how to check calculate portfolio diversification via Herfindahl index. HHI reduces the allocation map to one number: the sum of squared position weights. It does not price volatility. It does not estimate beta. It does not solve correlation. It measures capital concentration with mechanical precision.

Adapting the Herfindahl-Hirschman Index for investment portfolios

The Herfindahl-Hirschman Index was designed for market concentration analysis. In portfolio work, the same arithmetic applies. Replace corporate market share with portfolio weight. Square each weight. Sum the squares.

The formula is:

HHI = Σ(wi²)

Where wi is the weight of asset i.

Two scales are used:

The percentage scale is more common in portfolio diagnostics because the thresholds are easier to read. A one-asset portfolio scores 10,000. A two-asset 50/50 portfolio scores 5,000. A five-asset equal-weight portfolio scores 2,000. A ten-asset equal-weight portfolio scores 1,000. A twenty-asset equal-weight portfolio scores 500.

That minimum is important. HHI is bounded by the number of holdings. A portfolio with four positions cannot score below 2,500 if weights sum to 100%. A portfolio with eight positions cannot score below 1,250. The index penalizes concentration through squaring. A 40% position contributes 1,600 points alone. Four 10% positions contribute 400 points in total.

HHI is not a diversification label. It is a concentration statistic with a hard mathematical floor.

For global stock indexes, this distinction matters. An index can hold hundreds of constituents and still carry a high effective concentration if the largest constituents dominate capitalization weight. A cap-weighted equity benchmark with five mega-cap positions at 25% combined is not equivalent to an equal-weight basket with the same constituent count. HHI detects that difference directly.

Step-by-step calculation: from asset weights to HHI scores

The calculation has no modeling layer. It uses current portfolio weights. Market value weights are preferable to purchase-cost weights because concentration is a live exposure, not a historical allocation record.

A basic sequence:

1. Calculate each position’s current market value.

Use position quantity multiplied by current price. Include cash if cash is part of the risk allocation. Excluding cash raises the relative weight of risk assets. Including cash lowers total HHI if cash is a distinct allocation bucket.

2. Divide each position value by total portfolio value.

A $24,000 position in a $120,000 portfolio has a 20% weight. On decimal scale, that is 0.20. On percentage scale, that is 20.

3. Square each weight.

On the percentage scale, 20² equals 400. A 5% position contributes 25. The non-linear penalty is the entire mechanism.

4. Sum the squared weights.

The result is the portfolio HHI. No expected return input. No covariance matrix. No volatility assumption.

A portfolio example:

HHI = 900 + 400 + 225 + 100 + 225 + 25 + 25 = 1,900.

The portfolio sits in the moderately concentrated band. The top two allocations contribute 1,300 points, or 68.4% of the total HHI. That is the diagnostic value. The score is not just the final number. The contribution table identifies where concentration originates.

The same portfolio using decimal weights gives:

0.30² + 0.20² + 0.15² + 0.10² + 0.15² + 0.05² + 0.05² = 0.19.

Multiply by 10,000 to convert to the percentage scale. 0.19 × 10,000 = 1,900.

No information changes. Only notation changes.

Interpreting the 1,500 and 2,500 thresholds

The common HHI bands on the 0–10,000 scale are:

These bands are useful. They are not universal portfolio law. There is no single ideal HHI for all retail portfolios, multi-asset mandates, factor portfolios, or index replication strategies. A concentrated quality-equity mandate may intentionally run above 2,500. A global ETF allocation may target below 1,500. A cash-heavy tactical book may show low HHI while still carrying duration, currency, or reinvestment risk.

The bands work best as a control variable. They answer one narrow question: how evenly is capital distributed across the defined buckets?

The defined bucket matters. HHI can be calculated at several levels:

• Security level.

Each stock, bond, ETF, option position, or fund is a separate line. This detects single-name and instrument concentration.

• Issuer level.

Different securities from the same issuer are consolidated. Common stock, preferred stock, and corporate bonds from one issuer become one exposure.

• Asset-class level.

Equity, sovereign debt, credit, commodities, cash, and alternatives are treated as buckets. This measures allocation concentration, not instrument concentration.

• Region level.

US, Europe, Japan, emerging Asia, Latin America, and other regions are separated. This is relevant for global stock indexes and cross-border ETF portfolios.

• Factor level.

Growth, value, momentum, low volatility, quality, duration, credit spread, and currency carry are grouped where exposures can be estimated.

A portfolio can score differently under each lens. A book with 30 individual US technology stocks may show low security-level HHI and high sector-level HHI. A portfolio with five global ETFs may show high instrument-level HHI and moderate region-level HHI. The result is not contradictory. It is a function of aggregation.

The HHI answer changes when the exposure definition changes. The arithmetic stays constant.

For index investors, this is not a minor detail. A portfolio that holds the S&P 500, Nasdaq 100, and a broad US total-market ETF may appear diversified across three tickers. At the underlying security level, overlap can be substantial. At the factor level, the book may still be concentrated in US large-cap growth and high-duration equity cash flows. HHI by ticker will understate that structure.

Position count versus effective number of holdings

HHI can be converted into an “effective number of equally weighted holdings.” This is often more intuitive.

Using decimal HHI:

Effective number = 1 / HHI

Using percentage-scale HHI:

Effective number = 10,000 / HHI

A portfolio with an HHI of 2,000 has an effective holding count of 5. It may contain 20 securities, but its concentration is equivalent to five equally weighted positions. A portfolio with an HHI of 1,250 has an effective holding count of 8. A portfolio with an HHI of 500 has an effective holding count of 20.

This conversion is useful in portfolio review. Nominal holding count is a weak statistic. Effective holding count is a concentration-adjusted statistic. The gap between the two is the signal.

A large gap indicates weight skew. A portfolio with 60 holdings and 8 effective holdings is not broad in capital terms. It is a core-satellite book, whether intentional or not. A portfolio with 12 holdings and 10 effective holdings has limited weight skew. The second book may have fewer names but better capital symmetry.

The limits: HHI ignores correlation, volatility, and shared risk factors

HHI has one core defect. It does not account for correlation between assets. It only measures capital allocation concentration.

A portfolio can have a low HHI and still carry high common-factor exposure. Twenty equal-weight bank stocks produce an HHI of 500. That looks diversified by capital weight. It may still be concentrated in credit cycle risk, yield-curve risk, regulatory risk, and domestic macro beta. Ten equal-weight semiconductor stocks produce an HHI of 1,000. The score is below 1,500, but factor clustering remains.

The same applies to ETFs. A global equity ETF, a US technology ETF, and a Nasdaq 100 ETF may be separate instruments. They may also share large constituents. HHI by ticker does not detect overlap. Underlying holdings analysis is required.

HHI also ignores volatility. A 10% allocation to a 2-year Treasury ETF and a 10% allocation to a levered volatility product contribute the same 100 HHI points on the percentage scale. Their risk contribution is not comparable. HHI measures capital concentration, not variance concentration.

It also ignores directionality. A long equity ETF and a short futures position could offset market beta but still contribute to gross capital concentration depending on how the portfolio is measured. Long-short books require separate treatment: gross exposure HHI, net exposure HHI, and factor exposure HHI can produce different answers.

This is why HHI should sit beside, not replace, other risk metrics:

• Volatility and drawdown statistics measure realized return dispersion and downside path behavior.

• Beta and factor models measure sensitivity to broad market and style variables.

• Correlation matrices detect clustering across assets.

• Value at Risk and expected shortfall estimate tail exposure under distributional assumptions.

• Sharpe ratio and information ratio connect returns to volatility or active risk.

• Liquidity metrics measure exit capacity under stress.

HHI has no opinion on those variables. That is a strength when the task is narrow. It is a weakness when the task is total risk classification.

The nearest analogy is operational triage. A first-pass metric separates cases that need deeper work from cases that do not. It is not a full diagnosis. The same distinction appears outside markets in emergency guidance; a resource such as first aid and urgent-condition guidance separates immediate escalation from basic response. HHI does the portfolio equivalent: it flags concentration that needs further decomposition.

Integrating HHI into a broader risk management framework

HHI works best when calculated at fixed intervals and after material price moves. Monthly is sufficient for slow allocation books. Daily or intraday is appropriate for trading books, options overlays, or portfolios with high turnover.

A practical risk process uses HHI as a boundary condition:

1. Set a target band by mandate.

A diversified multi-asset ETF portfolio may set an HHI ceiling near 1,500 at the instrument or allocation-bucket level. A concentrated equity portfolio may accept 2,500–4,000. The number must match the strategy. A low number is not automatically superior.

2. Calculate HHI at multiple aggregation levels.

Ticker-level HHI is the starting point. Sector, region, issuer, factor, and currency HHI are separate diagnostics. A global stock indexes portfolio should at least split US, developed ex-US, emerging markets, and currency exposure.

3. Track HHI drift from price performance.

Concentration increases mechanically when winners compound faster than the rest of the book. No trade is required for risk to change. A 20% position can become 30% through relative performance. Its HHI contribution rises from 400 to 900. That is a 500-point increase from one line item.

4. Separate intended concentration from accidental concentration.

A benchmark-aware manager may intentionally hold a large index ETF. A stock picker may intentionally hold a 12% highest-conviction position. Those are mandate decisions. Accidental concentration occurs through overlapping ETFs, unmonitored factor exposure, or legacy positions left outside the allocation process.

5. Use contribution-to-HHI, not only total HHI.

A portfolio HHI of 1,850 is less informative than knowing that one position contributes 625 points and the next four contribute 700 combined. Contribution analysis turns the statistic into an action map.

The mechanics become more precise with rebalancing bands. For example:

These numbers are examples, not universal thresholds. The structure is the point. The portfolio has a measurable concentration budget. HHI converts that budget into observable points.

Worked comparison: two portfolios with the same asset count

Consider two seven-asset portfolios. Both have the same number of holdings. Their HHI profiles differ.

Portfolio A is just above the diversified threshold. Portfolio B is above the high-concentration threshold. The difference is driven almost entirely by the 45% US equity allocation. That line contributes 2,025 points. It accounts for 78.2% of Portfolio B’s total HHI.

The technical inference is direct. If the mandate does not permit high US equity concentration, Portfolio B requires adjustment. If the mandate is a US-led equity allocation, the score is not a violation. It is a quantified expression of the mandate.

This is where HHI avoids language drift. “Balanced,” “diversified,” and “global” are imprecise. A score of 2,588 is not.

Applying HHI to global stock indexes and ETF portfolios

Index portfolios create a specific measurement problem. The instrument count is often low, while the underlying constituent count is high. HHI can be calculated both ways.

An investor holding three ETFs at 50%, 30%, and 20% has ticker-level HHI of:

50² + 30² + 20² = 2,500 + 900 + 400 = 3,800.

That appears highly concentrated by instrument. If those ETFs hold thousands of underlying securities across regions, underlying security-level HHI may be far lower. Neither number is wrong. They answer different questions.

Ticker-level HHI measures implementation concentration. It captures dependence on fund structure, provider, vehicle liquidity, and trading mechanics. Underlying-level HHI measures economic exposure concentration. It captures issuer and constituent weights. Sector-level HHI captures macro and earnings-factor concentration.

For broad index portfolios, the useful hierarchy is:

• Ticker HHI for vehicle concentration.

• Underlying issuer HHI for single-company concentration.

• Country or region HHI for geographic concentration.

• Sector HHI for earnings-cycle concentration.

• Currency HHI for FX exposure.

• Factor HHI for style concentration.

The calculation remains identical. Only the grouping changes.

This matters in portfolios built from cap-weighted global equity indexes. A developed-market ETF and a global all-country ETF may overlap materially. A US total-market ETF and an S&P 500 ETF may overlap heavily in large-cap constituents. A Nasdaq 100 ETF layered on top of a broad US equity ETF increases technology and growth exposure even if ticker-level weight appears modest.

HHI will not solve overlap unless the inputs are decomposed. The input table dictates the output quality.

The technical use: rebalance thresholds and probability discipline

HHI is most useful when tied to predefined action levels. Without thresholds, it becomes a reporting statistic. With thresholds, it becomes a control variable.

A portfolio manager can define a rebalancing rule:

• HHI below 1,500: no concentration action unless factor metrics breach.

• HHI 1,500–2,500: review top five contribution lines.

• HHI above 2,500: rebalance unless concentration is mandate-approved.

• HHI change above 250 points in one review period: run attribution.

• Single line contribution above 1,000 points: review position cap.

A 250-point change is not trivial. It is equivalent to adding a 15.8% isolated position contribution, since 15.8² is approximately 250. It can also emerge from multiple smaller drifts. The source matters.

HHI should also be assessed against realized volatility. If HHI rises and volatility rises, concentration and risk are moving in the same direction. If HHI rises while volatility falls, concentration may be moving into lower-volatility assets. If HHI falls while correlation rises, capital is more distributed but exposures may still converge. These combinations produce different portfolio states.

The metric is not predictive by itself. A high HHI does not guarantee loss. A low HHI does not guarantee protection. The probabilistic statement is narrower: as HHI rises, portfolio outcome dependence on fewer allocation units increases. That raises sensitivity to idiosyncratic errors if those units are securities, and to macro classification errors if those units are asset classes or regions.

Final technical pivot

To check and calculate portfolio diversification via Herfindahl index, square each current portfolio weight and sum the results. Use the 0–10,000 scale for operational clarity. Below 1,500 indicates broad capital distribution. Between 1,500 and 2,500 indicates moderate concentration. Above 2,500 indicates high concentration.

The statistic is clean because it is limited. It measures allocation concentration only. It does not measure correlation, volatility, beta, liquidity, drawdown, or factor crowding. The correct use is as a first-line concentration control, repeated across ticker, issuer, sector, region, and factor groupings.

Technical pivot level: on the percentage scale, 2,500 is the concentration boundary. Above it, the portfolio has fewer than four effective equally weighted holdings. Unless the mandate explicitly accepts that structure, the probability of single-bucket dominance is no longer incidental. It is the portfolio design.