Concentration vs Diversification Balance

What you'll learn

The trade-offs between concentration and diversification in portfolios
How correlations drive diversification benefits and when they fail
Step-by-step calculations for portfolio variance, risk contributions, and effective number of bets
Advanced tools: diversification ratio, risk budgeting, tracking error, and active share
How to size positions with Kelly-style methods and decide on fractional Kelly
Practical frameworks to choose your balance based on edge, costs, and drawdown tolerance

Big wins come from concentration; staying in the game comes from diversification. The optimal balance aligns your edge with your risk capacity.

Concept explanation

Concentration means putting a larger share of your portfolio into a small number of high-conviction ideas. If your research gives you a real edge, concentration magnifies that edge. The flip side is that mistakes, bad luck, or surprises can hurt more, leading to bigger drawdowns.

Diversification spreads your money across many holdings, sectors, or asset classes so that no single mistake dominates results. The key fuel for diversification is imperfect correlation: when different holdings do not move together, the portfolio’s total risk falls faster than returns do. You can think of diversification like a team sport: one player having a bad day is fine if others are playing well.

However, diversification is not a free lunch at all times. In normal markets, correlations among different stocks and assets are moderate, giving good risk reduction. But in crises, correlations often jump, and diversification benefits fade right when you need them most. Also, excessive diversification can dilute your best ideas, raising costs and taxes while dragging returns toward the market average.

The right balance depends on your edge, conviction, time horizon, drawdown tolerance, tax situation, and the type of risks you face. Professionals formalize this balance with risk models, risk budgets, and position-sizing rules that translate ideas into portfolio weights.

Why it matters

Concentration and diversification are not opposites; they are tools. Concentration is about extracting value from your insights. Diversification is about surviving variance and uncertainty. Portfolios that ignore either side often underperform or fail to meet investor constraints. A concentrated genius can be undone by unforeseen shocks; a perfectly diversified portfolio without edge just becomes an index minus fees.

Real-world investing has frictions: trading costs, taxes, liquidity, and behavioral limits. Concentration increases idiosyncratic risk, potentially raising tracking error versus broad benchmarks. Diversification across too many small positions can rack up costs and may not materially improve risk after a certain point. Knowing where the marginal benefit of adding another position tails off is essential.

Advanced methods—like risk contribution analysis, diversification ratios, and Kelly-style sizing—turn intuition into numbers. They help you decide whether to add a 41st stock, scale a high-conviction idea, or trim correlated bets that secretly concentrate your risk.

Calculation method

Portfolio variance and the role of correlation

For a portfolio with weights vector w and covariance matrix Σ, total variance is:

\sigma_p^2 = w^T \Sigma \; w

For equal-weighted stocks with identical volatility σ and average pairwise correlation ρ, an approximation for portfolio variance is:

\sigma_p^2 \approx \sigma^2 \left( \frac{1}{n} + \rho \cdot \frac{n-1}{n} \right)

Insight: As n grows large, the first term 1/n shrinks, but the correlation term tends toward ρ. Diversification removes idiosyncratic risk but leaves correlated market risk.

Risk contributions (which holdings truly drive risk?)

Portfolio volatility is σ_p. The marginal risk contribution of position i is:

MRC_i = \frac{\partial \sigma_p}{\partial w_i} = \frac{(\Sigma w)_i}{\sigma_p}

The total risk contribution of position i is weight times marginal contribution:

RC_i = w_i \times MRC_i

These help you spot hidden concentrations. Two positions in the same industry may each be 5 percent by weight but 10 percent by risk due to high correlation.

Diversification ratio and effective number of bets

Diversification ratio (DR) measures how much volatility reduction you get from combining assets:

DR = \frac{\sum_i w_i \sigma_i}{\sigma_p}

Higher DR indicates stronger diversification benefits.
Herfindahl-Hirschman Index (HHI) measures concentration of weights:

HHI = \sum_i w_i^2, \quad N_{eff} = \frac{1}{HHI}

N_eff approximates the effective number of equally-weighted bets. It ignores correlation but is a quick diagnostic. Combine it with RC_i for a fuller picture.

Kelly-style sizing (how much to concentrate when you have edge)

For a single repeated bet with probability p of winning b to 1, the Kelly fraction is:

f^{*} = \frac{p \cdot b - (1 - p)}{b}

For a risky asset with expected excess return μ and variance σ^2 under a lognormal approximation:

f^{*} \approx \frac{\mu}{\sigma^2}

In practice, estimates of μ and σ are noisy. Many professionals use fractional Kelly, such as one-half or one-quarter Kelly, to reduce drawdowns and estimation risk.

Active share and tracking error for active vs diversified blends

Active share: fraction of your portfolio that deviates from the benchmark’s holdings.
Tracking error (TE): volatility of active returns versus the benchmark. If your concentrated sleeve has high TE, blending with a diversified core reduces overall TE. Risk budgeting allocates TE to the best ideas.

Case study

Assume:

Each stock has annual volatility σ = 25 percent
Average pairwise correlation ρ = 0.3 in normal markets
Equal weights

A) Five-stock concentrated portfolio

n = 5. Use the equal-vol, equal-correlation approximation:

\sigma_p^2 \approx 0.25^2 \left( \frac{1}{5} + 0.3 \cdot \frac{4}{5} \right) = 0.0625 \times (0.2 + 0.24) = 0.0625 \times 0.44 = 0.0275 \sigma_p \approx \sqrt{0.0275} \approx 16.6\%

B) Forty-stock diversified portfolio

n = 40:

\sigma_p^2 \approx 0.25^2 \left( \frac{1}{40} + 0.3 \cdot \frac{39}{40} \right) = 0.0625 \times (0.025 + 0.2925) = 0.0625 \times 0.3175 = 0.01984 \sigma_p \approx \sqrt{0.01984} \approx 14.1\%

Observation: Going from 5 to 40 stocks reduces volatility from about 16.6 percent to 14.1 percent—useful but not dramatic—because correlation limits the benefit. If correlations rise to 0.6 in stressed markets, the 40-stock volatility increases:

\sigma_p^2 \approx 0.0625 \times \left(\frac{1}{40} + 0.6 \cdot \frac{39}{40}\right) = 0.0625 \times (0.025 + 0.585) = 0.0625 \times 0.61 = 0.03813 \sigma_p \approx \sqrt{0.03813} \approx 19.5\%

Diversification can fade when correlations spike. This argues for mixing across asset classes or strategies, not just more stocks in the same market.

Now, add risk contributions. Suppose two of the 5 stocks are in the same industry and correlate at 0.8 with each other while the rest average 0.3. A full covariance-based RC_i calculation would show those two names contributing a larger share of portfolio risk than their 20 percent weights suggest, flagging a hidden concentration.

Finally, consider position sizing via fractional Kelly. If your expected excess return for a concentrated idea is 8 percent with σ = 30 percent, a naive Kelly fraction is roughly 0.08 / 0.09 ≈ 0.89, which is unrealistically high given estimation error. Using one-quarter Kelly implies about 22 percent of capital, often still too large for a single stock. This demonstrates why professionals cap single-name weights and use risk budgets.

Practical applications

Risk-budgeted core-satellite: Hold a diversified core (broad index or multi-factor fund) and a concentrated satellite of best ideas. Set a tracking error budget, for example 3 to 5 percent annualized, and size the satellite so total TE stays within plan.
Use diversification ratio and RC_i to guide adds and trims: Add new positions when they increase DR meaningfully and do not overly spike any one RC_i. Trim or pair-trade highly correlated positions that bloat a single risk cluster.
Cap single-name and cluster risk: For example, limit any one name to 5 percent weight or 10 percent of total risk contribution, and limit any one industry to 25 percent of risk. RC_i and sector-level RCs operationalize these caps.
Estimate correlations realistically: Use multi-year windows and stress scenarios. Build a stressed covariance matrix by blending normal and crisis correlations to avoid overestimating diversification.
Fractional Kelly for sizing: Start with your best estimate of expected excess return and variance. Apply a conservative fraction (one-half or one-quarter) to account for model error. Cap weights with practical limits from liquidity and compliance.
Define drawdown tolerances: Convert your target volatility to expected shortfall (average loss beyond a percentile) to check comfort. If expected shortfall is too high, tilt toward more diversification or reduce total risk.
Rebalancing cadence: Concentrated portfolios drift faster. Set rules to rebalance when weights or RC_i breach bands, balancing turnover with control of risk.

Common misconceptions

よくある誤解

- More positions always mean less risk. Beyond a moderate number, correlation dominates and risk reductions become small. - Diversification eliminates losses. It reduces idiosyncratic risk but cannot remove market-wide shocks. - If conviction is high, size without limits. Even great ideas face estimation error and tail risk; use fractional Kelly and risk caps. - Equal weights ensure fairness. Equal weights can hide correlated clusters; check risk contributions and correlations. - Diversification is only about the number of stocks. True diversification spans factors, sectors, geographies, and asset classes.

Summary

まとめ

- Concentration magnifies edge but raises drawdown risk; diversification reduces idiosyncratic volatility. - Correlation drives diversification limits; when correlations rise, benefits fade. - Use portfolio variance, risk contributions, and diversification ratio to quantify balance. - HHI and N_eff are quick weight-based diagnostics; combine with RC_i for correlation-aware insight. - Position size with fractional Kelly to temper estimation error and tail risk. - Apply risk budgets, caps, and stress-tested correlations to control hidden concentrations. - Choose a core-satellite approach to blend diversification with high-conviction ideas.

Glossary

Concentration risk: Exposure to large losses due to heavy weights in a few positions or correlated clusters.

Diversification: Spreading investments across different holdings so that not all move together, reducing idiosyncratic risk.

Idiosyncratic risk: Risk specific to a company or asset that can be diversified away.

Systematic risk: Market-wide risk that affects many assets and cannot be diversified away.

Correlation: A statistic from -1 to 1 describing how assets move together; higher means less diversification benefit.

Covariance: A measure of joint variability between assets; building block of portfolio variance.

Portfolio variance: The expected squared volatility of a portfolio, computed as w^T Σ w.

Risk contribution: How much each position contributes to total portfolio risk, often via marginal risk contribution times weight.

Diversification ratio: Sum of weighted individual volatilities divided by portfolio volatility; higher indicates more diversification.

Herfindahl-Hirschman Index: Sum of squared weights; higher implies more concentration. Inverse approximates effective number of bets.

Kelly criterion: A formula that maximizes long-term growth rate and suggests optimal bet size; often used fractionally in practice.

Tracking error: Volatility of portfolio returns relative to a benchmark, used to manage active risk.

Active share: The proportion of a portfolio that differs from its benchmark’s holdings.

Risk budget: A set allocation of total risk or tracking error across positions or strategies.

Expected shortfall: Average loss in the worst tail of the return distribution, used to gauge drawdown risk.

Concentration vs Diversification: Optimal Balance