Portfolio Diversification: HKSI Paper 12
Diversification combines assets whose returns are not perfectly correlated, which can reduce portfolio volatility without necessarily lowering expected return by the same amount.
- It cannot eliminate broad systematic risk.
- In Paper 12, understand the covariance formula, the role of correlation and how an investor's objectives constrain diversification.
On this page12 sections
- What diversification does
- Calculate the risk of a two-asset portfolio
- Correlation is not the same as causation
- Systematic and unsystematic risk
- Diversification must fit the client
- Original scenario: two assets
- Original practice question
- Common calculation errors
- How to use diversification in portfolio review
- Understand when correlation changes the answer
- What diversification cannot promise
- A second two-asset example
- Main mechanism
- Combine assets with imperfectly correlated returns
- Risk reduced
- Unsystematic or asset-specific risk can be diversified
- Risk that remains
- Systematic market risk
- Two-asset variance
- W1²σ1² + w2²σ2² + 2w1w2ρ12σ1σ2
What diversification does
Diversification spreads investment exposure across assets, issuers, industries, regions or risk factors. If one holding falls while another is stable or rises, the combined portfolio may fluctuate less than either asset alone. The benefit depends on how returns move together, not merely on the number of positions. Holding twenty securities from the same industry may leave the portfolio exposed to one common shock. A smaller portfolio across unrelated risk sources can offer more genuine diversification.
In portfolio theory, the key statistic is covariance, often expressed through correlation. When two assets are perfectly positively correlated, their returns move together and combining them does little to reduce volatility. If correlation is below one, a portfolio can benefit from imperfect co-movement. Negative correlation can provide a larger risk reduction, although correlations can change during market stress. A historical estimate is not a guarantee of future diversification.
Calculate the risk of a two-asset portfolio
For two assets, portfolio variance is w1²σ1² + w2²σ2² + 2w1w2ρ12σ1σ2. Here w1 and w2 are portfolio weights, σ1 and σ2 are standard deviations, and ρ12 is the correlation coefficient. The portfolio standard deviation is the square root of the variance. Keep variance and standard deviation distinct: the formula produces variance first, so take a square root before reporting volatility.
Suppose two assets each have annual volatility of 20%, the portfolio holds 50% in each, and their return correlation is 0.25. Using decimals, variance equals (0.5² × 0.2²) + (0.5² × 0.2²) + 2 × 0.5 × 0.5 × 0.25 × 0.2 × 0.2. The first two terms total 0.0200. The covariance contribution is 0.0050. Portfolio variance is 0.0250, and standard deviation is square root of 0.0250, or about 15.81%.
| Component | Calculation | Contribution to variance |
|---|---|---|
| Asset 1 | 0.5² × 0.2² | 0.0100 |
| Asset 2 | 0.5² × 0.2² | 0.0100 |
| Covariance | 2 × 0.5 × 0.5 × 0.25 × 0.2 × 0.2 | 0.0050 |
| Total variance | 0.0100 + 0.0100 + 0.0050 | 0.0250 |
| Portfolio standard deviation | √0.0250 | 15.81% |
If the assets were perfectly correlated at +1, the covariance term would be larger and the equal-weight combination would have 20% volatility, the same as either asset. With a correlation of 0.25, the portfolio falls below 20%. The expected return of the two-asset portfolio is the weighted average of expected returns, assuming the weights and return estimates are consistent. Diversification changes risk through co-movement; it does not make an asset's expected return disappear.
Correlation is not the same as causation
Correlation describes how returns have moved together over a measurement period. It does not prove that one asset causes the other's performance. An estimate can change with the sample period, frequency and market regime. A pair that appeared weakly correlated in calm markets may become more correlated in a crisis. Managers therefore use diversification as a risk-management principle, not as a promise that losses will be offset in every scenario.
The correlation coefficient ranges from −1 to +1. A value of +1 means perfect positive linear co-movement; 0 indicates no linear correlation in the sample; −1 means perfect negative linear co-movement. Zero correlation does not mean the assets are independent in every possible way, especially when return distributions are not normal. For exam calculations, use the coefficient provided and keep it in decimal form unless the question uses percentages.
Systematic and unsystematic risk
Unsystematic risk is specific to a company, issuer or limited group of investments. Examples include a product recall, a failed project or an issuer's governance problem. Diversifying across independent issuers can reduce the effect of one event. Systematic risk affects broad markets or economies, such as a global recession, interest-rate shock or market-wide liquidity event. A large number of securities cannot remove this shared exposure simply by increasing the count of holdings.
This distinction also explains beta in CAPM. Beta measures systematic sensitivity to the market, while standalone volatility reflects both systematic and asset-specific variation. In a diversified portfolio, specific risk can be reduced, but the portfolio still has market risk. Do not answer that diversification eliminates all risk. A broadly diversified portfolio can lose value when the risks affecting its asset classes are correlated or when expected returns fall across markets.
Diversification must fit the client
Diversification is not a goal detached from the investor's IPS. A client may have a concentrated employer-share position, a liability in a specific currency or a legal restriction. The manager should consider the total portfolio and client circumstances, not only the assets in a managed account. A recommendation should be reasonably suitable when applicable, and the investment strategy should reflect the objectives and constraints in the Paper 12 framework.
A client with a large HKD liability in two years may not benefit from adding a foreign-currency asset merely because it reduces equity concentration. The currency exposure could create a mismatch with the upcoming payment. Another client may need global exposure because all employment and property wealth is tied to one local economy. The right diversification depends on the client's total risks, cash needs, investment horizon and ability to bear losses.
Original scenario: two assets
A candidate portfolio has 60% in Asset A and 40% in Asset B. Both have 15% volatility, and correlation is 0.20. Variance is 0.6² × 0.15² + 0.4² × 0.15² + 2 × 0.6 × 0.4 × 0.20 × 0.15 × 0.15. The first term is 0.0081, the second is 0.0036 and the covariance term is 0.00216. Total variance is 0.01386, so standard deviation is √0.01386, about 11.77%.
The portfolio's volatility is lower than either asset's 15%, because correlation is well below one. This result does not mean the expected return is guaranteed or that the portfolio cannot lose. It only describes risk under the stated standard deviations, weights and correlation estimate. If correlation rises, the covariance term rises and the diversification benefit shrinks. If weights change, recompute every term rather than averaging standard deviations.
Original practice question
Two assets each have a standard deviation of 10%. An equal-weight portfolio has a correlation of +1 between the assets. What is the portfolio standard deviation?
- 0%
- 5%
- 10%
- 14.14%
Common calculation errors
Do not omit the covariance term. Doing so assumes the assets have zero covariance, which the question may not state. Do not square a correlation coefficient; the two-asset formula uses correlation multiplied by both standard deviations. Do not stop at variance if the question asks for standard deviation. Finally, ensure weights sum to one and are expressed as decimals: 60% is 0.60, not 60.
A useful check is to compare your result with reasonable bounds. With identical volatilities and a positive correlation no greater than one, a diversified mix should not produce volatility above the same asset volatility in the equal-risk case. Negative correlation can lower risk further, while a perfect negative relationship with carefully balanced exposures can theoretically offset returns. If your calculation gives a much larger result, check percentages, squares and the covariance term.
How to use diversification in portfolio review
A portfolio review should look through fund labels and examine underlying exposures. Two funds with different names can hold many of the same securities. A manager should consider issuer, sector, geography, currency, duration and factor exposure, then assess how holdings might behave under shared shocks. Correlation matrices can help, but they summarize historical data and should be paired with stress scenarios and judgment.
Paper 12's syllabus includes risk and return, portfolio theory, investor preferences and asset allocation. Diversification links these concepts: a client's risk preference informs the portfolio, while covariance explains why a combination's risk differs from the weighted average of individual risks. The public syllabus is version 3.7, effective 1 November 2024. Verify the guide update page before your sitting; this article uses original calculations, not HKSI sample items.
Understand when correlation changes the answer
Correlation affects portfolio risk because it measures how returns move together. Two assets with high standalone volatility can still contribute diversification if their returns are imperfectly correlated. Conversely, a portfolio spread across many securities can remain exposed to a common sector, interest-rate or currency factor. Count of holdings is a weak proxy for economic diversification.
Suppose a portfolio holds domestic equities, foreign equities and corporate bonds. The labels suggest variety, but an exposure review should ask whether the equities share a growth factor, whether the foreign holdings introduce currency risk, and whether the corporate bonds could weaken when equities fall. Diversification analysis compares sources of return and risk, not only product names.
What diversification cannot promise
The covariance term makes diversification possible in the portfolio-variance equation. If correlation is +1, assets move together perfectly in the model and the two-asset diversification benefit disappears. As correlation falls, the covariance contribution falls, all else equal. This describes expected co-movement under selected data and assumptions; it cannot guarantee protection in every market episode.
A second two-asset example
Assume two assets each have 10% standard deviation and half weights. With perfect positive correlation, portfolio standard deviation remains 10%. With zero correlation, variance is 0.5²×0.10² + 0.5²×0.10² = 0.005, so standard deviation is about 7.07%. The reduction comes from imperfect co-movement, not from lowering either asset’s own volatility.
If equal-volatility assets had correlation of −1 and equal weights, their return movements would theoretically offset. Real investments rarely preserve this relationship, and expected returns, liquidity and constraints still matter. Treat the calculation as an illustration, not a guarantee that a balanced pair cannot lose value.
The benefit of diversification must be weighed against mandate and implementation. A hedge that reduces modeled volatility might add cost, leverage, liquidity risk or tracking error. A client may also have restrictions that rule out certain assets. The suitable portfolio is not simply the one with the lowest calculated variance; it must meet the objective within the stated constraints.