Why a portfolio of good investments can be a bad portfolio

A portfolio is a system, not a collection

The natural way to build a portfolio is one decision at a time. A fund performs well, so it is added. A new theme sounds persuasive, so it receives an allocation. An employer awards shares. A pension uses one default fund while an ISA holds another. Every decision has its own explanation, but nobody checks the aggregate.

Portfolio construction reverses the sequence. It begins with the required outcome, acceptable failure, time horizon and available liquidity. It assigns roles and risk budgets. Only then does it choose securities or funds to implement those roles.

Every new holding is also a decision not to put that capital somewhere else. ‘Add’ is not a complete portfolio instruction. It needs a funding source and a reason the new combination is better.

The central idea is interaction

Harry Markowitz's 1952 paper Portfolio Selection formalised a durable insight: an investment should not be evaluated only by its own expected return and variance. What matters to portfolio risk is also how its return moves with the returns of the other holdings. The relevant object is a covariance matrix, not a league table of favourite assets.

The letters are less important than the logic. The weights matter. Each asset's volatility matters. The correlation between them matters. A volatile asset can improve a portfolio if it behaves differently enough and is sized sensibly. A stable-looking asset can add little diversification if it fails for the same reason as everything else.

Variance is not a complete definition of risk. It does not capture fraud, an inability to sell, a tax bill, a margin call or money being unavailable when a goal arrives. But the portfolio perspective survives every broader risk definition: judge an asset by its contribution to the whole, not by its biography in isolation.

Different labels can conceal the same holdings

Fund names create an illusion of variety. ‘Global’, ‘US’, ‘growth’, ‘quality’, ‘technology’ and ‘innovation’ sound like six categories. Their largest positions can overlap heavily, and even different companies can depend on the same interest-rate, valuation and economic-growth environment.

Look-through exposure is the first portfolio X-ray. For every fund, identify the underlying regions, sectors, largest companies, currencies and asset types. Then multiply those exposures by the fund's portfolio weight and add them across every account.

The interactive example is deliberately simplified. Actual index weights change, categories overlap and an individual company's revenues can be more global than its listing. The point is not to produce a perfect decimal. It is to expose a concentration that the fund names hide.

Count economic bets, not account lines. A global equity fund plus a US tracker can be a deliberate US overweight. A US tracker plus a growth index plus a technology fund is a stronger and more specific choice. Add employer shares and the portfolio becomes connected to the household's future earnings as well as its invested wealth.

Different securities can still share one risk driver

Overlap is broader than owning the same company twice. A property fund, a highly leveraged housebuilder and a buy-to-let property are legally different investments. All can suffer when financing costs rise and property transactions freeze. A high-yield bond fund and an equity-income fund hold different securities, but both can be exposed to weak corporate cash flow and recession.

Factor and scenario views complement the holdings view. Ask how much of the portfolio depends on strong growth, falling inflation, easy credit, stable currencies, rising valuations, liquid markets and low interest rates. If several holdings need the same environment, they should not receive full diversification credit merely because they use different wrappers.

Capital weight is not risk weight

A portfolio described as 60% equities and 40% bonds looks as though risk is distributed 60/40. It usually is not. The more volatile component can dominate changes in portfolio value.

Take a simplified example. Assume equities have annual volatility of 16%, bonds have 6% and their correlation is zero. At 60% and 40% capital weights, the equity term contributes about 94% of the calculated portfolio variance. The exact number is assumption-dependent, and realised losses will not follow it neatly. The lesson is stable: pounds allocated and risk contributed are different measures.

This does not mean every household needs institutional risk-parity machinery. It means a large allocation to a calmer asset can still provide less risk diversification than its capital weight suggests, while a small leveraged, option-based or concentrated position can matter far more than its percentage implies.

Correlation is a relationship, not a permanent property

Historical correlation is often treated like a product specification. It is an estimate from one period, frequency and currency. It can change when inflation, policy, market structure or investor behaviour changes.

François Longin and Bruno Solnik's research on major international equity markets found that correlation increased in bear markets, rather than simply with volatility in both directions. The finding is historical and specific to their data and method, but it captures an uncomfortable portfolio truth: some diversification weakens precisely when losses matter most.

The answer is not to declare diversification useless. It is to test more than the average relationship. Run at least three views: a normal environment, a shock resembling a known historical mechanism and a bespoke scenario aimed at the portfolio's most crowded assumption.

• Growth shock: equities and lower-quality credit fall together while unemployment risk rises.
• Inflation shock: equities and nominal bonds both fall as discount rates rise.
• Liquidity shock: private assets cannot be sold, spreads widen and the liquid sleeve funds every cash need.
• Currency shock: foreign assets move sharply in sterling even if their local prices are stable.
• Company shock: employer shares fall while bonuses, job security and local property demand weaken.

A stress test is not a forecast. Its job is to reveal whether one plausible mechanism can damage several positions and the household at the same time.

A good investment can be wrong for its assigned role

Portfolio labels should describe jobs. Growth assets seek long-term real wealth. Defensive assets are expected to reduce damage in selected shocks. Liquidity assets fund near-term spending. Diversifiers provide a genuinely different return source. Satellites express a narrower view within a capped risk budget.

Problems begin when an asset is excellent at one job and quietly assigned another. A volatile equity fund may be a reasonable twenty-year growth holding and a terrible house-deposit reserve. An unhedged global bond fund can be diversified by issuer yet fail as a sterling stabiliser because currency movement dominates its bond return. Vanguard's 2026 research on global bonds makes the same role distinction: broader issuer exposure can reduce volatility, but unmanaged currency can undermine the stabilising purpose of fixed income.

Liquidity must be designed at portfolio level

An illiquid investment can be entirely rational when its expected reward, governance and size fit the investor. A portfolio full of rational illiquid allocations can still be unable to pay a tax bill, fund retirement withdrawals or rebalance during a crisis.

Liquidity is not binary. Cash can usually be used immediately. A traded fund may sell quickly but at a poor price. Property can take months. Private funds can lock capital for years, call additional commitments and distribute cash on a schedule the investor does not control. Some open-ended funds holding illiquid assets can suspend dealing under stress.

The FCA's rules for certain open-ended funds investing in inherently illiquid assets explicitly address restricted access, valuation uncertainty and the risk of forced asset sales. At household level, the same principle applies: match the liquidity of the portfolio to the timing and flexibility of its obligations.

Do not count the same cash twice. Money assigned to an emergency fund cannot also support a private-market commitment and next year's house purchase. A liquid sleeve is not an embarrassment beside higher-returning assets. It is what prevents the portfolio from selling the wrong investment at the wrong time.

The household balance sheet is part of the portfolio

A brokerage account is not the investor's full economic exposure. Include the main home, investment property, pension, employer shares, deferred compensation, business ownership, debts and human capital. Future earnings are not a tradeable asset, but they change which financial risks the household can sensibly add.

Research by Joseph Blasi, Douglas Kruse and Harry Markowitz on employee ownership describes the diversification issue directly: risky employer stock can belong in an efficient portfolio, but the overall portfolio must be diversified and personalised to the worker's circumstances. The company does not become a bad investment merely because it also pays the salary. The combined exposure becomes more fragile.

Debt belongs in the picture too. A household with a large variable-rate mortgage already has sensitivity to interest rates and monthly cash flow. Adding highly leveraged property exposure may deepen rather than diversify that risk. Net worth can look broad while every major asset depends on one country, one housing market and one income source.

One portfolio cannot serve incompatible goals without rules

Retirement in thirty years, school fees in seven years and a house deposit in eighteen months do not share one risk budget. Combining them in a single account is possible, but only if the portfolio records which assets support which liability and how the allocation changes as each date approaches.

Otherwise, the long-term goal becomes an excuse to invest near-term money too aggressively, or the near-term goal forces the entire portfolio to stay too cautious for decades. The holding is not at fault. The architecture failed to separate horizons.

• Name each goal, amount, first withdrawal date and flexibility.
• Reserve unavoidable near-term spending in assets suitable for that horizon.
• Give long-term growth capital its own allocation and rebalancing rule.
• Model withdrawals, not only the final portfolio value.
• Reduce risk deliberately as a fixed-date goal approaches, rather than waiting for a favourable market.

More managers can mean more duplication, not more diversification

A multi-manager portfolio can contain several excellent managers who all own similar securities or depend on the same factor. Selecting each because it has a strong process does not guarantee that their combined active positions are useful.

Two global equity managers can cancel each other's distinctive views and leave an expensive index-like result. Three alternative funds can each report low equity beta while all depend on leverage, stable financing and liquid markets. Several credit funds can hold different bonds issued by businesses vulnerable to the same refinancing cycle.

For each manager or fund, distinguish market exposure, factor exposure and genuinely idiosyncratic decisions. Then ask whether the portfolio is paying several times for the same return source. Investment fees compound regardless of whether the underlying duplication was intentional.

Three worked examples

Example 1: the fund collector

Leah owns the five equity holdings from the opening example. She chose every fund for a defensible reason, but has no target asset allocation. The portfolio X-ray shows that US and technology exposure dominate, while employer shares connect invested wealth to her income.

The repair is not necessarily to sell every specialist holding. Leah first defines a global equity core, a maximum employer-share exposure and a total satellite budget. She decides whether the US and technology overweights are intentional. Duplicate funds that do not earn a distinct role are removed, and new contributions are directed to missing defensive and regional exposures.

Example 2: the income portfolio

Marcus owns dividend shares, high-yield bonds, a property fund and a buy-to-let flat. Each produces income. He describes the portfolio as diversified because the cash arrives from rent, dividends and coupons.

The common driver is economic resilience. In a recession, tenants can struggle, property values can weaken, company dividends can be cut and credit spreads can widen. The portfolio is diversified by payment label but concentrated in growth and credit risk. Marcus adds high-quality liquid assets and tests the income after defaults, vacancies, dividend cuts and refinancing costs rather than assuming every yield continues.

Example 3: the sophisticated illiquid portfolio

Aisha has private equity, venture funds, private credit and direct property. Each manager is credible and every investment has a long horizon. Together they represent 70% of her investable assets. Capital calls remain outstanding, and her liquid portfolio must also fund living costs.

The portfolio may show smooth reported valuations, but the household carries exit, financing and cash-flow risk. Aisha models a period with delayed distributions, simultaneous capital calls and falling public markets. She sets a minimum liquid reserve and a maximum future commitment rate. The issue was not manager quality. It was aggregate liquidity.

Build the architecture before choosing the furniture

A practical portfolio can be built in eight steps. It does not require optimising forecasts to two decimal places. In fact, unstable return and correlation estimates can make a mathematically precise portfolio less robust.

• Define the goals: amounts, dates, currencies, essential versus optional needs and flexibility.
• Set the household risk budget: loss in pounds, income resilience, debt, behaviour and capacity to wait.
• Reserve liquidity: emergency cash, planned withdrawals, tax, property costs and unfunded commitments.
• Assign portfolio roles: growth, defence, liquidity, genuine diversifiers and capped satellites.
• Choose the simplest credible implementation for each role, considering costs, tax and access.
• Run a look-through X-ray across accounts, funds, property, employer exposure and currencies.
• Stress the common drivers, including scenarios where normal diversification weakens.
• Write rebalancing ranges, contribution rules and conditions that trigger a strategic review.

The order prevents product availability from dictating strategy. Without it, an investor sees a compelling fund and invents a portfolio role after buying. With it, every holding has to compete for a defined job.

Run a portfolio audit once a year

• Can every holding's role be stated in one sentence?
• Which holding would be reduced to fund a new idea?
• What are the largest look-through company, sector, country and currency exposures?
• Which risks are duplicated through different funds or asset classes?
• How much portfolio variance and stress loss comes from each major sleeve?
• What happens if correlations move against the portfolio during a fall?
• How much can be accessed within one week, three months and two years?
• Are future private commitments and tax payments included in liquidity needs?
• Does employment, property or business ownership reinforce financial-market exposure?
• Is near-term goal money separated from long-term growth capital?
• Are fees being paid more than once for substantially the same exposure?
• Which positions have drifted outside their intended range, and what is the rebalancing action?

Do not redesign the portfolio because a different asset recently performed better. The audit is about structure. Strategic changes follow a changed goal, capacity, evidence or implementation, not envy.

The bottom line

A good company can be overpriced. A good fund can duplicate a fund already owned. A good diversifier can be too small to matter. A good illiquid asset can leave the household short of cash. A good long-term investment can be disastrous for a near-term goal.

Portfolio quality lives in the relationships: weights, overlap, common drivers, liquidity, currency, tax, goals and behaviour. The correct unit of analysis is the whole household portfolio under stress, not each product on its best day.

Stop asking only whether each investment deserves a place. Ask what job it performs, what it depends on and what the portfolio would lose by removing it. If the answer is unclear, the holding may be good and the portfolio may still be bad.