Modern portfolio theory (Markowitz)
Modern portfolio theory (MPT), formalised by Harry Markowitz in 1952, demonstrates that a portfolio is not judged security by security, but as a whole. What matters is not the return of an isolated line, but its contribution to the return-risk pairing of the whole.
The central intuition is that by combining assets whose performances do not move at the same time, you reduce overall volatility (standard deviation) without giving up expected return. Two portfolios showing 6% expected return can carry very different risks depending on the correlation of their components. Optimal diversification consists, for each accepted level of risk, in choosing the combination that maximises return.
The set of these superior combinations forms the efficient frontier: the curve of portfolios that, for a given risk, offer the best possible return, and vice versa. Any portfolio located below this frontier is dominated: somewhere else there exists an allocation that does better for the same risk. Building a portfolio means seeking to get closer to this frontier, not stacking up the best ideas of the moment.
To compare these portfolios, the Sharpe ratio measures the efficiency of the risk taken: Sharpe ratio = (portfolio return − risk-free rate) / volatility. A portfolio with 8% return, a 2% risk-free rate and 12% volatility shows a Sharpe of 0.5; at equal volatility, a Sharpe of 0.8 reflects a better return per unit of risk. It is this trade-off, and not the gross performance, that should guide your allocation choices.
Strategic vs tactical allocation
Strategic allocation sets the long-term split between major asset classes (equities, bonds, cash, real estate), calibrated to your horizon and your risk tolerance. Tactical allocation adjusts this target at the margin, according to market conditions. The first drives most of the performance; the second remains a fine-tuning.
The strategic split follows from two concrete parameters. The investment horizon first: capital that can be mobilised in three years is not managed like savings intended for retirement in twenty-five years, because the long run smooths out successive shocks. Risk tolerance next, that is, the loss you can bear financially and emotionally without selling at the worst moment. From these two parameters investor profiles are derived: cautious, balanced, dynamic.
Purely for illustration, a cautious profile could aim for 30% equities and 70% defensive assets, a balanced profile 55/45, a dynamic profile 80/20. These round numbers are not recommendations: they simply show how horizon and tolerance translate into weightings. Tactical allocation then consists in deviating by a few points around these targets, without ever turning a cautious profile into a dynamic one at the whim of the news.
Rebalancing is the discipline that holds the course. When equities rise sharply, their weight exceeds the target and the portfolio's risk increases without your knowledge; bringing the allocation back to its target, at regular intervals or when a threshold is crossed (for example 5 points of deviation), amounts to selling what has risen and reinforcing what has fallen. It is counter-intuitive, but it is precisely what anchors the steadiness of a portfolio over time.
Measuring risk: volatility, drawdown, VaR
The risk of a portfolio cannot be reduced to a single indicator. Volatility, drawdown (maximum loss) and VaR (Value at Risk) each shed light on a different facet. Combining them avoids the common error of confusing a convenient statistical measure with the risk actually experienced.
Volatility (standard deviation of returns) measures the amplitude of variations around the mean. It is useful, but it is an imperfect proxy for risk: it treats in the same way upside deviations, which no one dreads, and downside deviations, which hurt. It also assumes a well-behaved distribution of returns, whereas markets produce extreme movements far more frequently than the bell curve predicts.
Drawdown fills part of this blind spot: it measures the loss between a peak and the trough that follows, hence what you would actually have taken at the worst moment. A portfolio with low day-to-day volatility can suffer a drawdown of 40% during a crisis. It is often this figure, and not the standard deviation, that determines whether an investor holds their position or capitulates. VaR, for its part, answers a different question: what loss will not be exceeded, over a given horizon, with a set confidence level. A one-month 95% VaR of 8% means that one month in twenty, on average, the loss should exceed 8%.
The decisive limitation of VaR is that it says nothing about the magnitude of losses beyond the threshold: it bounds the frequency, not the severity of the extreme. Reading risk therefore means cross-referencing these three measures rather than relying on a single one, and supplementing them with stress scenarios that quantify the impact of a major shock.
Diversifying without diluting yourself
Diversifying reduces specific risk, particular to a security or a sector, but not systematic risk, that of the market as a whole. Most of the benefit is obtained with a reasonable number of positions: beyond that, you add complexity without reducing risk any further.
Diversification works thanks to correlations: combining assets that do not react identically to the same shocks dampens the variations of the whole. But it has a frontier. Once specific risk is largely neutralised, systematic risk remains, which you cannot eliminate through the number of lines, only modulate through your overall exposure to equities. Measured by beta, this risk rewards the risk premium you accept by staying invested.
Multiplying lines beyond what is necessary produces dilution: the portfolio becomes unmanageable, gets closer to an index while paying active management fees, and its best convictions are drowned out. On an equity portfolio, most of the reduction in specific risk is captured fairly quickly, and each additional line brings only a decreasing marginal gain.
- How many lines: by way of illustration, a portfolio of individual equities captures most of the diversification benefit at around 20 to 30 well-spread securities; below that, specific risk remains high; well beyond, the gain becomes negligible.
- Correlations: diversify according to what matters (geographic zones, sectors, asset classes), not by the number of securities alone; ten stocks from the same sector diversify almost nothing.
- Home bias: the tendency to overweight one's own country concentrates risk on a single economy and a single currency, whereas international exposure broadens the play of correlations.
Cognitive errors that destroy returns
The main destroyers of return are not the markets, but the investor's cognitive biases. These mental reflexes drive people to buy high, sell low and abandon a strategy at the worst possible moment. Naming them is the first step to opposing them with discipline.
These biases are all the more dangerous because they disguise themselves as common sense. In the moment, overweighting what has just risen or fleeing what has just fallen seems rational; that is precisely how a well-built portfolio deteriorates, one emotional decision after another. The remedy is not the absence of emotion, which is unrealistic, but a written framework (target allocation, rebalancing rules) that takes the decision away from instinct.
- Recency bias: giving excessive weight to recent events, assuming that a rise (or a fall) will continue, and adjusting one's portfolio to the latest trend rather than to one's horizon.
- Anchoring: staying fixed on a reference price, for example the purchase price, and refusing to sell until one is "back to breakeven", instead of judging the position on its prospects.
- Overweighting the familiar: concentrating one's investments on what one believes one knows (one's employer, one's country, one's sector), confusing familiarity with mastery of risk.
- FOMO (fear of missing out): rushing into an asset after a sharp rise, out of fear of missing the move, hence precisely when the risk of reversal is highest.
- Loss aversion: feeling a loss roughly twice as intensely as an equivalent gain, which drives people to sell winning positions too early and hold losing ones too long.
Black swan and tail risk
A black swan, a concept popularised by Nassim Taleb, denotes a rare, unpredictable event with major consequences, which is rationalised after the fact. Tail risk is the risk of these extremes that classical models underestimate. Protecting against it without paying too much for the insurance is a permanent trade-off.
The problem lies in the real shape of returns. Models based on the bell curve predict that violent shocks are almost impossible, whereas they occur far more often: markets have fat tails. A crash that the normal distribution would place at one event every several thousand years occurs, in practice, on the scale of an investor's career. Ignoring this reality amounts to confusing the absence of a recent crisis with the absence of risk.
Protecting yourself has a cost, and that is where the trade-off plays out. Permanently holding insurance against shocks (protective options, a large cash pocket) erodes returns in the years when nothing happens, that is, most of them. Conversely, temporal diversification (spreading your investments over time rather than placing everything at once) reduces the risk of poor timing at no explicit cost. Robustness in the face of extremes is therefore built mainly through the structure of the portfolio, not solely through paid hedges.
Concretely, the useful stance combines three reflexes: sizing your positions to survive an extreme scenario rather than to optimise the average scenario, keeping a liquidity margin that turns a crash into an opportunity instead of a forced-selling constraint, and testing your portfolio against historical shocks. The objective is not to predict the next black swan, by definition unpredictable, but to survive it without being forced to sell at the bottom.