Diversification Is a Discipline, Not a Number

Ask a room of investors how diversified a portfolio is and they will answer with a count: the number of positions, the number of markets, perhaps a correlation figure carried to two decimal places. The count feels like knowledge. It has the great virtue of being available, and the great defect of being a photograph: a record of relationships as they stood on the day of measurement, presented as if it were a property of the portfolio itself. Diversification is not a property. It is a relationship among parts, relationships change, and they change fastest in exactly the conditions diversification exists for. A number cannot hold still long enough to be the thing it names.

The point can be made without any crisis at all. A portfolio can add markets and become less diversified: add a dozen new positions that all rise and fall with the same underlying driver, size them generously because each one looks small, and the portfolio now has more lines, more apparent breadth, and a larger single bet than it had before. Conversely, a portfolio with few holdings, sized so that no one driver can dominate its fate, can be genuinely difficult to hurt. The count moved in the wrong direction in both cases. What actually decides the outcome is not how many things are held but how the losses arrive: together, in one correlated convulsion, or separately, in instalments the portfolio can absorb. Diversification is a claim about the second of those patterns, and a count of positions is, at best, weak evidence about it.

The standard counterargument is mathematical, and it is correct as far as it goes. The arithmetic of breadth and the basic arithmetic of portfolio variance both say that breadth helps: combine genuinely independent bets and the risk-adjusted result improves, roughly with the square root of the number of bets. We do not dispute the arithmetic. We dispute the premise smuggled inside the word "independent." Independence is not a fact about two return series that, once measured, stays measured. It is a condition that holds in some regimes and fails in others, typically in the regimes where it is most needed, because stress creates common causes: common funding, common margin clerks, common fear. Correlations estimated in calm weather describe calm weather. The formula tells you what diversification would be worth if the independence you fed into it survived. Discipline is the work of keeping the portfolio honest about whether it has.

That work is unglamorous and mostly consists of refusals. It is sizing by risk rather than by conviction, so that the positions a process loves do not quietly become the portfolio. It is rebalancing on a fixed schedule, which in practice means systematically selling what has grown and buying what has shrunk, against every instinct and every recent headline. Above all it is refusing to let winners concentrate. Every portfolio drifts toward its own undoing through success: the position that performed becomes the largest position, the largest position becomes the firm's fate, and the diversification that existed at inception has been spent, not lost, exchanged for the pleasant feeling of having been right. Nobody decides this. It happens by default, which is why it must be prevented by rule. In the fund we are building, which is designed to invest systematically across futures markets, diversification is therefore not something to be reported after the fact. It is something the rules are written to maintain continuously: risks sized to a budget, exposures rebalanced on a discipline rather than on a mood, and concentration treated as a cost that compounds silently and is collected all at once.

The drift is easy to underestimate, so it is worth doing the arithmetic once. Take an illustrative portfolio that starts with half its risk in one driver and half in another. Suppose the first doubles in value and the second goes nowhere. No one has made a decision, yet the first driver now accounts for two-thirds of the capital, and because risk scales with the size of the position, an even larger share of the portfolio's risk. A further doubling would take it to four-fifths. At no step did anyone choose concentration; each step was simply the absence of a choice. This is why we think of rebalancing not as a trading activity but as a decision that must be taken on a schedule precisely so that it cannot be postponed by how the portfolio feels. A rebalance deferred because the winner is still winning is a concentration decision made by default, and default is the one decision-maker we do not let near the portfolio.

There is a deeper reason to insist on the distinction, and it is about honesty in reporting, including to ourselves. A number can be audited once and quoted forever, and an entire industry has built its reassurance on quoted diversification: so many positions, such a low average correlation, such a smooth simulated path. Each figure may be accurate and the reassurance false, because the figures describe the portfolio's past configuration, while the risk lives in its future behaviour. A discipline, unlike a number, can be inspected in the present tense. Are positions being sized by a rule or by enthusiasm? Was the last rebalance done on schedule or deferred because it felt wrong? Has any single driver been allowed to grow past its budget because it was winning? These questions have answers today, not just in the post-mortem. A firm that can answer them is diversified in the only sense that will matter on the bad day. A firm that can only quote its count is diversified in the sense that a photograph of a bridge is a bridge.

We do not want to be misunderstood as arguing against breadth. Breadth is where diversification starts, and our research is built on the value of many modest, genuinely different effects over one brilliant one; a companion essay in this Library takes up breadth and its fine print directly. The argument here is about what breadth needs in order to still be there when it is called on. Markets will do what they do: regimes will turn, correlations will converge, and the portfolio will discover what it actually holds. The discovery is pleasant exactly in proportion to the maintenance that preceded it. Diversification is not the number of things you own. It is the practice that keeps them different, kept up on the ordinary days, when nothing is testing it and skipping it feels free. Like most disciplines, it is cheap every day except the one it is for.