Diversification is one of the most widely repeated principles in investing. It is also one of the most widely misapplied. The common interpretation — hold many assets and you are diversified — misses the point almost entirely.
Genuine diversification is not about the number of positions. It is about the independence of the risks those positions represent.
The Problem with Counting
A portfolio holding 100 stocks in the same sector is not diversified. When that sector corrects, every position moves together. The investor has the appearance of diversification — a large number of holdings — but very little of the substance. The risks are correlated; the outcomes cluster.
This distinction matters enormously during market stress. Correlations between assets tend to rise sharply when markets fall. The diversification investors thought they had — built on normal-market correlation estimates — compresses precisely when it is most needed. Many portfolios that looked well-spread in calm conditions turned out to be highly concentrated in practice.
What Diversification Actually Requires
True diversification requires sources of return that behave differently from one another — not just different names, but genuinely different exposures to different economic risks.
In quantitative terms, the concept is captured by covariance: the degree to which two assets move together. A portfolio is well-diversified when the covariance between its components is low. Adding a new position only improves diversification if it introduces a source of risk that is not already present.
This is why covariance estimation is one of the most important and most difficult problems in portfolio construction. A flawed estimate of how assets co-move leads directly to portfolios that are riskier than they appear.
How We Approach It
At KB Asset Management, we recognize that no single method for estimating how assets co-move is reliable in all market conditions. The raw statistical approach — estimating covariance directly from historical return data — works reasonably well in stable periods but produces noisy and unstable results over shorter horizons or during market transitions.
We apply techniques designed to reduce this estimation noise and produce more stable representations of the underlying risk structure. The goal is not to optimize against a particular market regime but to build a portfolio whose risk is genuinely distributed across different economic exposures — rather than merely distributed across a large number of tickers.
Achieving this requires judgments about what the data is actually telling us versus what is simply noise. That distinction cannot be resolved by counting positions.
The Practical Implication
Building a diversified portfolio requires accepting that some of the most effective diversifiers look uncomfortable in isolation. Assets that move differently from the rest of the portfolio — and therefore provide genuine risk reduction — may underperform in trending markets. The temptation is to remove them. That impulse is exactly what undermines diversification.
Diversification is only effective if it is maintained consistently. A systematic process holds its structure through discomfort. That consistency is the mechanism through which diversification delivers its benefit over time.
Past performance is not indicative of future results.