Short answer
Correlation measures how two assets' returns move together on a scale from minus one to plus one: plus one is perfect lockstep, minus one perfect opposition, zero no linear relationship. Blending imperfectly correlated assets lowers portfolio risk more than expected return, Markowitz's near free lunch, but correlations rise toward one in a crisis and over-diversifying dilutes an edge.
This article explains correlation and diversification as concepts. It is educational and general, not personalised investment advice, and nothing here is a recommendation to buy, sell, size, hedge, or allocate any position. Where it describes statistical behaviour, that is a widely-observed tendency, not a precise claim about any single asset or period.
Diversification is the one idea in finance almost everyone has heard and almost everyone half-understands. “Don’t put all your eggs in one basket” is the folk version, but the real insight is sharper and stranger: combining assets that do not move in perfect lockstep can lower a portfolio’s risk by more than it lowers its expected return, something close to a free lunch in a field that rarely offers one. Here is what correlation actually measures, how diversification lowers risk through it, why the effect weakens exactly when it is needed most, and whether any of it helps a deliberately concentrated momentum book.
The short version
Correlation measures how two assets’ returns move together, on a scale from minus one to plus one: plus one is perfect lockstep, minus one is perfect opposition, and zero is no linear relationship. Diversification exploits imperfect correlation: blending assets that are less than perfectly correlated lowers a portfolio’s risk more than it lowers its return, which is why Markowitz called it the closest thing to a free lunch in investing.
What correlation is
Correlation is a single number that summarises how two return series move relative to each other. Formally it is the covariance of the two returns divided by the product of their standard deviations, which rescales the relationship onto a clean, unit-free axis running from minus one to plus one. That rescaling is the useful part: it strips out how big each asset’s swings are and leaves only how together they swing, so two very different assets can be compared on the same footing.
Read the scale in three anchors. A correlation of plus one means the two assets move in perfect proportion: when one is up a given amount, the other is reliably up in step. Minus one means perfect opposition: one zigs precisely as the other zags. Zero means no linear relationship at all, the two wander independently. Most real pairs of stocks sit somewhere in the positive middle, moving loosely together because they share exposure to the same economy, the same market, the same interest rates, but rarely in perfect lockstep. Two caveats matter. Correlation captures only linear co-movement, so it can miss a real but curved relationship, and it is famously not causation: two series can correlate strongly by coincidence or because both respond to a hidden third thing.
How diversification lowers risk
The reason correlation matters to an investor is that it governs how risks combine. Expected return combines simply: a portfolio’s expected return is just the weighted average of its holdings’ expected returns. Risk does not. A portfolio’s variance depends not only on each asset’s own volatility but on how the assets co-move, and that cross term is driven by correlation. When two assets are less than perfectly correlated, some of their individual wobbles offset: one is having a bad day while the other is flat or up, and the blend is steadier than either piece alone.
This is the asymmetry that makes diversification powerful. Combine two assets and the portfolio’s expected return lands on the straight line between them, but its risk lands below the straight line whenever their correlation is under plus one, and the lower the correlation, the deeper below. You keep a weighted-average return while shedding more than a weighted-average share of the risk. Harry Markowitz formalised this in his 1952 work on portfolio selection, the foundation of modern portfolio theory, and the popular gloss of the result is exactly the one worth remembering: diversification is the only “free lunch” in investing, a genuine reduction in risk that does not have to be paid for with a matching reduction in expected return. The effect is strongest with low or negative correlations, real but weaker with moderate positive ones, and vanishes entirely at plus one, where two assets are effectively the same bet wearing two names.
How correlation shapes the blend of two equally-volatile assets:
| Correlation | How they move | Effect on portfolio risk |
|---|---|---|
| +1.0 | Perfect lockstep, one bet in two names | No reduction: risk is the weighted average, nothing offsets |
| +0.5 | Loosely together, the typical stock-to-stock case | Meaningful reduction: some wobbles cancel |
| 0.0 | Independent, unrelated drivers | Large reduction: risk falls well below the average |
| -0.5 | Tend to lean opposite ways | Very large reduction: losses in one often met by gains in the other |
| -1.0 | Perfect opposition | Risk can be driven to zero at the right weights (with equal volatilities, equal weights, a theoretical ideal) |
The same logic scales past two names. Spread capital across many holdings whose company-specific surprises are unrelated, one firm’s bad earnings, another’s lawsuit, a third’s lucky contract, and those idiosyncratic shocks increasingly wash out against each other, leaving mostly the market-wide movement that no amount of diversifying can remove. That split is the heart of it: total risk divides into idiosyncratic (specific to a name, diversifiable) and systematic (shared across the market, not diversifiable). Diversification is, precisely, the removal of the first kind.
The limits: when the free lunch gets expensive
Diversification is real, but it comes with two honest catches a careful reader has to hold in mind, and both are routinely soft-pedalled.
The first is the cruel one: correlations rise toward one in a crisis, exactly when the diversification is needed most. In calm markets a basket of loosely-correlated stocks behaves as the theory promises. But in a genuine panic, a liquidity crunch, a systemic shock, a forced-selling cascade, investors dump nearly everything at once to raise cash, and assets that normally drift apart suddenly plunge together. Correlations that averaged, say, a mild positive number in normal times can spike close to plus one in the worst weeks. The diversification you were counting on thins out precisely at the moment of maximum stress, which is why a portfolio that looks well-spread on a long-run correlation table can still take a coordinated hit when a true crisis arrives. Diversification protects against the idiosyncratic and the ordinary far better than against the systemic and the extreme.
The second catch is subtler: over-diversifying can dilute an edge. Each name you add removes a little more idiosyncratic risk, but the benefit shrinks fast, the bulk of the available risk-reduction is captured by a relatively modest number of well-chosen, loosely-correlated holdings, and beyond that point new positions mostly just pull the portfolio toward the market average. Someone who genuinely believes they can identify better-than-average opportunities pays for excessive diversification in forgone return: dozens upon dozens of positions increasingly recreate the index they were trying to beat, a phenomenon sometimes called “diworsification.” There is a real tension here between spreading risk and expressing a view, and it is the exact tension a concentrated strategy chooses to live inside. How many holdings actually captures most of the benefit is its own study, taken up in how many positions to hold.
Does diversification help a concentrated momentum book?
This is where the concept gets genuinely interesting, because a momentum strategy looks, at first glance, like the opposite of diversification. Concentration is a deliberate bet: a momentum book crowds into a small number of the strongest-trending names on purpose, because spreading too thin would dilute the very relative strength it is trying to harvest. Diversifying that away completely would defeat the point. So the naive reading is that momentum and diversification are at war.
The truer picture is that they operate on different layers. Even a concentrated book benefits from diversifying within the concentration, across names, sectors, and themes whose company-specific risks are unrelated. Holding a handful of strong movers that all happen to be the same theme, the same reaction to one macro driver, the same crowded trade, is not really a concentrated portfolio of several bets; it is one bet held several times, a correlation quietly near plus one. Spreading the same conviction across genuinely uncorrelated leaders keeps the intended concentration while still stripping out the idiosyncratic, single-name risk that no strategy is paid to carry. Concentration governs how many bets and how large; diversification, applied inside that, governs whether those bets are actually distinct. The two are complements, not opposites, once you see that correlation, not the raw count of holdings, is what determines whether risk is really being spread.
How a systematic process relates to this
A rules-based momentum publisher sits squarely on this tension, and the honest way to describe where it lands is through evidence rather than assertion. Shishin’s published track-record analysis includes an “edge-versus-beta” study that speaks directly to correlation and diversification at the portfolio level. Two findings are worth stating at a high level. First, the book’s market beta comes out close to nil, which is a way of saying its returns are not strongly correlated with the broad index: the strategy is not simply a levered version of the market moving in step with it. Second, the returns read as largely idiosyncratic, driven by the specific names the process selects rather than by riding the same systematic wave everyone else is on. In correlation terms, a near-zero market beta and an idiosyncratic return stream are the signature of a book whose behaviour is its own, not a repackaged index.
That is the descriptive connection, and it is a factor-analysis point, not a pitch. The same regression that measures a fund’s correlation to the market and to known return drivers is the one behind that edge-versus-beta finding, the broader framework for which is set out in factor investing, where the size premium and the other systematic tilts live. A concentrated momentum process does not diversify away its edge; the evidence that it is holding a genuinely idiosyncratic book, rather than a crowded, index-correlated one, is exactly what a low market beta and an independently verified record are meant to demonstrate.
So, does correlation-driven diversification help?
Yes, with a clear-eyed view of what it can and cannot do. Diversification is the one near-free lunch in finance: combining imperfectly correlated assets lowers portfolio variance by more than it lowers expected return, and the lower the correlation, the larger the gift, the point Markowitz made and the reason the idea has endured. But it protects against the diversifiable and the ordinary, not the systemic and the extreme, because correlations rush toward one in a crisis, and it can be overdone to the point of diluting a real edge into an expensive index. For a concentrated momentum book the resolution is not to abandon concentration but to make sure the concentrated bets are genuinely distinct: correlation, not the number of names, is what decides whether risk is actually being spread or merely reprinted.
Sources & further reading
- Markowitz, H. (1952). “Portfolio Selection.” Journal of Finance, 7(1), 77-91, the founding statement of diversification and the risk-return trade-off in modern portfolio theory.
- Ang, A. Asset Management: A Systematic Approach to Factor Investing, a standard reference on how correlations, factors, and diversification behave in practice, including the rise of correlations in crises.
- On the systematic tilts a diversified book is really exposed to, see factor investing; on how many holdings actually capture most of the benefit, how many positions to hold.
- For how a near-zero market beta and an idiosyncratic return stream read in a real record, see Shishin’s track-record analysis and its independently verified attestation.