——
Level — · Streak —
— XP
Mastery—weighted
Quizzes 30d—last 30 days
Cert progress—active level
Rank—global
Loading...
The page behind this dialog is live. Create a free account or sign in and you'll land right back on it.
The only free lunch in investing
Imagine two portfolios on January 1, 2000. Both hold five companies; both companies are excellent. Portfolio A holds five technology names — Cisco, Microsoft, Intel, Oracle, and Sun Microsystems. Portfolio B holds Cisco, Procter & Gamble, JPMorgan, ExxonMobil, and Pfizer — same number of names, same average quality, but five different industries. Over the next 30 months, the Nasdaq fell 78%. Portfolio A gave back roughly two-thirds of its value; Portfolio B held within 15% of its starting value. The five-name count was identical; the diversification was completely different. This lesson is about why those two portfolios behaved so differently — and why a portfolio's risk depends much more on what's in it than on how many things are in it.
works mathematically because variance in a combined portfolio depends not just on the variance of each position, but on the between them. When two assets have correlation below 1.0, combining them produces less risk than the weighted average of their individual risks. Harry Markowitz formalized this insight in his 1952 paper 'Portfolio Selection' — for which he eventually won the Nobel Prize — and famously called diversification 'the only free lunch in investing.' Every other return-improving move requires a tradeoff: more return demands more risk, lower fees come from less service, more conviction comes from more concentration. Diversification is the one exception: you can reduce portfolio risk for free, simply by combining assets whose returns don't perfectly track each other.
The trap is that diversification depends entirely on correlations being LOW — and most retail portfolios contain assets that LOOK diversified but actually aren't. Five technology stocks held in 2000 had correlations in the 0.7-0.9 range during the dot-com crash; they all moved together because they all depended on the same underlying drivers (capex on networking, enterprise software spending, internet adoption). The portfolio had five names but effectively one bet. The same trap occurs today with FAANG-style portfolios: Apple, Microsoft, Alphabet, Amazon, and Meta have lower nominal correlations than the dot-com five had, but they still all depend on advertising spend, consumer discretionary income, and cloud capex cycles. When those drivers turn — as they did briefly in 2022, when the cohort fell 30-65% — the portfolio falls together. Effective diversification requires asset classes (stocks, bonds, real assets), geographies (US, international developed, emerging), market caps (large, mid, small), and within equities, GENUINELY different sectors with different cash-flow drivers.
The empirical research on diversification's diminishing returns has been replicated dozens of times since Evans and Archer's seminal 1968 study in the Journal of Finance. The pattern is consistent: a single equally-weighted random stock gives roughly 50% of the diversification benefit available; 10 random stocks gives about 85%; 20 stocks gives about 95%; 30 stocks gives about 97%; beyond that, diminishing returns are essentially zero. The catch is the word 'random' — Evans and Archer drew names without regard to sector or correlation. If you're picking 20 names from the same sector, your effective diversification is much closer to 5-stock random than 20-stock random. The practical rule for active stock pickers: 15-25 names across 5+ sectors and 2+ market caps captures most of the available diversification benefit while staying within a count you can actually research and follow. Beyond 40-50 names, you've effectively built an expensive, tax-inefficient index fund — a real index fund will do the same job for 0.03-0.10% annual fees.
The cap-weighted S&P 500 — the one quoted on the news — weights every stock by its market capitalization. Apple, Microsoft, Nvidia, Alphabet, Amazon, Meta, and Tesla (the so-called 'Magnificent Seven') had grown to represent more than 30% of the index by mid-2024 — meaning the supposedly diversified S&P 500 was actually highly concentrated in seven mega-cap technology-adjacent names with substantially correlated drivers. The equal-weight S&P 500 (tracked by the RSP ETF since 2003) holds the same 500 names but weights each one identically (~0.2% per name), avoiding the mega-cap concentration. Over long periods, equal-weight has typically outperformed cap-weight by approximately 1 percentage point per year (S&P research, 2003-2023, with significant period-to-period variation). The outperformance comes precisely from genuine diversification: equal-weight rebalances away from whatever has run up and toward whatever has lagged, capturing the mean-reversion that cap-weight misses by structurally holding more of recent winners. The catch: equal-weight also has higher volatility (more small/mid-cap exposure) and worse tax efficiency in taxable accounts (more rebalancing turnover). The lesson for individual investors: 'I own the S&P 500' does not necessarily mean 'I'm diversified across 500 stocks.' Check the top-10 weighting before assuming. For a more genuinely diversified equity exposure, combine cap-weight US (VTI), international developed (VXUS), and possibly an equal-weight or small-cap tilt — none individually solves the concentration problem completely. Source: S&P Dow Jones Indices research; Markowitz, Journal of Finance, 1952.
The Portfolio Hub's allocation breakdown shows your exposure across sectors, market caps, geographies, and asset classes — the same view a financial advisor would build. Concentration within any single sector above ~25-30% surfaces a warning. The /screener page lets you find candidates that complement existing exposures rather than duplicating them — filter by sector and the screen will show low-correlation alternatives to your current concentration. The ETF Research module compares roughly 140 ETFs by holding overlap so you can avoid the 'I own SPY and QQQ and FAANG individually' triple-counting trap that destroys ETF-based diversification. The Insights tab on each holding flags significant correlation to other names you own. The reference architecture for a low-effort diversified core: VTI (total US market), VXUS (international), and BND (US bonds) — three funds, total expense ratio under 0.10%, captures roughly 95% of available global diversification. Active stock pickers can layer 10-20 individual names on top of this core to express conviction without sacrificing the floor of diversification.
The most damaging diversification trap is that historical correlations are a poor guide to crisis-period correlations. In a normal market, US stocks, international stocks, REITs, high-yield credit, and emerging-market debt show correlations of ~0.4-0.6 — meaningfully diversifying. During an actual stress event (October 2008, March 2020, the 2022 rate-shock drawdown), nearly all risk assets show correlations of 0.85-0.95 — they all move together because they all depend on the same risk-on/risk-off macro flow. The diversification you THOUGHT you had vanishes in the moment you most needed it. Three structural defenses. First: include genuinely uncorrelated asset classes — high-quality government bonds (Treasuries) historically have NEGATIVE correlation to equities during crises, which is why the classic 60/40 portfolio survived periods like 2008 with ~25% drawdowns vs 50%+ for stocks-only. (The 2022 exception, when bonds and stocks fell together, was the worst year in 100+ years for the 60/40 because it broke this normal pattern — interesting evidence that even the classic diversifier isn't bulletproof.) Second: hold some cash/short-duration as crisis dry powder for rebalancing. Third: size positions assuming crisis correlations of ~0.9, not normal-market correlations of ~0.5. The portfolio that's diversified at the levels the math implies in normal times is roughly half as diversified as it appears once stress arrives. Build for the stress case, not the median case. Ray Dalio's 'All Weather' framework (Bridgewater, 1996-present) is the institutional response to this: diversify across MACRO REGIMES (growth/inflation quadrants), not just across asset classes — different positions chosen specifically because they respond differently to inflation up vs down and growth up vs down.
It is necessary to avoid investing all of one's resources in securities with high covariances among themselves. We should diversify across industries because firms in different industries, especially industries with different economic characteristics, have lower covariances than firms within an industry. The fact that a portfolio of sixty different railway securities, for example, would not have the diversification of one of the same size with some railroad, some public utility, mining, manufacturing, etc. is well known.