Key point
Benartzi and Thaler (2001) called the bias '1/N' after watching 401(k) participants split contributions equally across whatever fund options the plan offered — typically allocating ~1/N to each fund regardless of whether the funds were genuinely diverse. Across plans, the menu drove the outcome: in Benartzi and Thaler's data, a plan whose menu was mostly stock funds produced roughly three-quarters equity allocations, while a mostly-bond menu produced roughly one-third — even though there is no reason pilots and professors differ that much in true risk preference. The portfolio was determined by the menu, not by the investor's actual risk preference.
Compare
| Naive-Diversification View | Correlation-Aware View | Why It Matters |
|---|---|---|
| 20 tech stocks across 5 sub-sectors = diversified | 20 tech stocks have ~0.7-0.9 cross-correlation in a sell-off — effectively one position | The 2022 NASDAQ drawdown took down all 20 simultaneously — diversification was illusory |
| 5 S&P 500 ETFs from different issuers = diversified | All 5 hold roughly the same 500 names — you own the same beta 5× | Net exposure is identical to one ETF; the count creates a false sense of breadth |
| 10 large-cap dividend stocks across sectors = diversified | Most large-cap dividend stocks correlate ~0.6-0.8 in equity drawdowns | Cross-asset diversification (bonds, cash, international, alternatives) does more work than stock-count |
Key point
Markowitz (1952) defined true diversification as adding low-correlation positions: a position correlated 0.3 with the rest of the portfolio reduces variance more than 5 positions correlated 0.8. The math says you need ~15-20 securities to capture most diversification benefit IF correlations are low — but if all 20 securities are large-cap US equities, you have captured the diversification benefit of 1-2 securities and stopped early.
Note
Forget counting positions. Instead, group your portfolio into three buckets: (1) same beta exposure (e.g., US large-cap equity — likely VTI, SPY, individual mega-cap stocks, most US large-cap mutual funds); (2) distinct risk factors (international developed, emerging markets, US small-cap, real assets, commodities); (3) low-correlation diversifiers (high-quality bonds, gold, cash). A portfolio with 30 positions that all live in bucket 1 is undiversified; a portfolio with 5 positions spread across all three buckets is meaningfully diversified. The bucket count matters more than the position count.
Try it
Key insight
Check-in
Check-in
Sit with the ideas.
Benartzi and Thaler's 1/N study (2001) found that 401(k) participants split contributions roughly equally across the funds offered, regardless of those funds' content. Which conclusion is most directly supported?