Talmud thirds
Talmud rule (thirds)
Where it stands today
Our calculation, 1926-07 – 2026-09
Growth of $10,000
Where it came from
The rule comes from the Babylonian Talmud, tractate Bava Metzia, folio 42a. The discussion there concerns where a person should keep money and starts from a verse about binding up money in one's hand. Rabbi Yitzhak says a person's money should always be in his possession, so that he is not left with no cash for spending. He then adds that a person should always divide his money into three parts: bury one third in the ground, invest one third in business, and keep one third on hand (text checked on Sefaria).
The Steinsaltz commentary reads the 'land' third as a hiding place - money buried in the ground - rather than as real estate, and ties the scheme to Deuteronomy 28:8, a blessing on granary, business and land. Another medieval commentary attributed to Rabbeinu Bechaye, on Genesis 32:9, links the idea to risk diversification, with land as the stable long-term holding and business as higher return with more risk. Modern writers prefer to read 'land' as real estate.
In the investment literature the saying was brought into modern asset allocation by Roger Gibson, a non-Jewish adviser who, per Moneycrashers, found it in a book of quotations. His book 'Asset Allocation' (first published 1989, revised 2008) is built around the thirds idea, mapping land to real estate (mainly REITs), business to US stocks and reserve to US bonds (Financial Planning magazine, September 2011).
Academic references are indirect. Several papers on the equal-weight '1/N' strategy mention the Talmudic passage. DeMiguel, Garlappi and Uppal (Review of Financial Studies, 2009, vol. 22 no. 5, pp. 1915-1953) tested 14 optimisation models against 1/N on seven datasets and found none consistently better in Sharpe ratio, certainty-equivalent return or turnover, because estimation error outweighs optimisation gains; they estimated that mean-variance strategies would need roughly 3,000 months of data for 25 assets and about 6,000 months for 50. Whether the paper itself cites the Talmud could not be checked (full text unavailable). Tu and Zhou (2011, CEMA working paper 715; titled 'Markowitz meets Talmud') showed 1/N often beats estimated Markowitz rules out of sample but that combining 1/N with a sophisticated rule beat 1/N in most scenarios.
Note the distinction: the Talmud rule is one-third in each of three asset types (an asset-class rule), while 1/N in the academic literature means equal weight across N assets. The thirds rule is a special case with N = 3.
What is in it
| Land (bury in the ground in the Talmud; real estate in modern readings) | 33.3% | |
| Business (merchandise/trade; modern: stocks) | 33.3% | |
| Cash in hand (reserve; modern: bonds or cash) | 33.3% |
How we model it
| US large companies (S&P 500) | 34% |
| Real estate (REITs) | 33% |
| Cash (Treasury bills) | 33% |
For a euro investor
How to build it with European funds
| Part | Share | Example funds (ISIN) |
|---|---|---|
| US large companies (S&P 500) | 34% | iShares Core S&P 500 UCITS ETF USD (Acc) IE00B5BMR087Vanguard S&P 500 UCITS ETF (USD) Distributing IE00B3XXRP09 |
| Real estate (REITs) | 33% | iShares US Property Yield UCITS ETF IE00B1FZSF77iShares Developed Markets Property Yield UCITS ETF IE00B1FZS350 |
| Cash (Treasury bills) | 33% | Xtrackers II EUR Overnight Rate Swap UCITS ETF 1C LU0290358497iShares USD Treasury Bond 0-1yr UCITS ETF (Acc) IE00BGSF1X88 |
The idea
Spread wealth across fundamentally different holdings, keeping liquid cash for needs. Modern reading: diversify across asset types without forecasting or optimisation.
How people buy it
Modern proxies: a US stock index fund, a REIT index fund, and a bond or money-market fund (Moneycrashers and Gibson's mapping). Gibson Capital expands the idea to seven asset classes.
Rebalancing as published
Not specified by the Talmud. Gibson's results below were reported for the mix without rebalancing; later advisers rebalance annually (not sourced).
Variants
- Gibson's expanded seven-asset-class version (interest-generating vs equity buckets).
- Equal weights over stocks, REITs and bonds (Moneycrashers).
- General 1/N rule over N assets.
Criticism
- Moneycrashers summary: fine for individuals, but an institution with many assets and optimisation tools found Markowitz superior (secondary source).
- A 2013 Brazilian thesis found tail-dependence models and Markowitz had better ex-ante performance than the Talmud rule and the Tu-Zhou model (as relayed by the search result; not read directly).
- Newfeld's study of the 1/3-1/3-1/3 rule found it can beat a minimum-variance counterpart, but rarely (relayed, not read).
- Talmudic text is about holding practice, not return optimisation; 'land' as buried money differs from modern real estate.
Through the eras
| Era | Dates | Total | Per year | Worst drop | S&P 500 |
|---|---|---|---|---|---|
| Roaring Twenties boom | 1926-07 – 1929-08 | +93.6 % | +23.2 % | −4.0 % | +153.7 % |
| Wall Street crash and Great Contraction | 1929-09 – 1932-07 | −60.7 % | −27.4 % | −69.9 % | −78.2 % |
| New Deal recovery | 1932-08 – 1937-02 | +172.6 % | +24.5 % | −29.4 % | +268.7 % |
| Recession within the Depression | 1937-03 – 1938-03 | −40.5 % | −38.0 % | −41.7 % | −49.5 % |
| Rearmament and World War II | 1938-04 – 1945-09 | +155.9 % | +13.3 % | −19.5 % | +205.8 % |
| Postwar expansion | 1945-10 – 1965-12 | +768.9 % | +11.3 % | −17.5 % | +1212.9 % |
| Golden age of the 1950s | 1950-01 – 1959-12 | +300.1 % | +14.9 % | −12.5 % | +438.8 % |
| Go-go years and nifty-fifty | 1966-01 – 1968-12 | +33.9 % | +10.2 % | −10.3 % | +34.0 % |
| The Great Inflation (stagflation) | 1969-01 – 1982-07 | +163.7 % | +7.4 % | −23.7 % | +84.6 % |
| 1973-74 oil shock and bear market | 1973-01 – 1974-12 | −22.0 % | −11.7 % | −23.6 % | −41.7 % |
| Second oil shock and the Volcker disinflation | 1979-01 – 1982-08 | +64.1 % | +14.5 % | −7.7 % | +59.2 % |
| Great bull market in stocks and bonds | 1982-08 – 1999-12 | +746.2 % | +13.0 % | −15.7 % | +2060.4 % |
| Black Monday | 1987-08 – 1987-12 | −11.8 % | — | −15.7 % | −21.8 % |
| Savings-and-loan crisis and 1990 recession | 1990-07 – 1991-03 | +7.8 % | — | −10.3 % | +7.4 % |
| 1994 bond market massacre | 1994-02 – 1994-12 | +0.2 % | — | −3.8 % | −3.2 % |
| Asian and Russian financial crises | 1997-07 – 1998-10 | +8.6 % | +6.4 % | −11.1 % | +22.4 % |
| Dot-com bubble and crash | 2000-03 – 2002-10 | +1.5 % | +0.6 % | −10.4 % | −37.2 % |
| Recovery and housing boom | 2002-11 – 2007-09 | +89.1 % | +13.8 % | −6.7 % | +97.0 % |
| Global financial crisis | 2007-10 – 2009-03 | −37.9 % | −27.2 % | −41.1 % | −44.7 % |
| The decade after the crisis | 2009-04 – 2019-12 | +231.9 % | +11.8 % | −11.9 % | +411.2 % |
| Fourth-quarter 2018 sell-off | 2018-09 – 2018-12 | −7.6 % | — | −7.6 % | −14.4 % |
| Covid crash and stimulus boom | 2020-02 – 2021-12 | +29.2 % | +14.3 % | −14.9 % | +53.6 % |
| Inflation shock and rate hikes | 2022-01 – 2022-12 | −14.5 % | −14.5 % | −17.5 % | −19.9 % |
Calendar years
Any N-year period
| Any N-year period | Worst | 10 % worse than | Median | 10 % better than | Best |
|---|---|---|---|---|---|
| 1 y (1192) | −51.0 % | −6.2 % | +10.8 % | +25.4 % | +131.6 % |
| 3 y (1168) | −30.5 % | −0.3 % | +9.7 % | +17.8 % | +33.5 % |
| 5 y (1144) | −13.1 % | +2.1 % | +9.1 % | +15.9 % | +27.4 % |
| 10 y (1084) | −2.5 % | +5.1 % | +9.5 % | +13.5 % | +16.5 % |
| 15 y (1024) | +1.6 % | +6.5 % | +9.4 % | +13.1 % | +15.3 % |
| 20 y (964) | +3.2 % | +7.0 % | +10.0 % | +12.4 % | +13.8 % |
| 30 y (844) | +7.1 % | +8.1 % | +10.1 % | +11.3 % | +12.6 % |
How it behaved in history
- 2008. Gibson's mix reportedly held up in 2008 mostly because of bonds. www.financial-planning.com
Published results (by others)
- Three-part mix held without rebalancing gained about 5% in the 2000-2002 bear market while a one-third-bond / two-thirds-stock mix lost about 22% (2000-2002): +5% vs -22% www.financial-planning.com
- Optimised portfolios did not consistently beat 1/N across 14 models and 7 datasets (paper published 2009): none consistently better on Sharpe ratio, certainty-equivalent return, turnover ideas.repec.org
- Estimated Markowitz rules often underperform 1/N; combination with 1/N improves them (paper 2011): qualitative econpapers.repec.org
What we could not verify
- Whether DeMiguel et al. (2009) cite the Talmud directly - full text not accessible.
- The name of the Bava Metzia passage attribution date; the compilation date of the Talmud is general knowledge, not sourced here.
- Gibson's 1989 anecdote comes from a secondary summary.
- Newfeld and the Brazilian thesis were seen only through search summaries.
- Claim in Aish that researchers confirmed the rule against Markowitz: no named study.
Sources
- www.sefaria.org
https://www.sefaria.org/Bava_Metzia.42a - steinsaltz.org
https://steinsaltz.org/daf/bavametzia42/ - www.financial-planning.com
https://www.financial-planning.com/fp_issues/2011_9/the-talmud-strategy-2674819-1.html - www.moneycrashers.com
https://www.moneycrashers.com/talmud-portfolio/ - ideas.repec.org
https://ideas.repec.org/a/oup/rfinst/v22y2009i5p1915-1953.html - econpapers.repec.org
https://econpapers.repec.org/RePEc:cuf:wpaper:715 - aish.com
https://aish.com/ten-timeless-rules-of-investing-from-the-talmud/
Calculated on historical data. Past results do not predict the future. Models before the 1990s use stand-ins for classes that had no index yet; fees, taxes and fund costs are not included. This is mathematics, not an investment recommendation.