Invest for what you keep
Subscribe

Choose · what to own · Decision 01 of 03 · Asset allocation

Three books for three purposes, and a house counts for a fifth of its value

Pick what the money is for; the purpose sets the book of stocks, Treasuries and gold. Count the house at a fifth of its value and let the liquid book complete the target. Every figure here is before tax, from two starts, 2006 and 2010.

3booksOne per purpose of the money: 65 / 10 / 25, 50 / 20 / 30, 45 / 25 / 30 in stocks, Treasuries and gold, rebalanced once a year.
0.72–0.78Return per unit of risk since 2006, against 0.67 for 60/40; from 2010, 0.90–0.92 against 0.87.
−13% to −14%In 2022, the year stocks and Treasuries fell together, against −17% for 60/40. Gold fell 1%.
0.2× valueWhat a house adds to the stock count. The sample household's savings hold 56% stock on the Balanced FI book, not 50%.
The count

Pick what the money is for, then count the house

The purpose sets the book and a target share of net worth in stocks. A house supplies a fifth of its value toward that target; the savings hold the rest. The sample household's numbers are a starting point, not yours.

01What is the money for?

The Balanced financial independence bookstocks / Treasuries / gold, rebalanced once a year · target 50% of net worth in stocks
Stocks 50%Treasuries 20%Gold 30%
$694,000what $100,000 became, 2006 to 2026; 60/40 made $523,000from 2010: $533,000 against $469,000
−$28,000the worst fall per $100,000 at the peak; 60/40 fell $30,000from 2010: −$19,000 against $21,000
−13%in 2022, when stocks and Treasuries fell together; 60/40 lost 17%in 2008: −14% against −16%
0.77return per unit of risk since 2006; 60/40 0.67from 2010: 0.92 against 0.87

02Count your home

Start from
Sample households; every field below is yours to change. Nothing you type leaves this page.

Net worth $1,600,000: home equity $700,000 (44%) and savings $900,000. At beta 0.2 the house supplies $300,000 of stock exposure against a Balanced financial independence target of $800,000, 50% of net worth. The savings hold the other $500,000: 56% of the liquid book in stocks, against 50% for the book alone.

The liquid book, completedwhat the savings hold, Treasuries and gold in the book's own ratio; the Balanced financial independence book alone is 50 / 20 / 30
Stocks 56%Treasuries 18%Gold 27%
$166,000the reserve: 12 months of spending and payments, in bills, outside the bookpayment $3,794 a month
surviveshome equity after a 30% fall in the house: $250,000leverage 2.1x on equity
keepthe loan costs 3.00%, less than a 10-year Treasury pays (4.68%): a spare dollar goes to the book10-year Treasury yield, FRED DGS10, 2026-08
out of the moneythe going rate (6.95%) is not a point below the coupon30-year fixed, Freddie Mac via FRED, 2026-09-17

The books: alphalab household_aa campaign on ETF total returns (SPY, IEF, GLD; TLT, VNQ and BIL/SHY as references; bitcoin from FRED CBBTCUSD), rebalanced once a year for the books, monthly for the tests; nominal, gross of costs and taxes. The house: asset-allocation research, house_model (v3, 2026-09-24), beta 0.2 ruled 2026-09-24 on a range of 0.1 to 0.42. Before costs and taxes; Implement shows what is kept. Arithmetic, not advice: the presets are illustrations a reader replaces with their own numbers.
What the allocation is worth

$100,000 in the Balanced FI book became $694,000 since 2006; a plain 60/40 made $523,000

From a 2010 start, which skips 2008, the gap narrows to $533,000 against $469,000. What both starts agree on: each book returned more per unit of risk than 60/40, fell less in 2022, and Balanced FI and Legacy fell less at the worst.

Two panels of horizontal bars. Top: $100,000 put in became, by 2026, $804,000 in Grow wealth, $694,000 in Balanced financial independence, $644,000 in Leave a legacy and $523,000 in a plain 60/40 from a 2006 start; from a 2010 start, $660,000, $533,000, $488,000 and $469,000. Bottom: the worst fall per $100,000 held at the peak, $35,000 to $25,000 for the books against $30,000 for 60/40 from 2006, and $22,000 to $18,000 against $21,000 from 2010.
Stocks (SPY), Treasuries 7-10y (IEF) and gold (GLD), rebalanced each January; 60/40 the same two funds. 2006 to 2026 and 2010 to 2026, nominal, before costs and taxes. Rounded to the nearest $1,000.
+1.1 to +2.3pts a yearThe books over 60/40 from 2006; +0.3 to +2.3 from 2010. The return gap depends on the start; the risk-adjusted direction did not.
35%Grow wealth's worst fall from 2006, four points deeper than 60/40's 30%: what 65% stocks costs in a 2008, and the loss that lifestyle signs up for.
4–5%a yearTurnover of the once-a-year books, against 12 to 14% rebalanced monthly, for a return within 0.2 points. In a taxable account that difference is Implement's subject.
Two bar panels of calendar-year returns. 2008: stocks −37%, Treasuries +18%, gold +5%, 60/40 −16%, the three books −21%, −14% and −11%. 2022: stocks −18%, Treasuries −15%, gold −1%, 60/40 −17%, the books −14% to −13%.
Calendar-year total returns, 2008 and 2022. 2008 was a stock crash: Treasuries rose 18% while stocks fell 37%. 2022 was a rate shock: Treasuries fell 15% with stocks, and gold fell 1%.

Why these weights. An optimizer fed the 2006 window wants 43 to 45% gold and no Treasuries; fed the 2010 window it wants 70 to 85% stocks and no Treasuries. Gold at 15 to 30% is in both answers; the Treasury weight is the leg the two decades disagree on, whose only job is the 2008 kind of year and whose argument today is a 4.7% starting yield. The books were chosen to be acceptable from both starts, not optimal in either, and any weight can move five points without leaving the noise.

60/40 is the floor, and three books clear it from both startsBitcoin is sized by the loss a household can hold

Research · what stays out

Why no commercial real estate, and no long Treasuries: neither paid for its risk from either start

Listed commercial real estate returned 6.1% a year since 2006 for more risk than stocks and a 73% fall; long Treasuries 2.8% for stock-sized swings and a 48% fall. Neither improved a 60/40. A household that owns a house already holds property.

Two scatter panels of return against volatility for six assets, from 2006 and from 2010. Long Treasuries sit at 2.8% and 2.4% a year for 15% volatility and a 48% worst fall; listed real estate at 6.1% and 8.6% for more volatility than stocks and a 73% fall from 2006. Both sit below the line from the cash point through stocks, gold and Treasuries in both panels.
Compound annual return against volatility, the six funds, 2006 to 2026 and 2010 to 2026. The gray lines from the cash point are equal return per unit of risk. Bitcoin, since 2016, is off the scale and noted in the panel.
0.59from 0.64Return per unit of risk of a 60/40 once ten points of listed real estate are carved from the stock sleeve, from 2006; 0.81 from 0.86 since 2010. A second, more volatile equity, called diversification.
−23%A 60/40 in 2022 with long Treasuries in place of intermediate, against −17%: 6.7% points paid for a swap that changed the full-window return from 8.2% to 8.3%.
−73%The deepest fall of the six from 2006, listed real estate in 2008 and 2009, with a beta to stocks above one. From 2010 it fell 42% for 8.6% a year against stocks' 14.2%.

Long Treasuries and listed real estate do not pay for their risk

Research · the house

How to view your home: stock exposure at a fifth of its value, and a loan that is leverage, a rate and an option

On annual returns a Los Angeles house moved with the stock market at a beta of 0.23 over 1988 to 2026 and 0.42 over 2006 to 2026, a window dominated by 2008. The working number is 0.2: a $1,500,000 house supplies $300,000 of stock exposure. The house does not change the book's risk; leverage sets the reserve, and the loan's rate decides prepaying.

Three stacked horizontal bars, one per lifestyle, for a household with a $1,500,000 house, a $800,000 loan and $900,000 of savings: the house supplies $300,000 of stock exposure in every row, and the liquid book adds $740,000 for Grow wealth (82% of the book against 65% for the book alone), $500,000 for Balanced financial independence (56% of the book against 50% for the book alone), $420,000 for Leave a legacy (47% of the book against 45% for the book alone).
The sample household: a $1,500,000 house, a $800,000 loan at 3%, $900,000 of savings; net worth $1,600,000, house equity 44% of it. The lifestyle's target applies to net worth; the savings are a smaller share of it, so they hold more stock than the plain book, not less.
56%not 50%Stocks in the sample household's savings on the Balanced FI book, once the house is counted. Grow wealth: 82%; Legacy: 47%.
17%not 4%The house's true volatility once the price index's smoothing is removed, for 4.8% a year unlevered since 2006. A number the household never lives through: the house is never marked and, with a reserve, never sold into a fall.
13%vs wiped outThe same house, 20% down: bought in January 2010, 13% a year on equity and never underwater; bought in January 2006, the equity gone in 2008 and 2.6% a year on the cash put in. Leverage decided how much; the entry decided the sign.
Scatter of return against volatility: cash, Treasuries, the Balanced FI book, gold and stocks, and a Los Angeles house without a mortgage at 4.8% a year, shown twice: at the 4% volatility the price index reports and at the 17% true volatility once the smoothing is removed. A note records the same house bought in 2006 (equity wiped out in 2008) and in 2010 (13% a year on equity).
Monthly, January 2006 to June 2026. House: Case-Shiller Los Angeles price change plus net rent (4% gross yield, 25% costs); true risk by Geltner's reverse filter. Buyers: 30-year fixed, refinance allowed at 80% loan to value.

The loan. A fixed-rate mortgage cannot be sold, so it has no market value to the household. It is leverage, which sizes the reserve (12 months of spending and payments in bills for Balanced FI, $166,000 for the sample household) against a 30% fall in the house; a fixed rate to compare with the 10-year Treasury (4.7%: the 3% loan stays, a 6.5% loan is prepaid first); and a refinance option, in the money a point below the coupon with the loan under 80% of value, out of the money for both sample loans at a 6.95% going rate.

The house is stocks, at a fifth of its valueThe mortgage is leverage, a rate and an option

Research · the diversifier

The power of gold: the one asset that held in 2008 and in 2022, and the reason a book holds it at 30%

A portfolio of stocks and Treasuries has one hedge, and it works when the shock is a growth shock. When the shock is inflation and rates, both halves lose. Gold rose 5% in 2008 and fell 1% in 2022; since 1972 it was the best of the three assets in the 2 regimes of 4 where stocks and bonds both lost. The weight is a view on the decade; the two bad years are evidence.

Four panels, one per economic regime, each with the real annualized return of stocks, bonds, gold and the 60/30/10 blend, 1972 to 2026. Stocks win the two below-trend-inflation regimes; gold wins the two above-trend regimes, where stocks and bonds both lose; the blend is never the worst of the four in any panel.
Real returns by regime, 1972 to 2026: industrial production up or down over twelve months (FRED INDPRO), inflation above or below its three-year average (FRED CPIAUCSL). Inflation ran above trend in 44% of months. Shiller stocks and yields, LBMA gold.
15 to 30%The gold weight both starts agree on: an optimizer wants 43 to 45% from 2006 and 15 to 25% from 2010. The stock and Treasury weights swing with the start; gold's does not.
+1.5pts a yearWhat 10% gold did to a 60/40's worst decade since 1972, real (−2.3% against −3.8% a year), at +0.4 points of return.
10.2%a yearGold's worst real decade since 1972. It lost 62% from 1980 to 1999 and regained the 1980 peak in 2007. A backtest that starts in 2006 holds only the decades gold won; the sleeve is held for the regime, not the price.

Gold is a view on the decade, not a backtest result

The choice

Hold all three. Pick the book by what the money is for, not by a risk questionnaire: Grow wealth 65 / 10 / 25, Balanced financial independence 50 / 20 / 30, Leave a legacy 45 / 25 / 30, rebalanced once a year. Count the house at a fifth of its value and let the savings complete the target; the loan sets the reserve and the prepay test, not the mix. Long Treasuries and listed real estate stay out. Gold at 25 to 30% is a view that the coming decade rhymes with 2022 more than with 1995, and the page labels it as one.

Book weights are provisional (any can move five points without leaving the noise); the house beta of 0.2 is a ruling on a range of 0.1 to 0.4; bitcoin, capped at 3 / 2 / 0% by lifestyle and rebalanced on its own band, added +2.1 points a year to Grow wealth since 2016 for a worst fall +0.3 points deeper. Nothing here is a view on where any asset goes next.

For practitioners: the books on six measures from both starts, the time series, every mix, the completion curve, and the 1972 record
BookStartReturn a yearVolatilitySharpeWorst fall20082022$1 became
Grow wealth65 / 10 / 25 2006 10.7% 13.1% 0.72 −34.5% −21.4% −13.8% $8.04
2010 12.0% 11.8% 0.90 −22.2% — −13.5% $6.60
Balanced financial independence50 / 20 / 30 2006 9.9% 11.0% 0.77 −27.7% −13.8% −12.6% $6.94
2010 10.6% 10.0% 0.92 −18.5% — −12.3% $5.33
Leave a legacy45 / 25 / 30 2006 9.5% 10.2% 0.78 −25.0% −11.1% −12.5% $6.44
2010 10.0% 9.4% 0.92 −18.2% — −12.2% $4.88
60/40stocks and Treasuries only 2006 8.4% 10.5% 0.67 −30.2% −15.8% −17.1% $5.23
2010 9.7% 9.7% 0.87 −21.4% — −17.0% $4.69

Annual rebalancing, 2006 to 2026 and 2010 to 2026. The Sharpe standard error over twenty years is about 0.22, so the gaps to 60/40 are inside noise in both windows; the claim is the direction, held in both windows tested, not the size. Monthly rebalancing changes return by 0.0 to 0.2 points and moves the worst fall by about a point; no ranking changes.

Growth of one dollar on a log scale, 2006 to 2026, for the three books and 60/40, ending at $8.04, $6.94, $6.44 and $5.23; beneath, the fall from the peak, deepest in 2008 and 2009 at −35% for Grow wealth and −30% for 60/40.
Growth of $1 on a log scale and the fall from the daily peak, month-end values, 2006 to 2026.
Scatter of every 5% mix of stocks, Treasuries, long Treasuries, listed real estate and gold from 2006, return against volatility, with the best-possible line along the top edge. The three books sit just under the line; 60/40 sits below it at 8.2% a year; thirds sits on the line at 8.6%.
Every 5% mix of stocks, Treasuries, long Treasuries, listed real estate and gold, rebalanced monthly, 2006 to 2026. The best-possible line never holds listed real estate and never more than 3% long Treasuries.
Three rising curves: the share of the liquid book held in stocks against house equity as a share of net worth, for each lifestyle at a house beta of 0.2 with a band from 0.1 to 0.3. All three rise with the house share; the sample household, at 44%, is marked on each.
Stocks in the liquid book against house equity as a share of net worth, house levered 2x, at beta 0.2 with a 0.1 to 0.3 band. Dashed: the book's own stock weight.
Scatter of every 5% mix of stocks, bonds and gold, 1972 to 2026: real annualized return against the worst ten-year real return. The stocks-and-bonds-only line runs from 100% stocks (6.9% a year, worst decade −5.9%) through 60/40 (5.3%, −3.8%); the same mixes with 10% gold sit above and to the right of it, 60/30/10 at 5.6% and −2.3%.
The 1972 record: every 5% mix of Shiller stocks, 10-year Treasuries and LBMA gold, real, 1972 to 2026, scored on return and on the worst decade. Of the 231 mixes, 77 contain gold and beat 60/40 on both scores.
Strategy cards

The 2 cards behind this decision

The routes that carry them
  1. The 60/40 mixEvidence supports 60/40 is the floor, not the answer: three books of stocks, Treasuries and gold returned more per unit of risk from both starts tested and lost less in 2022; the return gap depends on the start, the direction did not. 3–40 bps a year · mixed · The stock-bond mix alone swings twenty-year returns by up to three points a year
  2. GoldEvidence supports Gold earns its place because it held in the year stocks and Treasuries fell together; its weight, 25 to 30% in the books, is a view that the coming decade rhymes with 2022 more than with 1995, and a sleeve that can lose for twenty years while it waits. 10–40 bps a year · mixed · card only, read coming

A card states what a strategy is, its evidence with period and source, when it worked and failed, what access costs, its tax profile, capacity and behavior demand, and a verdict. Open cards carry no evidence yet and are published only when filled.

Which account holds each piece is the next section's job: Account placement, under Implement.

7 short reads in this decision, about 34 minutes, below.

Start here3 reads · 16 min

The few ideas everything else rests on.

  1. The house is stocks, at a fifth of its value

    A house moves with the stock market at a beta near 0.2, so a $1.5M house supplies $300,000 of stock exposure. Counted, the savings of a household with a house hold more stock than the plain book.

    6 min

  2. The stock-bond mix alone swings twenty-year returns by up to three points a year

    Since 1985, moving from an all-equity portfolio to 60/40 has changed trailing 20-year annualized returns by as much as 3.1 points a year, before any strategy or wrapper choice.

    5 min

  3. 60/40 is the floor, and three books clear it from both starts

    Three books of stocks, Treasuries and gold returned more per unit of risk than 60/40 from a 2006 start and a 2010 start, and lost less in 2022. The return gap depends on the start; the edge does not.

    5 min
Concepts used here
Basis pointBetaDrawdownGlide pathHolding periodLeverageRebalancingTax drag

How it works4 reads · 18 min

The mechanics and the numbers behind them.

  1. Long Treasuries and listed real estate do not pay for their risk

    From 2006 and from 2010, long Treasuries returned under 3% a year for stock-sized swings and listed real estate returned less than stocks for more risk. Neither improved 60/40.

    5 min

  2. Gold is a view on the decade, not a backtest result

    Gold held in 2008 and 2022 and sits at 15 to 30% in the optimizer's answer from a 2006 start and a 2010 start. It also lost 62% from 1980 to 1999. Its weight is a view on the next decade.

    4 min

  3. Bitcoin is sized by the loss a household can hold, not by the optimizer

    Three percent of bitcoin added two points a year to the Grow wealth book since 2016 for a worst fall a third of a point deeper. The optimizer wants 69%. Once a year cannot hold the cap.

    4 min

  4. The mortgage is leverage, a rate and an option, and it has no price

    A fixed-rate mortgage cannot be sold, so it has no market value to the household. It is leverage on the house, a fixed rate to compare with today's, and a refinance option that pays when rates fall.

    5 min
Concepts used here
BetaDrawdownHolding periodLeverageRebalancingTerm structure
Concepts

Terms this topic relies on

All concepts
BetaDrawdownGlide pathLeverageRebalancing