Portfolio Building: The Mix Decides It
Portfolio building is deciding what proportion of a total sits in each kind of asset, and why. The individual selections matter far less than the mix. Diversification means holding things that behave differently, not holding many things, and correlations tend to rise in a crisis.
How it works
Portfolio building is deciding what proportion of the whole sits in each kind of asset, and why. The individual selections matter far less than the mix, which is the part most people settle last.
The mix explains most of the result. Which shares sit inside an equity allocation moves the outcome less than how much of the total is in equities at all, which is risk management at the level of the whole.
And diversification only works if things differ. Holding many names from one sector is a single position written out many times. Independence of behaviour is the property that matters, not the number of lines on the statement.
Which they stop doing exactly when it matters. Correlation, the degree to which two holdings move together, tends to rise in a crisis, so the spread is measured in calm conditions and tested in bad ones.
Keeping the mix honest
Rebalancing sells what rose, which feels wrong. Restoring the intended proportions means trimming the winner and adding to the laggard, and that discomfort is the mechanism keeping the allocation the one you chose.
Most portfolios are far more concentrated than they look. Two funds can hold the same companies, whether bought as index funds or through ETF investing — an exchange-traded fund being a listed basket — and home-country bias stacks one economy underneath both.
The largest exposures are usually not in the account at all. A salary depends on one employer and one sector, and a home is a borrowed position in one street, so real estate belongs on the same page as the holdings.
And cash is a position, not an absence of one. Holding it is an allocation with its own expected real return, which inflation erodes, so calling it uninvested hides a decision that was made anyway.
In practice
The fee compounds against you every single year. Compounding an annual charge alone over thirty years removes 1.5% of the pot at five basis points, 5.8% at twenty, 20.2% at seventy-five and 36.5% at one hundred and fifty. A basis point is one hundredth of a percentage point.
Participation is not an input to any of this. Volume says how busy a market was on a given day, which matters for entering a trade and not for deciding what fraction of a total belongs in equities.
It is judged over decades, not quarters. On this history 95% of bars sit below a prior peak, the median drawdown is 1.36%, and the longest stretch below a peak runs 73 bars — an ordinary buy and hold picture.
And a crash moves everything the same direction. An opening gap does not respect an allocation; assets that behaved separately for years reprice together overnight.
There is no stop on an allocation. A stop loss exits one trade at a level, and a mix has no equivalent, so the protection has to sit in the proportions before anything falls.
Every rebalance costs 2% of a bar. A round trip on this history takes 2% of a median bar’s range and 45% of the smallest, and outside retirement accounts it realises a taxable gain as well.
Writing the allocation down
Write it down before anything falls, because that is the only time you will be honest. One page with three parts: the target proportions, the trigger that starts a rebalance, and the things you will not do.
The proportions come first, as fractions of the whole rather than amounts. Name each asset kind and say what it is expected to do in the years the others are doing badly.
The trigger is a rule, not a mood. A date or a drift threshold both work, and contributions arriving on a schedule — dollar cost averaging — quietly do part of the job by buying whatever sits below its target.
The refusal list is the part that saves money. Write down that you will not chase what rose and will not abandon the mix in a bad quarter, because an allocation you cannot hold through a decline is the wrong one.
What portfolio building is not
It is not stock picking. The mix decides more than the selections do.
It is not a number of holdings. Twelve names can be one position.
It is not fixed once chosen. Left alone it drifts into whatever rose.
And it is not only what sits in the account. The job and the house are positions too.
When it fails
In a flat decade the fee is the only thing you control. A market stuck in a trading range offers no return to choose from, but the charge is chosen: seventy-five basis points removes 20.2% of a thirty-year pot on its own.
The second failure is correlation arriving all at once. Holdings picked for independence in calm conditions move together in a crisis, and beta, how much a holding moves with the market, stops telling them apart when everything sells.
A third is hidden concentration. Two funds holding the same largest companies, an employer in the sector of the biggest allocation, and a house in the economy underneath both.
A fourth is rebalancing too often, or never. Too often spends the round-trip cost and realises gains; never leaves the whole mix pointed at whatever has already risen.
A fifth is judging comfort by depth alone. The deepest drawdown here is 3.76%, but the ulcer index, the root mean square of the drawdown series, is 1.67% — a ratio of 0.44 to the maximum. Discomfort comes from persistence, not depth.
And a sixth is an allocation chosen for a spreadsheet. Sharpe ratio and standard deviation describe a mix; neither says whether its owner will still hold it after a bad year.
The original data
Of the 24,971 videos scanned, 147 carry “portfolio” in the title at a median of 11,537 views across 117
channels, maximum 3,381,118. “Correlation” appears 9 times at a median of 1,162 across 5 channels, “index
fund” 30 times at a median of 74,230 across 27 channels, “compound interest” twice at a median of 881, “net
worth” twice at 1,057, and “risk of ruin” not once. The scan is research/broker-coverage.json, run over
research/search-study-corpus.jsonl.
The thing being built is discussed nineteen times more often than the property that decides whether the
building works. A portfolio without a view on correlation is a list of holdings, not a portfolio. The fee
ladder in research/series-measurements.json, produced by site/measure_series.py, is the other half: at
seventy-five basis points a year the charge alone removes a fifth of a thirty-year pot, and at one hundred
and fifty over a third. In a flat decade that is the only number the holder actually controls. List
everything you own including the house and the job, then look at what fraction of your total position depends
on one economy or one employer.
Related
Correlation is the property deciding whether a spread of holdings is actually a spread, and it is the page to read next. Index funds are how most people implement an allocation cheaply, which matters because the fee is the most reliable lever available. And risk management is the same discipline applied to one position rather than the whole.
For years I held a spread of names and called it diversification, until a bad week moved every one of them together. What I actually owned was a single bet written out several times, and the number of tickers had been hiding that from me. Since then I write the mix down first and check what my holdings have in common rather than what makes them different. It is a duller exercise than picking, and it has mattered a great deal more.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.