Portfolio construction
Positions are opinions, a portfolio is a machine, and construction starts one step earlier than most investors begin: with what the machine is for, before the core-view-hedge layers, the hedging ladder, or any asset gets chosen.
Positions are opinions; a portfolio is a machine. Construction is the discipline of assembling opinions across asset classes so the machine survives every plausible weather, and it begins one step earlier than most investors start: with what the machine is FOR.
Objectives first, because they change the answer
"Optimized for what?" has different answers with different portfolios behind them: maximum long-run compounding (tolerates deep drawdowns), a target with a deadline (a house in five years cannot ride a 50% drawdown), income now, or capital preservation. Time horizon, liquidity needs and the true pain threshold (the one revealed in 2008 and March 2020, not the one claimed in questionnaires) are constraints that belong on paper before any asset is chosen, because every later decision inherits them.
The layer model: core, views, hedges
CORE 60-90% the strategic mix that earns market returns across
regimes: diversified equities, duration, real assets.
Built with frontier logic; touched rarely.
VIEWS 10-30% the theses: single names, sector tilts, the trades
from the thesis docs. Sized by the risk rules; each
with its written plan.
HEDGES 0-10% positions whose JOB is to pay when the rest hurts:
index puts, duration in deflation risk, gold against
debasement. Judged by protection delivered, never by
standalone return.The layers exist to keep honest accounting: the core is judged against a benchmark, the views against their theses, the hedges against the disasters they insure. Mixing the books (a "view" that quietly becomes core because selling feels like defeat) is how portfolios drift into shapes nobody chose.
What each class actually contributes
| Class | The job it can do | The regime that hurts it |
|---|---|---|
| Equities | The growth engine; the compounding core | Recessions; rate shocks to long duration |
| Government bonds | Deflation insurance; ballast when growth breaks | Inflation, which flips the correlation |
| Credit | Yield between the two | Late cycle; it is equity risk in a bond costume |
| Commodities | Inflation participation; supply-shock hedge | Disinflationary calm; roll bleed |
| Gold | Debasement and crisis hedge; real-rate asset | Rising real rates |
| Crypto | Asymmetric upside; debasement thesis, levered | Liquidity drains; it is high-beta in stress |
| Cash / T-bills | Optionality with a yield; the dry powder | Inflation quietly; melt-ups loudly |
The classic failure is owning seven things that all need the same weather: falling rates and calm inflation. 2022 taught the 60/40 that lesson. Construct by asking which REGIMES each holding pays in (growth up or down, inflation up or down, liquidity in or out), and check every cell has something. The correlation matrix of the last five years is a description of the last regime, not a promise about the next.
Hedging: the ladder from free to expensive
Hedging has a cost ladder, and professionals exhaust the cheap rungs before paying for the expensive ones:
- Rung 1: don't hold the risk. Trimming an oversized winner is the only free hedge. Most hedging questions are sizing questions wearing a disguise.
- Rung 2: offsetting assets. Duration against equities (in low-inflation regimes), gold against debasement, cash against everything. Costs only expected return.
- Rung 3: reduce the specific exposure. Short the sector against the single name, hedge the currency on the foreign position: keeps the thesis, sheds the passenger risks.
- Rung 4: explicit insurance. Index puts, collars on concentrated stock, tail funds. A known premium for convex protection; a permanent standing budget (say 0.5-1% a year), or targeted around identified events, never improvised after the fall began.
Two rules keep hedges honest: a hedge must be sized to matter (a 1% put sleeve does not save a 100% equity book), and a hedge that paid must be MONETIZED (rolled, rebalanced into the fallen assets), or the protection evaporates with the rebound it just paid for.
Rebalancing: the machine's engine
Weights drift with performance, so an untended portfolio becomes concentrated in whatever rose: yesterday's winners, at yesterday's prices. Rebalancing (on a calendar, or at threshold bands like "when a sleeve drifts 20% from target") mechanically sells high and buys low, converts volatility into a small extra return, and, more importantly, keeps the portfolio being the one you chose. It is also where taxes and costs live: bands beat calendars for tax efficiency, and new contributions are the cheapest rebalancing instrument of all.
The construction review, quarterly
- Sum the book by DRIVER: what is the true equity beta, duration, dollar, inflation and liquidity exposure, across all wrappers?
- Which regime is this book implicitly betting on, and is that a choice or an accident?
- What single event costs the most, and is its hedge in place and sized?
- Have any views quietly become core? Any hedges become hopes?
- Strategic asset allocation
- The long-term policy mix across asset classes, set from objectives and regime logic, touched rarely. The single decision that explains most of a diversified portfolio's outcome.
- Core-satellite
- A cheap diversified core doing the compounding, with a bounded sleeve of active views around it. The structure that keeps opinions from silently becoming the whole portfolio.
- Regime
- A market era with its own rules: the inflation seventies, the QE 2010s. Backtests assume tomorrow is drawn from the sample's regime; report performance per named regime instead of one blended number.
- Hedging ladder
- The cost-ordered menu: hold less of the risk, own offsetting assets, hedge the specific exposure, buy explicit insurance. Professionals exhaust the free rungs before paying for the expensive ones.
- Tail hedge
- A standing position whose job is to pay in crashes: far out-of-the-money puts, long volatility. A drag most years by design; the premium buys the year that matters.
- Monetizing a hedge
- Harvesting a hedge that paid: rolling it, rebalancing the gains into fallen assets. Skipped, the protection evaporates with the rebound it just paid for.
- Rebalancing
- Trading back to target weights on a schedule or at thresholds: mechanically selling what rose and buying what fell. The discipline that keeps an allocation being the allocation you chose.
- Threshold bands
- Rebalancing when a sleeve drifts a set distance from target rather than on a calendar. Fewer trades, better tax outcomes, same discipline.
- Drift
- Weights wandering from target as performance compounds, concentrating the book in whatever already rose. The silent force rebalancing exists to answer.
- Duration (asset sense)
- A bond's price sensitivity to interest rates: duration 7 means roughly a 7% price fall per one-point yield rise. The dial bond managers use to set how much rate risk a portfolio carries.
- Collar
- Buy a protective put, fund it by selling a call above: downside floor purchased with surrendered upside. The standard dressing for a concentrated winner.
- Dry powder
- Unspent capacity, held on purpose: cash or unused risk budget available when prices are best. Not a failure to be invested; an option on everyone else's forced selling.