Portfolio construction

Diversification, honestly

Diversification is the only free lunch the market serves, and it is routinely oversold by its own menu. The true claim is narrow and powerful: combining assets that do not move together lowers the portfolio’s volatility (Lesson T2) without lowering its expected return by the same amount — the zig of one holding fills the zag of another, and the compounding of Lesson 21 prefers the smoother path. The false claim is that a long list of tickers is therefore diversified. Thirty technology stocks are one bet wearing thirty costumes: what they share (a sector, a factor, a customer) is what will move them together, and they will move together precisely on the day diversification was for.

The honest measure is not the count of names but the count of independent ways to be wrong. Different sectors, different drivers, different asset classes (Lesson I3’s ballast exists for this) — each genuinely new exposure retires a scenario in which everything falls at once. Folder 2’s sectors and breadth were the vocabulary for this; a diversified portfolio is breadth, owned deliberately.

One more honesty: the togetherness is not a constant. Correlations measured across calm years describe calm years. In a market-wide panic, holdings that zigged and zagged politely for a decade can fall in the same week — frightened sellers sell what they own, not what they meant to, and the shared fact of being owned becomes the correlation. That does not retire diversification; it narrows its honest promise. Genuinely different exposures still fail in genuinely different scenarios — a rate shock, one sector’s bust, one company’s fraud — even if a general panic can take most prices down together, by different amounts and with different recoveries.

Allocation by horizon

Before any name is chosen, a portfolio is an answer to one question: when is this money needed? Money with a date inside a few years belongs where its value cannot be halved by a bad season — I3’s shorter, higher-grade loans — because a drawdown (Lesson 21) does not care about your deadline, and selling into one converts a fluctuation into a fact. Money with a decade or more can ride equities’ weather, because time is the specific asset that converts volatility from a danger into a toll.

The practical shape this produces, for a working reader, is buckets before tickers: dated money in ballast, long money in the equity allocation, and the equity allocation itself divided between a broad core (the index exposure of Folder 2, owned plainly) and the chosen names the journal’s work informs. The proportions are personal — they belong on the investor’s plan page (Lesson I6) — but the order of decisions is not: horizon first, allocation second, names third. Most portfolios are built in exactly the reverse order, and their owners spend years discovering why the order mattered.

Rebalancing: the mechanical correction

Allocations drift: the winning bucket grows, and a 70/30 portfolio after a strong equity year can be silently a 75/25 one — riskier than its owner chose, at the top of the move that made it so. Rebalancing is the rule that reverses the drift: on a schedule, or when a bucket crosses a band (five percentage points is a common teaching shape), sell the overweight back to target and add to the underweight. Read what the rule actually does: it sells what has risen and buys what has lagged, in fixed proportions, at whatever prices the day offers — no forecast, and no knowledge of whether either price is a high or a low. A calendar doing the work a crystal ball cannot.

Its second job is Folder 5’s: rebalancing is a drawdown instrument. The portfolio that rebalances into a falling equity bucket is buying the drawdown in pre-committed slices — the investor’s version of the plan made when calm. The rule must be written (I6) because in the moment, every instinct will argue against both directions of it: trimming winners feels like betrayal, adding to losers feels like madness. The calendar does not feel either. That is the point of the calendar.

The portfolio is a Folder 5 citizen

Everything Folder 5 built for single trades scales to the whole. The portfolio has a drawdown line (at what depth does the plan change — and the honest answer, for long money, is usually it doesn’t, which is itself a plan worth writing before the drawdown). It has position limits (the largest single name’s share, capped on the plan page, so no company’s bad year becomes the portfolio’s). And it has a review rhythm — the Sunday fifteen minutes (Lesson 24), where drift is measured, bands are checked, and the one-decision rule keeps the portfolio from becoming a trading account with a longer memory.

Construction, then, in one sentence for the plan page: diversify across ways of being wrong, allocate by when the money is needed, and rebalance on a rule. The names are the visible part of a portfolio. These three decisions are the load-bearing part.

Where this breaks

Diversification’s measurements are mostly taken in calm weather. Correlations estimated across quiet years describe quiet years; in a general fall, shared ownership itself becomes the correlation, and holdings chosen to move differently can drop together. What survives that failure is narrower and still real: different holdings fail in different scenarios — a rate shock, a sector bust, a single company’s fraud. The rebalance, too, promises less than its nickname: it sells what rose and buys what lagged at that day’s prices, with no claim either price was a high or a low.

Practice

Run the rule on a fictional portfolio. $50,000 targets 70/30. A year later the equities stand at $44,100 and the ballast at $14,700. (1) The total, and the current weights. (2) The plan’s band is five percentage points: does a rebalance trigger, and if so, how many dollars move, and in which direction? (3) In one sentence: why does the rebalance, correctly described, make no claim about highs or lows?

Answers: (1) $58,800, split 75/25. (2) Yes — the drift has reached the band’s edge: sell $2,940 of equities and add it to ballast, restoring $41,160 / $17,640, exactly 70/30 of $58,800. (3) The rule reads proportions, not prices: it sells what rose and buys what lagged at that day’s prices, and nobody can know at the time whether those prices were tops or bottoms.