3. Rebalancing: what it gives, what it costs

Risk 5: Building the portfolio

A portfolio that is never touched stops being the portfolio you designed. No mistake is required: it is enough for the assets to move at different rates, which is the only thing they ever do. Rebalancing returns it to its original shape, and it is one of the few management decisions you can discuss with the arithmetic in hand.

The first thing to understand is drift

Take a portfolio of 60% shares and 40% bonds. Suppose ten years in which equities return 10% a year and bonds 2%. Touching nothing, the 60 turns into 155.6 and the 40 into 48.8. The portfolio is worth 204.4 and the weight of equities has gone from 60% to 76%.

Nobody has decided to raise the risk: it has gone up on its own, and right after an excellent stretch for equities, when valuations are at their highest. Drift always increases your exposure to whichever asset has risen most, which is rarely the one it makes sense to increase.

That is rebalancing's main job: keeping the level of risk you chose. It is not a technique for making more. It is a technique for not ending up, without noticing, in a portfolio you would never have signed off.

The side effect: buying cheap and selling dear without having an opinion

Rebalancing forces you to sell what has risen and buy what has fallen. It is a discipline that runs against the impulse: you cut back what is doing well, which is the last thing you feel like doing, and you add to what is doing badly, which is the second to last.

There is a part of this that can be quantified. When two assets have similar expected returns, plenty of volatility and little correlation, the rebalanced portfolio compounds above the average of its components. That difference is called the rebalancing premium, and it comes to roughly half the variance the mixture eliminates:

• Two assets at 50%, volatility of 20% each and correlation of zero: the effect comes to around 1% a year.
• The same two assets with a correlation of 0.8: the effect falls to around 0.2% a year.

It is worth being honest here. It is not free money or an independent source of return: it is an arithmetic consequence of how returns compound, and it depends entirely on the assets behaving in a particular way. If one of them enters a long, sustained trend, rebalancing will have had you selling the good one to buy the bad one, year after year. In the portfolio in the example, rebalancing through that decade of strong equities would have subtracted return, not added it. And it would still have been the right decision in risk terms.

Calendar or bands

The two families of rules are straightforward and they behave differently:

• By calendar. You rebalance on fixed dates: monthly, quarterly, annually. Advantage: it is impossible to argue with and it requires no watching. Disadvantage: it trades when there is no need and does not trade when there is. A monthly portfolio can generate something like three hundred and sixty trades over thirty years, many of them over deviations of a tenth of a point.
• By deviation bands. You rebalance when a weight moves away from its target by more than a threshold: five points in absolute terms, or 25% in terms relative to the target weight. Advantage: you only act when the risk has genuinely shifted. Disadvantage: somebody has to look.

A 60/40 with a five-point band triggers when equities reach 65% or drop to 55%. Over three decades that usually sets off something like ten or fifteen rebalances instead of three hundred and sixty. The resulting risk control is practically the same, with a fraction of the trades. The conclusion that comes up again and again when rules are compared is that there is no universally optimal frequency, and that the differences in return between reasonable rules are small next to the differences in cost.

The cost, which is where everything is decided

Three items, in increasing order of importance:

• Commissions. Small today in liquid assets, but not zero.
• Spread and impact. What you lose between what you buy and what you sell. Irrelevant in a large index, very relevant in a thin asset.
• Tax. The item that settles the argument. In Spain, capital gains are taxed within the savings base on a scale that runs from 19% to 30%. Move eight points of a 100,000-euro portfolio: you sell 8,000, of which let us say 3,000 is gain. At 21%, the bill is 630 euros, 0.63% of the portfolio. Compare that with the rebalancing premium of between 0.2% and 1% a year and you will see that the order of magnitude is the same.

Out of that come two practical consequences. The first: rebalancing with new money, with the dividends or with regular contributions avoids almost all the cost, because it buys without selling. The second: in an account with deferred taxation the calculation changes completely, and the bands can be much narrower.

2008 and 2009: the discipline put to the test

An investor with a 60/40 who started 2008 finished it with shares falling around 37% and aggregate bonds rising around 5%. The portfolio was worth 79.8 and the weight of equities had fallen to 47%: thirteen points below target, well outside any reasonable band.

Rebalancing meant buying some twelve points of portfolio in shares in the worst atmosphere in eighty years, and the market went on falling until March 2009, by which point it was down close to 57% from its peak. Afterwards, 2009 closed with a total index return of close to 26%.

The lesson is not that rebalancing picks the bottom, because nobody knew where the bottom was. It is that the rule takes the decision almost no investor takes on his own, and it does so without needing to be right about the future.