2. What to look at every month, and what never to look at

Risk 9: A risk system of your own

A risk system has two parts: the limits, written once, and the review, which repeats. The second is where almost everybody goes wrong, and not for looking too little. Intuition says that reviewing more often is being more responsible, and that the more information comes in, the better the decisions will be. The arithmetic says exactly the opposite, and it is worth seeing before arguing about it.

The asymmetry between seeing noise and seeing signal

Take an equity portfolio with an expected 8% a year and 20% volatility. The signal, that is the expected return, grows with time. The noise, that is the dispersion, grows with the square root of time. That difference explains everything.

• Over one day, the expected signal is around three hundredths of a point and the daily noise is around 1.26 points. The ratio is roughly one to forty.
• Over one year, the signal is 8% and the noise 20%: roughly one to two and a half.
• Over ten years, the accumulated signal is around 80% and the noise some 63 points: for the first time the signal is bigger than the noise.

From which comes the uncomfortable rule: looking more often brings no additional information, it brings more noise per unit of signal. Today's price contains no data about whether your portfolio is well built; it contains the mood of a market over six and a half hours.

The practical consequence is the one that hurts. Each observation is not neutral: it can turn into a decision. And since the probability of seeing a loss is close to a half over a day and considerably lower over a year, whoever looks daily submits to some two hundred and fifty occasions a year to feel pain, against one. Kahneman and Amos Tversky showed that the asymmetry between the pain of losing and the pleasure of gaining does not net out: it accumulates.

What to review every month

Everything that follows changes slowly, is measured without ambiguity and does not depend on this week's price:

• Real exposure by source of risk. Not the list of names, but the factors: how much sits behind a single sector, a single currency, a single interest rate, a single country. It is the review that discovers that twelve positions were three bets.
• Concentration. The weight of the largest position, the weight of the largest five and the weight of the largest factor. Three numbers that fit on one line and that answer level 7's question: where is the bet I did not know I had?
• Liquidity. Two figures: the months of spending covered outside the portfolio, and what proportion of the portfolio you could sell on a bad day without moving the price against you.
• Drift from the plan. Weights move on their own: whatever rises takes up more room. If your target was 8% per position and the market has carried it to 14%, the market has taken a position sizing decision on your behalf that you never approved.

What to review once a year

• The thesis of every significant position, and whether it still says the same thing as the day you wrote it.
• The real costs paid in euros: commissions, the bid-offer spread, currency conversion and taxes.
• The withdrawal rate and the contribution plan from level 8, against the spending you have actually had.
• The limits themselves, with the rule that any change takes effect only after a delay.

What never to look at

• The price every day. It produces no knowledge, only decisions, and it produces them at the worst possible moment.
• Cumulative return against an index over short periods. Over a few months that difference is almost entirely chance, and its only practical effect is to push you into changing the portfolio right after a bad stretch.
• The value of the portfolio in euros on the day of a big move, which is exactly when the figure has the greatest power to dictate an order.
• News with the names of your positions on it, which replaces a review of factors with a collection of anecdotes.

Philip Tetlock found that the best forecasters update often but in small increments and on real evidence, and that whoever reacts to every headline ends up with an opinion that is the average of their last few frights.

A real case: the information frequency experiment

In the late nineties, a group of researchers including Richard Thaler, Amos Tversky and Daniel Kahneman designed a simple experiment. Some participants were shown the results of their investment very frequently; others saw them aggregated into blocks of several years. All of them could split their money between a safe asset and a risky one, with the same underlying returns in both groups.

Those who saw frequent results put around 40% into the riskier asset. Those who only saw long aggregates put close to 70%. The same investment, the same distribution of outcomes and the same person: the only thing that changed was how often they were shown the scoreboard. They called that effect myopic loss aversion, and Thaler and Shlomo Benartzi later proposed it as the explanation for why equities have historically had to pay such a high premium over debt.

The empirical accompaniment comes from Brad Barber and Terrance Odean, who studied tens of thousands of real accounts at a US broker in the nineties. The average household obtained around 16.4% a year against 17.9% for the market; the most active quintile came in at around 11.4% net. The difference did not come from picking worse shares, but from trading, and trading comes from looking.

The conclusion is straightforward: reduce how often you see the scoreboard and you will have improved your decisions without changing a single position.