Anchoring: The Price That Haunts You

Clean anchors and dirty ones

An anchor is a number that pulls an estimate toward itself, and the useful distinction is between anchors that describe the market and anchors that describe the investor. A price where the stock has repeatedly reversed, a level where a large volume of options is written, a round number where orders cluster, a published estimate of fair value — these are clean anchors, because they are facts about other participants that a stranger with no position would also cite. What the position cost, what it was worth in March, and what the investor was once told about it are dirty anchors: they are true statements about a history that the market has no knowledge of, and using them as reference points substitutes that history for current analysis. The mechanism is not mysterious. Anchoring works by making a number available, and adjustment away from an available number is famously insufficient — estimates land closer to the anchor than the evidence warrants. Applied to a portfolio, that produces a set of decisions that look like analysis and are actually comparisons: is the position up or down against what was paid, is the house up or down against what it was worth last spring. Each of those questions has an answer that feels like information and contains none, and each of them crowds out the question that does. Anchoring is also expensive in a measurable way, which is worth working through because it links directly to the previous lesson. A 20% loss needs a 25% gain to undo, so the recovery target is itself a more demanding number than the loss suggests. Layer a high-water mark on top and the demand grows: a position at $96 asking for $132 is asking for a 37.5% move, and the high has become load-bearing in the investor’s reasoning — “it was worth that in March” is a fact about March. Every one of those numbers can be computed, and none of them is a valuation. The position and its references — Purchase price $120, price now $96: Unrealised loss 20%; recovery needed 25% ← · Published fair value $104: The gap between the anchor and the value estimate is $16 ← · High-water mark $132: A further 37.5% from here — a fact about March · Which numbers the market knows: Only the ones not derived from your own history A useful self-check when a number appears in your reasoning: would a stranger with no position in this name also cite it? If not, the number is about you rather than about the asset.

Resetting the frame

Because anchors are set by availability rather than by belief, the defence is to change what is available. The cheapest version is a fresh-eyes ritual: for each holding, write the thesis as though nothing were owned, decide buy, hold or sell from scratch, and then compare that answer with the actual stance. Divergence between the two is the anchor talking, because the only thing that changed between the two exercises is the presence of a position and a remembered price. The second defence is to give the anchor somewhere legitimate to live. Cost basis matters for tax, so it belongs in a record alongside the acquisition date rather than at the centre of the decision. A high-water mark matters for nothing except the historical fact that the market once paid more, so it can be dropped from the argument entirely. Once those two are filed, the remaining references are the ones a stranger would recognise: the price, the estimate, the level, the yield — the inputs of a valuation rather than the residues of a trade. It is worth being fair to anchoring at the end, because it is not a defect of intelligence and it is not confined to beginners. Professionals mark their estimates to the last thing they published and adjust insufficiently in the same way, which is why the ritual is procedural rather than motivational: writing a thesis from scratch, with the position and the entry price out of view, is a way of producing the estimate you would have produced without the anchor. The point is not to feel unbiased. It is to notice, on a schedule, which question you are actually answering. • Write the thesis as if nothing were owned, then compare with the actual stance. • File the cost basis where tax treatment lives; drop the high-water mark from the argument. • Keep only the references a stranger would cite. • Run the ritual on a calendar, because availability is what sets the anchor. The most expensive version of anchoring is not a held loser; it is a refusal to buy a good business that has doubled since it was first noticed. “Missing it” is a statement about the date the decision was first considered, and the position does not know that date.

The anchor you do not notice: your own estimate

The visible anchors in this lesson are prices — what you paid, what it was worth last year, what a friend said. The more dangerous one is quieter, and it operates on your estimate rather than your decision. If you look at the quote before you value the business, the quote becomes the starting point for the valuation, and every subsequent assumption tends to adjust toward it. The result looks like independent analysis: you build a model, you pick growth and margins, and your answer lands within a few percent of the price you saw. That is not a coincidence and it is not evidence that the price is right. The same mechanism explains why so many published price targets cluster near the current price. A target is a forecast with a strong incentive to be defensible today, and a number far from the market invites argument, so the estimate gets dragged toward the anchor it was supposed to be independent of. This is a social version of anchoring as much as a cognitive one, and the practical tell is the size of the gap: targets that sit within a few percent of the price are describing the price, not the business. The fix is procedural and costs nothing. Do the valuation **before** you look at the quote, write the range down with its assumptions, and only then compare it with the price. If the answer changes after you see the market’s number, the change is the interesting event and you should be able to say exactly which assumption moved and why. An estimate produced after the price is a rationalisation with a spreadsheet attached, and it will feel just as rigorous from the inside as one produced before it. The same company, two orders of operations — Value first, then look at the price: A range you can disagree with — and a real question if the price is outside it · Look at the price, then value: An answer that lands near the price, with assumptions quietly reverse-engineered · Read the targets first: A consensus range you have adopted without noticing you adopted it ← If you have already seen the price — and you almost always have — write down what the price implies about growth and margins before you model anything. That converts the anchor from a contaminant into an explicit assumption you can attack.

The anchors the market supplies for you

Anchoring experiments usually plant a number in a lab. The market plants them for you, dozens at a time, and the ones embedded in the data are the hardest to notice because they look like information rather than like a starting point. The most persistent is the **52-week high**. It is displayed on every quote page, it is what a stock’s fall is measured against, and it converts a price into a story: a stock at forty that was at ninety “has room”, while a stock at ninety that was at forty “is extended”. Neither statement contains a fact about value. The all-time high does the same work over a longer memory, and it is why the phrase “still below its high” appears in arguments that are otherwise about cash flow. The second family is the round number, and its power is mechanical as well as psychological. Orders cluster at round strikes and round prices, so the level is real in the order book and the psychological pull on the person reading it is reinforced by the liquidity that sits there. That is why “it broke the hundred” is treated as an event when the price moved from 99.80 to 100.10, a difference of thirty cents and no change in the business. The third is the price you first saw the stock at — an IPO price, the print on the day you put it on a watchlist, the level at which you first considered buying. That number is pure noise relative to the business, and it has more influence on when people buy than most fundamentals do, because a position that was “too expensive at 140” feels like an opportunity at 120 without any earnings having changed. The defence is procedural rather than psychological, which is the useful part. Estimate the value first, from the numbers, and write the estimate down before opening a quote page — the order of operations is what removes the anchor, because a number formed before the anchor appeared cannot have been pulled by it. Then, when reading a price, notice which comparisons you are making automatically. If you are measuring the price against a past price rather than against your own estimate, the anchor is doing the analysis. The test is simple and uncomfortable: if your view changed when the price changed, and the business did not, something other than the business is speaking. • The 52-week high turns a price into a story, and the story is not about value. • Round numbers are anchors *and* real order-book levels, so the pull has two sources. • The price you first saw is the least defensible reference you carry. • Estimate before you look, and write it down; the order of operations defeats the anchor. One habit catches most of it: when the price moves, ask what changed in the business. If the honest answer is nothing, the new number is an anchor, and the old one was too.

Sources

• Anchoring and insufficient adjustmentKahneman, “Thinking, Fast and Slow”
• Anchoring in financial estimates and pricesThaler, “Misbehaving”
• Drawdown convexity and the recovery arithmeticStandard return mathematics; drawdown research