Margen
In 2001, in a run of laboratory negotiations, the side that named the first figure closed better than the side that waited. It happened whether the buyer opened or the seller did: whoever let their number out first came away ahead. And that contradicts what you have been told all your life, that whoever speaks first loses.
The set-up was Adam Galinsky and Thomas Mussweiler's, and what it measures is clean. They had students negotiate the sale of one and the same industrial plant in pairs. One played the buyer, the other the seller, both with the same file in front of them and the same time to study it, and for each pair they noted two things: who had said the first price and what price they had settled at. The final price did not land halfway between the two, as you might expect: it trailed after the first one. They published it in the Journal of Personality and Social Psychology under the idea that first offers work as anchors.
The mechanism, seen from the inside, is sillier than it sounds. You arrive with a band: you would pay up to X and you would settle for Y. If the other side lets their figure out before you do, your band disappears from the conversation and what gets argued over is theirs. You can fight it, you can pull it down a long way, you can walk out offended. What you are not going to do is close far away from it, because from that moment on both of you are hunting for arguments inside the ground they drew. Even the concessions you wring out of them, you wring out of a starting point they chose.
That said, the old saying is half right, and it is worth knowing which half. Opening only helps you if you know what the thing is worth. Picture yourself selling a car the market prices at 3,400 euros and asking 1,900 because it strikes you as reasonable: nobody is going to correct you upwards, and the person sunk by your anchor is you. Which is why the real rule is about information and not about turns: with the prices checked, open; without them, name no figure and ask questions.
And there is a second limit, one the experiment does not cover but any real table does: a wildly disproportionate number does not anchor, it offends. Ask for three times what something is worth and the other side does not adjust towards you, they get up or they stop taking you seriously. An anchor works when it is high and defensible at the same time, and what makes it defensible is being able to say where you got it.
With pay, all of this is starker than anywhere else, because the information gap is at its widest: the company knows what it pays twelve people in your job and you do not even know the range. Getting hold of it is not a matter of intuition, it is a matter of half an hour. Your sector's collective agreement is published and lists the scales by grade. Comparable job ads that are still open carry a range. And there are almost always two or three people you can ask without the question costing you anything. Once that half hour is done, your figure stops being a bet of yours and becomes a number with a source, which is precisely what turns it into an anchor rather than a whim.
Then there is the other half of the time, which will come round just as often: the other side says the first number. That is where the finding becomes genuinely useful, because Galinsky and Mussweiler also tested how an anchor is switched off. The answer is not heroic in the least. Resisting does not work: the effect turns up in people who know perfectly well that it exists, and being warned does not inoculate you. What works is shifting your focus before you answer, and the focus that works is not the one you would pick on your own. Think about your target, the ambitious figure you had in mind before sitting down; or think about the other side's limits, the most they can pay and what alternatives they are left with if you walk today. With any of those three in front of you, the opener's advantage disappeared. And there is an uncomfortable detail the same authors measured a year later: whoever clings to their walk-away price, the number below which they leave, closes worse than whoever clings to their target. The walk-away is there so you do not cross it. What you take to the table is the other number.
It is worth saying, since these were students in a simulation, that nobody has measured this on the purchase of your house. But the mechanism does not depend on the laboratory: it depends on the first figure taking the middle of the table, with everything else arranged around it.
So the next time you have to talk about money with somebody, settle one thing before you walk through the door. If you have checked what it is worth, say your number in the first two minutes and say where it comes from. And if you have not checked, name no number at all and spend the whole meeting asking.