How Anchoring Affects Our Decisions
Plus, how language can conceal with precision, and how slope graphs can help you see shifts over time. Bonus: a musical short on slippery slopes.
Issue No. 8
A warm welcome to the 100 or so subscribers who have joined us since last time. It’s wonderful having you here. I’ll briefly reintroduce myself—one summer ten years ago, I put together an illustrated book on bad arguments, and I’ve been writing on critical thinking, off and on, ever since. My approach is to take common knowledge that’s been around for a while, and reframe it in a way that makes it memorable.
Two bits of bonus content this week. For all subscribers, there’s a new musical short. This one’s on slippery slopes. In case you missed it, the first one was on false dilemmas.
And for paid subscribers, you can now listen to issue no. 7, on slippery slopes. Did I have the audacity to sing that excerpt from Hamilton? You’ll have to listen to find out.
Earlier in the week, my publisher informed me that for Amazon Prime members in the US, you can now get the print edition of An Illustrated Book of Loaded Language at a discount. The offer runs through next Wednesday, October 11. There are a set number of units available, so the deal only lasts until those sell through.
1 Reasoning
Anchoring
Imagine these two scenarios. You want to buy a TV.
First scenario, you look to one friend, and you see that they’ve bought a $3,000 model. You look to another friend, and see that they’ve bought a $2,500 model. You just want a regular TV, so you decide to buy a less expensive one, at $2,000.
Second scenario, you look to one friend, and see that they’ve bought a $500 model. You to look to another friend, and see that they’ve bought a $250 model. You just want a regular TV, so you decide to buy a less expensive one, at $100.
This tendency to get tied to some initial reference point, and then adjust from there, is known as anchoring. And it’s a form of bias that can affect our decisions.
In an experiment, two groups were asked what they thought Mahatma Gandhi’s age was at the time of his death. But before asking them that question, the first group was asked whether Gandhi was older than or younger than nine at the time of his death. And the second group was asked whether Gandhi was older than or younger than 140 at the time of his death. After answering the original question (what Gandhi’s age was), the first group ended up giving a lower age (50), on average, than the second group (67).1
The takeaway here is that some initial information you show someone can affect their subsequent thinking. It anchors them to a reference point. What’s truly interesting is that the bias can affect our judgment, even when we’re explicitly told that a piece of information we’re shown is irrelevant,2 which means we’re more susceptible to that initial stimulus than we think.
Another way to realize the effect of anchoring is by trying to guess—in five seconds, say—what the answer to the following expression is: 8 × 7 × 6 × 5 × 4 × 3 × 2 × 1. Now do the same thing, but with this expression: 1 × 2 × 3 × 4 × 5 × 6 × 7 × 8. On average, people give a larger number for the first expression, the one that starts with an 8, than for the second expression, which starts with a 1. In an experiment done on two groups of high-schoolers, one group’s average answer for the first expression was 2,250 and the other group’s average answer for the second expression was 512, the correct answer being 40,320.3
We find anchoring in negotiations, where two sides may be trying to agree on the value of a company that’s for sale. The first person to share a number determines the anchor. If you’re looking to buy, too high a number and you might overpay. And if you’re looking to sell, too low a number and you might leave money on the table.
Similarly, with salary negotiations, when you go first and say how much you’d like to get paid, you determine the anchor. Say you’re at a job that pays you $100,000, and you’re interviewing for a new job, where you ask for $110,000. Not an unreasonable figure. It’s a 10% increase for you. The company might prevaricate for a bit to show you’re really twisting their arm and then eventually agree. But what you won’t ever know is how much they would have been willing to pay you had you not set $100,000 as your anchor, and on the basis of that determined that $110,000 was a good salary to ask for.
I’m reminded of this passage from a book on negotiations I read many years ago:
When the famous film director Billy Wilder went to hire the famous detective novelist Raymond Chandler to write the 1944 classic Double Indemnity, Chandler was new to Hollywood. But he came ready to negotiate, and in his meeting with Wilder and the movie’s producer, Chandler made the first salary offer: he bluffly demanded $150 per week and warned Wilder that it might take him three weeks to finish the project.
Wilder and the producer could barely stop from laughing, because they had been planning to pay Chandler $750 per week and they knew that movie scripts took months to write. Lucky for Chandler, Wilder and the producer valued their relationship with Chandler more than a few hundred dollars, so they took pity on him and called an agent to represent Chandler in the negotiations.
—Chris Voss (Never Split the Difference)
In negotiations, therefore, it can be useful to let the other person go first, as long as you’re aware that the other person, if they’re experienced, might well set an anchor that’s egregiously low to throw you off. To break out of an anchor, generally, and in this case as well, it helps to reference something objective, by saying, for instance, I’ve looked at industry reports, and jobs at this level for someone with my experience tend to pay within such and such a range. In some US states, it’s now mandatory to publish pay scales in job postings.
We see anchoring in marketing-speak as well. A company wanting to persuade you to buy their product will establish a reference point that allows them to show that the new product is better than the reference point, and therefore worth buying. This isn’t necessarily a bad thing, but it’s helpful to be aware of the approach so you can guard yourself against it.
For instance, on Dyson’s website, if you look up the newest cordless vacuum cleaner, you’ll find it says that this new model has
100% more suction power.
A reader might assume it’s 100% more suction power than the previous model, but a footnote mentions that the newer model has 100% more suction power compared to the V8, a model that was released six years ago. (At the time of writing, that footnote isn’t there on the website anymore.)
Similarly, at corporate events announcing a new model of a device, you might hear about how the newest model is 2.5X faster than the previous model. That relativeness aims to persuade you that this new thing is better than the old thing, irrespective of whether or not you actually need it to be 2.5X faster. But since it’s better, you feel a pull to buy it.
At the grocery store too, you’ll find examples of anchoring. You pass by a drink, like the one pictured below, and the label will say that it’s on sale. Rather than paying $2.29, you now pay $1.67. Relative to $2.29, $1.67 is a good deal. But relative to zero dollars—what you would have spent had you merely walked past this item—$1.67 is a lot more. Items on sale at a store encourage us to buy them because they make us feel like we’re getting a good deal relative to the struck-through original price.
With anchoring, it can help to consciously check for it any time you find yourself on the verge of buying something, believing something, accepting something, or making a judgment based on incomplete information.
(A question to ponder: Any mystery novels come to mind where the author uses anchoring to mislead their readers for dramatic effect?)
2 Rethinking Language
Concealing with precision
A statistic can sound authoritative. We might be more inclined to believe a speaker who drops statistics that one who doesn’t, because we might assume they’ve done their homework and know what they’re talking about. Sometimes that’s true. Other times, a statistic can be a distraction and an attempt at misdirection.
I was in a room one time watching a presentation about some service I knew a bit about. Midway, the presenter shared the line,
We’ve seen a 50% increase in users over the past month.
I remember thinking to myself: hardly anyone uses that service. How’s it being framed as a breakout win here. But when I looked at the data, sure enough, the service used to have four users, and now it had six. So technically it had seen a 50% increase in users, but it wasn’t a material increase by any stretch of the English language. I did learn an important lesson that day. That precision, ironically, can be used to create ambiguity.
On a well-known life coach’s website, one finds the line,
One year of joining this program increases life happiness by 60%.
And the line,
One year of joining this program increases daily motivation by 100%.
In both cases, there isn’t a footnote to spell out how those numbers are computed. A reader who doesn’t know any better might assume that numbers imply rigor and therefore a program that’s battle-tested and worth paying for. It’s only when we stop and ask where those numbers come from, do we find out if the numbers have a basis in reality.
In an article published during the summer of 2021 about car break-ins, we find the following headline:
Car break-ins are up 753% in S.F. tourist hub. The aftermath happens elsewhere
One’s immediate reaction upon reading that headline is utter shock at the extraordinary increase in crime, especially considering the awfully precise jump of 753% it mentions. Upon reading the article, however, one finds that the increase is as big as it is because 2020 was when the pandemic hit, which led to a significant drop in car-break-ins. And that break-ins in 2021 are on-par with break-ins in 2019. The original article’s title4 has since been changed to make it less sensational.
Precision in this headline, therefore, actually concealed the fact that a rare event (a pandemic) caused a drop so steep that any subsequent year would have inevitably seen a bigger than usual rise.
It’s always a good habit to read with skepticism—not necessarily suspicion, but at least skepticism—any line that drops a statistic. As we saw, a statistic might sound authoritative, but it can also be a mere red herring that helps an opinion pass as fact.
3 Rethinking Images
How slope graphs can help you see shifts over time
In the UK, viewers pay an annual license fee in exchange for watching live television, irrespective of how it’s transmitted—terrestrial, satellite, or cable. I recently came across this graphic online, which shows the increase in the TV license fee over the past twelve years.
What the graphic mainly shows is the year-over-year change in the TV license fee, and how an expected upcoming one in 2024 might be significantly higher. Using bars to show relative magnitudes is a great choice—a reader can do a side-by-side comparison of fees and get a sense for whether a change was small or big.
The one thing I wish the graphic did is tell me, at a glance, whether most years saw increases, decreases, or no change. To do that, with bars, I have to compare the filled-in part of one red bar with the non-filled-in part of another bar (negative space). And I have to do that for every pair of years. Not a bad approach, but it got me thinking of another approach that might work well.
One idea to make the graphic clearer is to switch to a type of visualization called a slope graph, which is excellent at showing precisely that sort of shift—did most years see increases, decreases, or no change? Here’s a rough idea for what a reworked graphic might look like, with red lines showing the five years in which the fee increased, and a steep increase in the sixth such year. That steep increase is now much more imposing.
Today, the word essay falls with a dull thud. It reminds many people of the exercises imposed at school or college to test knowledge of the reading list … Essayer, in French, means simply to try. To essay something is to test or taste it, or give it a whirl. One seventeenth-century Montaignist defined it as firing a pistol to see if it shoots straight, or trying out a horse to see if it handles well. On the whole, Montaigne discovered that the pistol shot all over the place and the horse galloped out of control, but this did not bother him. He was delighted to see his work come out so unpredictably.
—Sarah Bakewell (How to Live: Or A Life of Montaigne)
Next time, we’ll cover hasty generalizations.
Until then,
—Ali
P.S. The answer to last issue’s question, about what JavaScript concept bears a resemblance to that quote from Slow Man, is closures. A closure is this idea that a function that’s created within another function has access to the information in that outer function even if that outer function were to no longer be around. As in, when the function that gives birth to another function dies, the new function continues to bear that old function’s information so long as the new function is still alive.
Explaining the Enigmatic Anchoring Effect: Mechanisms of Selective Accessibility, bear.warrington.ufl.edu
A New Look at Anchoring Effects: Basic Anchoring and Its Antecedents, researchgate.net
Judgment under Uncertainty: Heuristics and Biases, psych.ubc.ca