Anchoring bias is one of the most robust effects in psychology. Many studies have confirmed its effects, and shown that we can often become anchored by values that aren’t even relevant to the task at hand. In one study, for example, people were asked for the last two digits of their social security number. Next, they were shown a number of different products, including things like computer equipment, bottles of wine, and boxes of chocolate. For each item, participants indicated whether they would be willing to pay the amount of money formed by their two digits. For example, if somebody’s number ended in 34, they would say whether or not they would pay $34 for each item. After that, the researchers asked what the maximum amount was that the participants would be willing to pay.
Even though somebody’s social security number is nothing more than a random series of digits, those numbers had an effect on their decision making. People whose digits amounted to a higher number were willing to pay significantly more for the same products, compared to those with lower numbers.9 Anchoring bias also hold up when anchors are obtained by rolling some dice or spinning a wheel, and when researchers remind people that the anchor is irrelevant.4
Given its ubiquity, anchoring appears to be deeply rooted in human cognition. Its causes are still being debated, but the most recent evidence suggests that it happens for different reasons depending on where the anchoring information comes from. We can become anchored to all kinds of values or pieces of information, whether we came up with them ourselves or we were provided with them,4 but apparently for different reasons.
When we come up with anchors ourselves: The anchor-and-adjust hypothesis
The original explanation for anchoring bias comes from Amos Tversky and Daniel Kahneman, two of the most influential figures in behavioral economics. In a 1974 paper called “Judgment under Uncertainty: Heuristics and Biases,” Tversky and Kahneman theorized that, when people try to make estimates or predictions, they begin with some initial value, or starting point, and then adjust from there. Anchoring bias happens because the adjustments usually aren’t big enough, leading us to incorrect decisions. This has become known as the anchor-and-adjust hypothesis.
To back up their account of anchoring, Tversky and Kahneman ran a study where they had high school students guess the answers to mathematical equations in a very short period of time. Within five seconds, the students were asked to estimate the product:
8 x 7 x 6 x 5 x 4 x 3 x 2 x 1
Another group was given the same sequence, but in reverse:
1 x 2 x 3 x 4 x 5 x 6 x 7 x 8
The media estimate for the first problem was 2,250, while the median estimate for the second was 512. (The correct answer is 40,320.) Tversky and Kahneman argued that this difference arose because the students were doing partial calculations in their heads, and then trying to adjust these values to get to an answer. The group who was given the descending sequence was working with larger numbers to start with, so their partial calculations brought them to a larger starting point, which they became anchored to (and vice-versa for the other group).5
Tversky and Kahneman’s explanation works well to explain anchoring bias in situations where people generate an anchor on their own.6 However, in cases where an anchor is provided by some external source, the anchor-and-adjust hypothesis is not so well supported. In these situations, the literature favors a phenomenon known as selective accessibility.
The selective accessibility hypothesis
This theory relies on priming, another prevalent effect in psychology. When people are exposed to a given concept, it is said to become primed, meaning that the areas of the brain related to that concept remain activated at some level. This makes the concept more easily accessible, and more able to influence people’s behavior without their realizing.
Just like anchoring, priming is a robust and ubiquitous phenomenon that plays a role in many other biases and heuristics—and as it turns out, anchoring might be one of them. According to this theory, when we are first presented with an anchoring piece of information, the first thing we do is to mentally test whether it is a plausible value for whatever target object or situation we are considering. We do this by building a mental representation of the target. For example, if I were to ask you whether the Mississippi River is longer or shorter than 3,000 miles, you might try to imagine the north-south extension of the United States, and use that to try to figure out the answer.7
As we’re building our mental model and testing out the anchor on it, we end up activating other pieces of information that are consistent with the anchor. As a result, all of this information becomes primed, and more likely to affect our decision making. However, because the activated information lives within our mental model for a specific concept, anchoring bias should be stronger when the primed information is applicable to the task at hand. So, after you answered my first Mississippi question, if I were to follow it up by asking how wide the river is, the anchor I gave you (3,000 miles) shouldn’t affect your answer as much, because in your mental model, this figure was only related to length.
To test this idea, Strack and Mussweiler (1997) had participants fill out a questionnaire. First, they made a comparative judgment, meaning they were asked to guess whether some value of a target object was higher or lower than an anchor. For example, they might have been asked whether the Brandenburg Gate (the target) is taller or shorter than 150 meters (the anchor). After this, they made an absolute judgment about the target, such as being asked to guess how tall the Brandenburg Gate is. For some participants, however, the absolute judgment involved a different dimension than the comparative judgment—for example, asking about a structure’s width instead of its height.
The results showed that the anchor effect was much stronger if the object dimension was the same for both questions,7 lending support to the theory of selective accessibility. This does not mean that the anchor-and-adjust hypothesis is incorrect, however. Instead, it means that anchoring bias relies on multiple, different mechanisms, and it happens for different reasons depending on the circumstances.
Bad moods weigh us down
The research on anchoring has turned up a number of other factors that influence anchoring bias. One of these is mood: evidence shows that people in sad moods are more susceptible to anchoring, compared to others in good moods. This result is surprising, because usually, experiments have found the opposite to be true: happy moods result in more biased processing, whereas sadness causes people to think things through more carefully.4
This finding makes sense in the context of the selective accessibility theory. If sadness makes people more thorough processors, that would mean that they activate more anchor-consistent information, which would then enhance anchoring bias.8