P/S 04 · Heuristics, biases and experimental design
Anchors and Frames
Interpret experiments on anchoring, framing and time pressure.
Passage
People often make judgments quickly by using mental shortcuts called heuristics. Heuristics are efficient, but they can lead to systematic errors, or biases. In an anchoring effect, an initial number influences later estimates even when the number is irrelevant. In a framing effect, people choose differently between options with the same outcomes depending on whether the options are described as gains or as losses. Many psychologists describe thinking as having a fast, automatic mode and a slower, deliberate mode, and they expect biases to be stronger when people have little time to think.
Part 1. Researchers recruit 120 adults and randomly assign them to one of two groups of 60. Each participant is asked to estimate how many minutes it takes to drive from one end of a city to the other at midday. Before giving an estimate, participants in the low-anchor group are asked, ‘Does the drive take more or less than 20 minutes?’ Participants in the high-anchor group are asked, ‘Does the drive take more or less than 90 minutes?’ All participants then write down their own estimate. Table 1 shows the mean estimates.
The research assistants who recorded the estimates did not know which anchor each participant had been given. The expected value of an option is the sum of each outcome multiplied by its probability, a quantity that is used in Part 2.
Table 1. Mean estimated driving time by anchor condition.
| Group | Anchor (minutes) | Participants | Mean estimate (minutes) |
|---|---|---|---|
| Low anchor | 20 | 60 | 36 |
| High anchor | 90 | 60 | 62 |
Part 2. In a second study, a different set of 400 adults read a description of a town in which a factory closure will affect 600 jobs. Participants are randomly assigned to one of four groups of 100. All are asked to choose between two plans, A and B. In the gain frame, Plan A is described as ‘200 of the 600 jobs will definitely be saved,’ and Plan B as ‘a one-in-three chance that all 600 jobs are saved and a two-in-three chance that none are saved.’ In the loss frame, Plan A is described as ‘400 of the 600 jobs will definitely be lost,’ and Plan B as ‘a one-in-three chance that no jobs are lost and a two-in-three chance that all 600 jobs are lost.’ Half of the participants in each frame must decide within ten seconds, and the other half are given as much time as they want. Table 2 shows the percentage who chose Plan A, the sure option.
Table 2. Percentage of participants who chose the sure option (Plan A).
| Frame | No time limit | Ten-second limit |
|---|---|---|
| Gain frame | 72 | 80 |
| Loss frame | 38 | 25 |
After the study, one participant told the researchers, ‘I knew the drive was long, since I can easily remember the last three times I got stuck in traffic.’ The researchers note that the participants were not offered any payment based on the accuracy of their estimates or on the outcomes of their choices.
Questions
Question 1 of 6
What is the main conclusion that can be drawn from Table 1?
The two distances are 16 and 28. The comparison does not show which anchor is stronger, because the true value is not given.
Both means moved toward their anchors and differ from each other by 26 minutes. A single accurate drive time cannot produce two different means.
The participants made their own estimates, but the estimates differ in the direction of the anchors.
Why the answer is what it is
The group asked about 90 minutes gave a mean estimate of 62, and the group asked about 20 minutes gave 36. The only systematic difference between the groups was the anchor, so the anchor shifted the estimates.
Review: Cognition & Decision-Making · Social Structure, Inequality, Health, & Research Methods
Question 2 of 6
Why is random assignment important in Part 1?
Random assignment is about how participants are allocated to conditions. Representativeness depends on how they are sampled.
Assignment does not make estimates accurate.
Random assignment works together with a comparison group, not in place of one.
Why the answer is what it is
If participants chose their own group, the groups might differ in their knowledge of the city. Random assignment spreads such differences evenly across the groups, so that a difference in estimates can be attributed to the anchor.
Review: Cognition & Decision-Making · Social Structure, Inequality, Health, & Research Methods
Question 3 of 6
Plan A and Plan B in Part 2 have the same expected outcome. What is the expected number of jobs saved with Plan B (in the gain frame)?
This would be the result of taking 1/6 of 600. The chance of saving all 600 is one in three.
This is the number of jobs lost with Plan A in the loss frame, not the expected number saved with Plan B.
This is the number of jobs saved if the one-in-three outcome occurs. The expected value must take the probability into account.
Why the answer is what it is
The expected number of jobs saved with Plan B is (1/3) × 600 + (2/3) × 0 = 200, the same as the 200 jobs that Plan A saves for certain. In the loss frame, Plan A also leaves 200 jobs saved (600 − 400), so the two frames describe the same choices.
Review: Cognition & Decision-Making · Social Structure, Inequality, Health, & Research Methods
Question 4 of 6
Which statement best describes the pattern with no time limit in Table 2?
The table shows the opposite: 38% in the loss frame and 72% in the gain frame.
This is the reverse of the pattern. The sure option was chosen by 72% in the gain frame.
The two values differ by 34 percentage points, which is a large effect.
Why the answer is what it is
Plan A and Plan B are the same in both frames, but the sure option was chosen by 72% in the gain frame and by only 38% in the loss frame. People tend to avoid risk when choices are described as gains and to accept risk when they are described as losses.
Review: Cognition & Decision-Making · Social Structure, Inequality, Health, & Research Methods
Question 5 of 6
With a ten-second limit, the difference between the frames is 55 points, compared with 34 points with no limit. This finding is most consistent with the idea that:
The size of the framing effect is different with and without a time limit.
The effect grew with the time limit, which reduces deliberate thinking.
The study did not measure anxiety, so the data cannot show it.
Why the answer is what it is
Time pressure limits slow, deliberate thinking, and the framing effect grew from 34 to 55 points. Fast, automatic processing is more strongly affected by how a choice is framed.
Review: Cognition & Decision-Making · Social Structure, Inequality, Health, & Research Methods
Question 6 of 6
The participant who said, ‘I knew the drive was long, since I can easily remember the last three times I got stuck in traffic,’ is best described as using:
The representativeness heuristic is about judging by similarity to a typical example. The participant based the judgment on the ease of recalling examples.
Nothing in the statement shows that the participant ignored contrary evidence.
The sunk cost fallacy concerns continuing to invest because of past costs, and it does not describe this statement.
Why the answer is what it is
The participant judges the length of the drive by how easily examples of traffic come to mind. Judging frequency or likelihood by ease of recall is the availability heuristic.
Review: Cognition & Decision-Making · Social Structure, Inequality, Health, & Research Methods
Built from these lessons