Can some of the anchoring effect can be explained by the use of a kind of implicit confidence interval?
Suppose that I (subconsciously) have an estimate of 20% for the proportion of UN countries that are African. Further suppose that I think a 95% confidence interval ranges from 10% to 30%.
If I start at a high anchor, I will adjust downwards until I’m within the 95% CI, i.e., 30%. If I start at a low anchor, I adjust upwards until I’m within the 95% CI, i.e., 10%. In my head, I may consider 10% and 30% as not statistically different from one another.
I’m not talking about exact statistical inference, but I wonder if this process is part of what’s going on in the subject’s head.
I have tried a classroom bargaining experiment, where I give random “valuations” to students. I then assign random ownership (so that half the class become sellers). Without knowing what the item is (it’s just “some good”), the initial offerers tend to have a disadvantage because they use their own valuations as anchors.
When I change the setup by telling them that “it’s a used Toyota,” the final bargained prices tend to more closely (but not perfectly) split the surplus.
I’m reminded of a story that my father tells about being in the army and learning to shoot. After missing the target, the instructor told them to use “bold sight adjustments” because shooters tend to be too timid in adjusting their aims. The phrase “bold sight adjustments” became part of our family vocabulary.