Econ 522 – Dan Quint – March 2009 – Results from the In-Class Experiment

Tuesday’s in-class experiment was borrowed from Victor Matheson (2005), “Rationality, Tort Reform and Contingent Valuation: A Classroom Experiment in Starting Point Bias,” College of the Holy Cross Working Paper #05-09. Everyone was given a handout with the following story:

You have been asked to serve on a jury on a lawsuit dealing with personal injury. In the case before you, a 50-year-old construction worker was injured on the job due to the negligence of his employer. As a result, this man had his right leg amputated at the knee. Due to this disability, he cannot return to the construction trade and has few other skills with which he could pursue alternative employment.

The negligence of the employer has been firmly established, and health insurance covered all of the related medical expenses. Therefore, your job is to determine how to compensate this worker for the loss of his livelihood and the reduction in his quality of life.

Half the class was asked the following three questions:

(a) Should the plaintiff in this case be compensated more or less than $10,000?

(b) How much should the plaintiff receive?

(c) Are you male or female?

The other half of the class was asked the same three questions, but with “$10,000” replaced by “$10,000,000” in question (a).

The point of the experiment was to show that people’s opinion of what is fair (their answer to question (b)) is influenced by which version of question (a) they were asked. On average, those who were given the higher “prompt” – were asked about $10,000,000 – proposed damages which were three times higher than those given the lower “prompt”:

Full
Sample / Asked
$10,000 / Asked
$10,000,000
number of observations / 58 / 32 / 26
average / 1,988,362 / 1,030,313 / 3,167,500
geometric mean / 895,111 / 579,870 / 1,527,334
10th percentile / 220,000 / 160,000 / 675,000
25th percentile / 600,000 / 412,500 / 750,000
50th percentile (median) / 750,000 / 750,000 / 1,750,000
75th percentile / 1,750,000 / 1,000,000 / 5,000,000
90th percentile / 5,000,000 / 1,500,000 / 6,000,000

(A number of people gave formulas – such as “15 years salary”; I converted these to numbers using reasonable guesses at salary, retirement age, pension, etc. The results are similar if you look only at the people who reported numbers.)


Here is the distribution of peoples’ answers to question (b), separated by which version of question (a) they were asked:

So 41% of those given the low prompt, but only 12% of those given the high prompt, suggested damages less than $700,000; 81% of those given the low prompt, and 35% of those given the high prompt, suggested less than $1.4 million; and 35% of those given the high prompt, and just one (out of 32) given the low prompt, went above $3 million.

The takeaways:

1.  There is broad disagreement over how much a leg is worth

2.  How people are asked the question can have a big effect on their answer