Book Review: Stumbling on Happiness - Page 2

Other studies Gilbert cites involve pizza, ice cream, completion of boring tasks, failure to receive a gift certificate, and immersion of part of the body in ice water. In study after study, volunteers’ actual pleasure or pain resembled more closely the average pleasure/pain levels of a previous group of volunteers than their own actual predicted feelings. Thus, Gilbert argues, the only way to predict our own future happiness is to ask someone who is experiencing the event we are considering and ask them about their feelings.

There are two problems with this argument. The first is that Gilbert ignores the concept of variance in the studies he cites. In the studies Gilbert describes, average participant satisfaction levels mirror average prior participant satisfaction levels pretty closely. As anyone with even a rudimentary knowledge of statistics can tell you, two samples with the same average can differ dramatically in their variance, that is the range and frequency of high to low responses. Further, sampling responses tend to be normally distributed in a bell-shaped curve around an average response, when in fact the actual distribution of the population may look very different.

For instance, surveying a sample of Americans on their opinion of George W. Bush on a scale of 1-5 might lead you to believe that we’re all pretty neutral about the president, when in fact roughly two thirds of the country feels he’s incompetent and one third feels he’s the right man for our times. Given the strength of emotion about the president, you’d likely find far more responses on the poles (1s or 5s) than in the middle. By looking at the average of responses without considering the variance, you’d miss this polarity and come to the mistaken conclusion that most of us are without strong emotion either way.

Gilbert presents studies that report only average satisfaction levels and omits information on the variance of participants’ responses. Therefore, depending on which individual we choose to ask about their actual satisfaction level of an event they are experiencing presently, we could get extremely varying responses. If we had the luxury to ask enough individuals to get a statistically valid sample size, Gilbert’s argument might hold water. Instead, however, he suggests that asking one person would be sufficient.

Article Author: Justin McHenry

Justin McHenry is president of Index Credit Cards, a site offering credit card comparisons, research, tips, calculators and a listing of over 1,200 currently-available credit cards.

Visit Justin McHenry's author page — Justin McHenry's Blog