CHAPTER 2

Given the need to test for collective rationality in a wide variety of opinion distributions for which the sizes and shapes of the signal and noise distributions are known, a computer simulation was used to generate the raw data used for Chapter 2's analysis. The units of analysis in the simulation are collective opinion distributions made up of 1,500 respondents divided into signal and noise groups. The signal distribution is defined as the aggregated preferences of those hypothetical respondents possessing a balance of factually correct information. The noise distribution is defined as the aggregated opinions of those hypothetical respondents drawing from a balance of factually incorrect or misleading information as well as of those possessing no relevant information at all.

To assess how different ways of collapsing a given opinion distribution might affect the accuracy of signal estimates, opinion data were generated into a six-category response scale. Although it is unlikely that a six-point scale would be used in survey research, this response scale has a desirable property for studying the effects of aggregation: six categories can be collapsed directly into a dichotomous scale (treating the first three categories as one option and the second three categories as a second option) as well as into a three-point scale (with categories one and two as a first option, three and four as a middle option, and four and five as a third option).