A lot of organizations need to learn about the public's issue priorities, their support for or opposition to different policies and candidates as well as their responsiveness to potential messages.

Empirical Intelligence provides organizations with the data they need using the most appropriate methods. Some of the major methodologies include Likert scales, MaxDiff, experimental research, and conjoint analysis.

Traditional Survey Questions

There are many different open-ended and closed-ended question formats that are best for different types of projects. Some of the most popular formats are Likert scales, which are used to assess the intensity of people's attitudes, and numerical rating scales, which are used to get a quantitive measure of public opinion.

A major challenge with many of traditional types of survey questions is that respondents interpret the scales in different ways. Differences in scale interpretation can reflect differences across individuals and differences across cultures.

MaxDiff Analysis

MaxDiff (also called best-worst scaling) is a method that allows us to learn about respondents’ evaluations of a list of items. How much do they like each item or how important do they think each of them is?

MaxDiff has several huge advantages over more traditional survey questions:

• The task is simpler for respondents: they only have to say which item is best and worst or most and least important.
• It allows us to ask respondents about a long list of items (usually up to about 40). Respondents only see 4 or 5 on each screen. That way, we get data about a lot more items while keeping the task simple for respondents.
• It prevents respondents from telling us that they like everything or that everything is important.
• It avoids the problem of respondents interpreting scales differently by only asking them about the best and worst or most and least important item out of each list.

Experimental Research (Monadic Testing)

Survey respondents are randomly assinged to one of two or more versions of a stimulus. We then measure public opinion in each group and compare the results. This is a great approach for testing messages or for testing the influence of factors on public opinion.

Conjoint Analysis

Conjoint analysis is a more sophisticated type of experiment that shows survey respondents hypothetical policies or candidates with varying features or attributes. They are asked to select one (usually they can select none as well). Respondents go through several such choice tasks. We then combine all the data and calculate how much more or less particular characteristics make people support a candidate or policy.

Conjoint analysis also allows us to build a simulator that shows us how well a candidate or policy does compared to the competition. We can adjust the attributes of the candidate or policy until we find the optimal offer that maximizes public support.