Choice modeling is a method used to determine how an individual or group would respond to a specific choice in a specific situation. Unlike some simulations, choice modeling takes into account environmental, marketing and society-based influences on choice in addition to other common factors. These processes are often used as a method of increasing a company’s profits through a better understanding of their customer base.
While choice modeling can be very accurate, an experiment is only as good as its design. For a full and accurate representation of choice, it is important that the underlying experiment is as complete as possible. As a result, there are several steps and techniques used to make a good simulation.
The first step in a choice modeling experiment is setting up the experimental framework. During this phase, modelers determine the exact goals and methods of the simulation. By keeping the focus of the model as narrow as possible, they can make the experiment more accurate.
There are two main points that come up at the end of this phase that influence the rest of the experiment. First is determining how many people the experiment needs in order to be accurate. With too many people, the responses become bloated with extra information, but with too few, there won’t be an accurate cross section.
The other point of importance is the experimental questionnaire. The questions need to be focused and to the point without being leading or suggestive. Older choice modeling procedures used reveal preference questions, while newer ones often use stated preference. If everything is equal, the two techniques would reveal the same answer, but that doesn’t mean each doesn’t have its own strengths.
In revealed preference systems, questions attempt to bring out information that the question answerer may not know on a conscious level. Information about tendencies and mental state are often best found in this manner. Stated preference systems ask direct questions where those answering rate their like and dislike of specified objects and situations. These work well when learning about buying preferences and personal habits.
Once the questionnaire is finished and the number of participants is determined, the testing begins. During this experiment, individuals are brought in to answer questions. Their answers are supposed to represent a specific group of people; single white males, women making over a certain income per year, and so on.
The respondents’ information is then placed into a simulation. The most common form of choice modeling uses each of the recipient’s answers to make a specific person. For example, if there were 50 people studied, there would be 50 computer-modeled individuals. These simulated people are confronted with a series of choices, and the model spits out a response. This is repeated thousands of times to get an aggregate total response to the situation.
