Discrete Choice Modeling: Choices Speak Louder Than Words

September 25, 2018

By: Evan Tincknell

An important yet difficult challenge for any energy-efficiency program is the ability to quantify savings directly attributable to program offerings. To do so, one needs to understand the degree to which customers would have taken comparable energy-saving measures in the program’s absence. Other valuable insights such as which aspects of a product or offering people find most enticing or how much customers would be willing to pay are key to implementing an effective program. Every purchase involves choice, and those choices are rarely as simple as they appear. While it’s nearly impossible to define an individual’s exact purchase behavior, discrete choice modeling can reveal complex patterns in the choices that groups of people make.

Developed by economists and psychologists in the late 1970’s, discrete choice modeling pairs a specialized survey design with regression-based analytics to better understand and predict customer preferences under various market scenarios. A discrete choice survey presents participants with a series of ‘choice sets’ and then asks the participants to make a series of hypothetical purchase decisions, choosing between several products with varying characteristics. By aggregating the outcome of many individual purchase decisions, discrete choice analysis can predict the types of products customers find most attractive, how purchase tendencies would shift in the absence of discounts, and which products are most effective to discount. Common applications of this method include pricing analysis, product concept testing, product branding or positioning, and market share forecasting. In market research applications, discrete choice experiments are most often used with big-ticket items, such as airline tickets or cars.

One particularly powerful application for discrete choice modeling within the energy sector is to estimate the net impacts of energy-efficiency programs. This modeling can be used effectively to optimize rate structure offerings, pricing plans, design effective demand-response, and energy-efficiency program offerings, as well as assess market conditions and remaining market potential for a product, service, or program.

In recent energy-efficiency evaluations, we have used discrete choice surveys to assess price sensitivity of program-discounted energy efficient light bulbs. Although light bulbs are a relatively low-cost item and a fairly routine purchase decision for most shoppers, different customers focus on different product characteristics. Some might only consider price or gravitate toward the cheapest available product of a given wattage, while others may care more about light color, expected bulb life, or energy savings. The discrete choice survey presents respondents with a few product options with varying characteristics and uses hundreds of decisions from different customers to interpret preferences across products. The results allow us to simulate markets featuring either discounted or non-discounted energy efficient bulbs, in effect projecting how market shares would shift in the absence of the program.

In addition to modeling price sensitivity and predicting market shift with different price options available, the discrete choice model also measures the relative importance of various product characteristics. In the case of light bulbs, we can include light color or expected bulb life in the survey and then use the results to predict which light colors are most popular or how strong preferences are for light color relative to bulb life. We can also estimate how similar products fare or the relative importance of price for different types of products, which can help identify what products are most impactful to discount and what level of discounting is most effective.

While the discrete choice method relies on customer self-report, it’s methodology avoids many of the biases associated with more direct survey questions about decision-making processes or willingness to pay. By asking customers to make trade-offs between price and other non-price attributes, this method reveals the true effect of price or other considerations on customer choice and avoids much of the social desirability bias commonly associated with more direct questions on product pricing.

Discrete choice modeling analysis offers a wide array of analytic possibilities and insight into customer behavior. It is a time-proven method that holds tremendous promise in the relatively new area of research for energy-efficiency program evaluators. Energy-efficiency programs seeking to influence larger, more carefully considered purchase decisions may, therefore, look towards discrete choice modeling experiments offer a wide array of analytic possibilities and are readily customizable, making them a promising tool for informing both program design and program evaluation. In the context of constantly shifting markets, we look forward to helping our clients take advantage of this innovative and flexible approach to maximize program efficacy and provide their customers with the products they value most.