Standard practice in discrete choice experiments is to present information on alternatives in matrix form. However, how respondents are asked to read the matrix is at odds with learned visual routines and reading patterns, and this can impact respondents’ choices.
A discrete choice experiment (DCE) is a common stated preference technique used to elicit willingness-to-pay (WTP). In a DCE a respondent is asked to choose between several alternatives described by multiple attributes. For example, consider a choice between two types of coffee from a vending machine (alternatives). The coffee can be organic or fair trade, the cup can be recyclable and you can pay using a credit card (attributes). Standard practice is to show this information in a matrix where each row is an attribute and each column an alternative. Respondents are then asked to consider each alternative and all attribute information, i.e. they are asked to read from top-to-bottom column-by-column. Problematically, this is at odds with how people are taught to read. Several studies show that learned reading patterns dominate when people search for and process information, for example, items at the top of a list receives more attention than items at the bottom of a list and items in the top left corner of a matrix draws attention first, and search proceeds from there.
-To explore the implications of how we present alternatives to respondents we ran a simple split sample experiment where half the respondents were randomly allocated to the standard matrix display and half the respondents to a transposed matrix display, where each row was an alternative and each column an attribute – bringing theory in line with learned reading patterns, says Erlend. The results show that the two ways of displaying alternatives provides different preferences, but that the commonly observed ordering effect disappears. This is an effect that essentially shows a higher probability of choosing alternative one over alternative two over alternative three, when they are ordered from left to right. Furthermore, it became evident that respondents’ use of simplifying strategies was not affected by how information was presented, which suggests that this is a more fundamental process. - Finally, Erlend explains, we found evidence that comparisons by attributes rather than alternatives is more apparent under the standard matrix display compared to the transposed one. However, more research is needed and the paper outlines a few potential fruitful avenues for exploration.
These findings were presented in a paper at the Workshop on Non-Market Valuation (WONV) in Leeds earlier this year and won the Best Paper Award.
The paper: The Effect of Display Orientation on Processing Strategies, was presented by CERE's Erlend Dancke Sandorf, one of the authors of the paper.