Sophisticated choice questions and statistical methods are a way to obtain a good insight into people’s preferences. Statistical researcher Roselinde Kessels of the University of Amsterdam uses such methods to study topics from people’s expectations of a coach trip to how they view vaccinations.
Serene peace and silence reign in the courtyard and corridors of the University of Antwerp. There are no crowds of students sitting in cram-full study areas or queuing up for coffee. This setting suits Roselinde Kessels well. In Antwerp, where she works seventy percent of the time, she has been able for eight years now to devote herself fully to her research. Kessels, who is a Belgian national, owes this great freedom to a grant from FWO, the Flemish fund for academic research, and NWO, its Dutch sister organisation. She received the grant twice in a row.
Kessels combines her research position in Antwerp with a part-time job at the Faculty of Economics and Business at the University of Amsterdam (UvA). Kessels has been a lecturer in econometrics and a researcher at the University of Amsterdam for two years now. She obtained a degree in business engineering and a doctorate in statistics at KU Leuven. After, she did one year of research at North Carolina State University in the United States and worked for Procter & Gamble.
‘This method can be used to map consumer preferences.'
Procter & Gamble is exactly the kind of company that would share Kessels’ interests: people’s behaviour and preferences. Kessels conducts her research on the basis of so-called Discrete Choice Experiments (DCEs). In these experiments, a large group of people is asked a series of choice questions, with each individual being presented a different series. The various choice sets are designed very carefully and with the aid of statistical methods, which makes it possible to draw far-reaching conclusions on the basis of relatively simple research design.
‘The basis of the DCE method has largely been borrowed from marketing research,’ explains Kessels. ‘This method can be applied in many areas. It is used to map consumer preferences and also to measure the value of non-market goods, such as matters relating to safety, health and the environment.’
Many of the studies in which Kessels is involved focus on social policy issues. For example, she investigated the effectiveness of Dutch vaccination programmes for parents and their children. The respondents made choices concerning the effectiveness of the vaccine, the seriousness of the condition that it treats and the accessibility of the medicine. The researcher concluded that policymakers should offer broad access to medicine and emphasize the seriousness of the condition.
Although Kessels regularly collaborates with large companies who want to gain insight into the preferences of their customer base, she prefers to work in the academic world. ‘Many people with my background have employment in the private business sector, but I was not able to do fundamental research and further develop the DCE method in that environment. In the business sector, research of this nature is generally less in-depth.’
Kessels cites a recent study with ICB, the Belgian bus and coach institute, as an example of a DCE where she collaborated with the business community. She and her colleagues explored the preferences of people who travel by bus or coach or would like to do so. The 274 respondents completed questions online that consisted of a choice between two types of coach travel. In every virtual bus journey, a number of variables had changed, such as the presence or absence of internet and catering or the amount of legroom, while the price and duration of the journey also varied. For each choice set, the respondents selected their preferred bus journey, and they did so a total of sixteen times.
‘Proposed choice sets are all formulated in such a way that each variable can be weighted. It results in a utility model in which every variable can be expressed in financial value,’ says Kessels. ‘Discrete choice experiments make it relatively easy to determine what people prefer most. Just as expected, the price was the most important factor. Legroom and travel time are also factors that are considered to be of great importance.’
Such experiments are mainly carried out online or on paper, but the DCE method does attempt to imitate reality as closely as possible. ‘We do this by means of introductory, so-called “cheap talks,” which puts respondents in the mood and makes them answer more truthfully,’ says Kessels. ‘For example, in the case of the study about coach journeys, respondents could decide for themselves where they wanted to board and whether the trip was a business trip or a trip with friends.’
'Our research usually correctly identifies the main direction of the preferences, which is of great value, especially when making policy decisions.’
However, the real work is done right at the start, when setting up and selecting the choice sets, Kessels’ field of expertise. ‘Designing discrete choice experiments is at the core of my work. This means that the proposed choices should have the potential of leading to reliable and useful data. If illogical choices are proposed, you will end up with bad data and the results will not be reliable.’ Kessels uses Bayesian designs to improve the quality of the experiments. ‘The new designs that I have tested have always proved to be successful.’
When Kessels refers to ‘successful research,’ this concerns the statistical reliability of research. Whether the results are correct in practice is another matter. ‘The results of discrete choice experiments are difficult to test against reality and not much research has been done on them. In the real world, too many factors play a role, and on top of that, people's preferences are constantly changing. What counts is that our research usually correctly identifies the main direction of the preferences, which is of great value, especially when making policy decisions.’
More information? Email Roselinde Kessels.