Foreword
At TDL, our role is to translate science. This article is part of a series on cutting edge research that has the potential to create positive social impact. While the research is inherently specific, we believe that the insights gleaned from each piece in this series are relevant to behavioral science practitioners in many different fields. At TDL, we are always looking for ways to translate science into impact. If you would like to chat with us about a potential collaboration, feel free to contact us.
Introduction
The concept of nudging has recently grown in popularity. This is partially due to how exciting and innovative these types of interventions can be. But, what might be more important than their innovativeness and excitability is if they actually work. And if they do, which conditions are important for implementing nudges, and what can we learn from studying them on a large scale?
As an applied behavioral science research firm, The Decision Lab is interested in learning more about the effectiveness of nudges and how they can be better implemented to drive social change. To further this interest, we reached out to Dr. Dennis Hummel and Prof. Alexander Maedche to learn about their work on studying the effectiveness of nudges and their attempt at classifying them with the purpose of guiding future research.
A full version of some of Dennis and Alexander’s studies are available here:
Improving Digital Nudging Using Attentive User Interfaces: Theory Development and Experiment Design
How would you describe the focus of your research in simple terms?
Nudges are “any aspects of the choice architecture that alters people’s behavior in a predictable way without forbidding any options or significantly changing their economic incentives”.1 As nudges are very popular, they affect our everyday lives. For example, through governmental nudge units, such as the Behavioral Insights Team in the UK or the former Social and Behavioral Sciences Team in the US. Yet, it has never really been investigated on a large scale if nudges really work and, if so, under which conditions. One of the authors of the original nudging book has even dedicated a separate journal paper on “nudges that fail”.2 With our research,3 we aimed to estimate the effectiveness of nudging. Moreover, we wanted to design a classification system that can serve as a guide for future nudging studies.
How would you explain your research question to the general public?
Well, we follow two main research questions. On the one hand, we want to judge whether the hype around nudging can be documented with scientific data. We ask ourselves by how much nudges decrease or increase an outcome compared with a control group that received no nudge.
In addition, we want to know what would be the influencing factors of a divergent effectiveness. As nudging is a broad concept, some types of nudges or contexts could be more effective than others. On the other hand, we asked ourselves whether there is a way to classify all nudges studies into one comprehensive system, such as a taxonomy or a morphological box.
What did you think you’d find, and why?
As we followed an explorative approach, we did not formulate any explicit hypotheses. However, we of course expected that nudging would be highly effective. Based on previous literature reviews on nudging, we thought that in particular defaults would be among the most effective types of nudges. Moreover, we expected that nudges might differ by the context, for example, energy, the environment, etc. and by distinguishing between offline nudging and digital nudging which is a rather new concept introduced by Weinmann et al.4 Finally, we also hoped to find rather practical information such as in which countries the fewest nudging studies have been conducted or which types of nudges have been used rarely to offer avenues of future research to other researchers. As for the classification system, we were entirely curious and open as taxonomies or morphological boxes are developed along the process.
What sort of process did you follow?
We first conducted a systematic literature review. Literature reviews are performed about as follows: After defining a goal, a search strategy, keywords and databases, we ran a keyword combination in several academic databases. We had a broad set of keywords and found about 2,500 papers which then had to be screened based on the title, the keywords and the abstract. After the screening, we read 280 papers in full to distill the 100 relevant papers for our analysis (it was really a coincidence that it ended up on such a round number). These papers were then analyzed in detail extracting the type of the nudge, the effect size, the context and other relevant information. In the end, we created a database, which is available on request, with more than 300 different nudging treatments and extracted more than 20 characteristics for each treatment. To design the morphological box, we followed the recommendations from Nickerson et al.5
What did you end up finding out?
We found much more than we could ever present in one academic paper. First, our analysis revealed that only 62% of the nudging treatments are statistically significant which is much lower than we initially expected. Nudges have a median effect size of 21%, which depends on the type of nudge and the context. As expected, defaults are the most effective nudges while precommitment strategies (i.e. you commit now to do something in the future) are the least effective. Moreover, digital nudging is similarly effective as offline nudging but it offers new possibilities to individualize nudges. This means that digital nudges can be adapted more easily to the individual characteristics of the decision-makers (see a brand-new study for more information: Ingendahl et al., 2020). Finally, we developed a morphological box which categorizes empirical nudging studies along eight dimensions.
