Conference material

Behavioural Insights: Fact or Fad?

Plenary session 1 consisted of a short presentation, followed by a discussion session among the panellists as well as Q&A with the audience. Read on for the summary of the discussion points of plenary session 1. For more details on the presentation, you can follow the link to the presentation slides.

Date Posted

1 Jan 0001


Issue 17, 14 Jun 2017

Plenary Session 1: Behavioural Insights (BI): Fact or Fad? Will BI ever become a mainstream policy tool?

By Professor Dilip Soman, Professor Ho Teck Hua, Dr Rory Gallagher, Dr Leong Wai Yan and Mr Ivan Yeo (Moderator)

Short Presentation: Behavioural Nudges — Examples and Issues

Before the commencement of the plenary session, Professor Ho shared some of his work on using behavioural nudges to promote good health and environmental sustainability in Singapore. They include studies on getting people to stop littering, and incentivising taxi drivers and overweight people to exercise. Some interesting findings were:

  1. Incentivising taxi drivers to exercise. Taxi drivers who participated in the rental credit treatment (they received credit that could be used to offset their daily rental fee if they met a physical activity goal for that month) walked more than their counterparts in the cash treatment, even though both conditions offered the same monetary incentive. The result suggests that incentive targeting a salient and painful expenditure is more effective.
  2. Incentivising overweight people to exercise. Heterogeneous treatment effects were observed; the incentive for weight loss was effective at getting male participants to lose weight but seemed to have a limited effect on female participants. Professor Ho opined that this could be because female participants’ main motivation for exercising was their health rather than financial incentives. The study found indications of habit formation among participants; they continued to lose weight in the post-intervention period, when the incentives had been removed.
  3. Getting People to Stop Littering. The peer-enforcement treatment (the “Tell Others Not to Leave Their Rubbish Behind” poster) was more effective than the altruistic treatment (the “Tell Yourself Not to Leave Your Rubbish Behind” poster). The study found indications of habit formation in both treatment groups as the change in overall littering behaviour persisted even after the messaging campaign had concluded.

Professor Ho discussed three central issues in the field of behavioural insights. Firstly, on short-term versus long-term behavioural changes, he felt that most studies focused on the former but not the latter. He opined that if a certain behavioural intervention was not able to result in habit formation, it might not justify the cost of the intervention. Secondly, on heterogeneous treatment effects, practitioners must develop a methodology to customize treatments for different subjects. Lastly, on scaling up habit formation, practitioners must continue to explore scaling up successful behavioural interventions to engineer habit formation on a national level.

Link to Professor Ho’s presentation slide

Plenary Discussion

The plenary discussion spanned a few key themes, from ways to help policymakers acknowledge their own biases and the importance of behavioural insights (BI), to possible ways to “mainstream” BI and address the related challenges. Read on for a summary of the key insights.

Policymakers have biases too

  • Policymakers must be aware of their own biases so that they can take concrete steps to manage them. Otherwise, they run the risk of introducing bias into policy design.
  • Given this, the best way to make informed decisions is to be empirical and experimental. This could involve the use of machine learning techniques to collect data to show which types of interventions work best for certain groups of participants. Despite being non-causal, such information can help policymakers make more informed decisions.
  • Machine learning, however, could sometimes lead to public backlash, e.g.: in the case of Cambridge Analytical1 when machine learning resulted in too personalised and targeted interventions. The government needs to guard against behavioural interventions that are too predictive and targeted as they could erode public trust and confidence in the institution.

The importance and role of BI

  • BI is a good complement, rather than a substitute, for classical economics. For example, UK supermarkets were able to decrease single-use plastic bag usage substantially since a 5-pence charge had been introduced in 20152 and this feat could not have been achieved alone by BI. Furthermore, Pigouvian tax is a tried-and-tested method that is easily scalable and implementable. Nevertheless, it is important for the government to communicate to the public that such taxes are used to discourage undesirable behaviours and not to increase the government’s coffers. One possible way to mitigate public unhappiness about such taxes is to direct the collected revenue to fund meaningful initiatives.
  • Complementing classical economics, BI should be an integral part to policy design because policies are designed for humans, and not econs. But in order for BI to become a mainstream policy tool, policymakers need to stop viewing it as a separate entity and start fully integrate human psychology into their policy making process.
  • Behavioural interventions are highly contextual and the experimentation process can be lengthy before a specific BI treatment is ready to be implemented on a national level. For this reason, policymakers need to anticipate future research needs, especially in the realm of BI, since they might not have the luxury of time to conduct experiments when the issue arises.

How can BI be a mainstream policy tool?

  • From an organisational perspective, the cost of experimentation must decrease to increase the willingness and ability to conduct experiments (based on the simple law of Demand and Supply). These costs include substantial cultural costs such as ethical considerations and getting people to accept no result or negative results, as well as the cost of collecting and analysing data. In Canada, the government established a Research Ethics Board to serve as a check-and-balance in the system and practitioners have been receiving strong support from the Canadian government.
  • In addition, practitioners need to have patience in terms of timelines, and cultivate trust by working closely with policymakers and understanding their considerations.
  • From an application perspective, there are two ‘lenses’ through which BI can be a mainstream policy tool:
  1. Look out: Policymakers can use BI when working with external parties to influence and regulate firms’ behaviour. For example, Instagram had faced scrutiny over various issues such as cyberbullying and possible impact on mental health. Policymakers could use BI to influence the market by mobilising advocates to confront errant firms. Such an approach had been successful in getting firms to change their policies.3
  2. Look in: Rather than viewing BI only as a policy tool to change the behaviour of citizens, practitioners should also explore the possibility of using BI to change the behaviour of policymakers who are subjected to the same biases. By acknowledging their biases, policymakers are then able to address them and make more informed decisions.

Challenges and addressing them

  • The diminishing effect of nudges: The effect of nudges may decline over time if citizens are more aware of the government using nudges to steer behaviours. But it is also worth noting that choice architecture would become more effective as we refine our understanding of it. In some contexts, becoming more aware of nudges could actually steer people towards the right behaviour because it conforms to social norms and is viewed as the “right” decision by the citizens themselves.
  • Practitioners should not be too concerned about the possible diminishing effect of nudges as it might not even matter. A thought experiment to illustrate this point would be to envision a world where the government has successfully nudged its citizens to choose healthier products and recycle their trash. After these habits become a social norm, citizens will automatically engage in these behaviours without nudges, and pass them on to their children. In such a world, the government would no longer need to nudge its citizens and the question on its effectiveness is hence rendered irrelevant.
  • Failed experiments: A pertinent question for policymakers is how they can continue to affirm the value of evidence-based approach to policy design when their experiment “fails”. First, valuable insights can still be drawn from these “failed” experiments, including identifying barriers impeding the success of the experiment (e.g. from a design or implementation process) as they can be shared with other policymakers working on a similar topic. This is a possible way to frame the value of a “failed” experiment, rather than simply being resigned to an unfavourable outcome.
  • Second, the result that something would not work in itself, is a valuable insight, as it would be better to know that a possible intervention does not work at the experimental stage, than at full roll-out of a policy. These factors affirm the importance of weaving a culture of experimentation into policy work, as part of the ecosystem that supports evidence-based policymaking.
  • More importantly, there is a tendency for sharing only successful trials, without a balance from those that have failed. This is a serious problem as policymakers might develop an over-optimistic view of behavioural interventions.
  • Cross-Sharing: There were suggestions for policymakers to capitalize on various experiments conducted service-wide by setting up a “database” for sharing. This would facilitate peer-to-peer learning and provide a better assessment of how behavioural interventions work in different contexts, including their scalability.
  • Fixation on BI: While BI is a powerful tool, practitioners have a tendency to assume the problem lies in the last-mile, without considering if it is better to focus on the first-mile instead. Whenever possible, they should look upstream and change the environment since it is less effective to get people to change their behaviour via the use of choice architecture. For example, the UK Sugar Tax4 aimed at manufacturers instead of consumers, thereby reducing the sugar content of the products upstream rather than encouraging people to choose healthier alternatives.


  1. Z. Kleinman, “Cambridge Analytica: The story so far,” 21 March 2018.
  2. R. Morelle, “Plastic bag use plummets in England since 5p charge,” 30 July 2016.
  3. K. Steinmetz, “Inside Instagram's Ambitious Plan to Fight Bullying,” 8 July 2019.
  4. N. Triggle, “Soft drink sugar tax starts, but will it work?” 6 April 2018.