Chapter 1. Overview and key lessons

Smart meters and electricity consumption

Chapter 2 examines the increasing popularity of smart meters. In the past decade, governments have rolled out smart meters to replace traditional analogue meters in many regions of the world. In contrast to analogue meters, smart meters track real-time energy use and automatically send data to energy suppliers. When coupled with feedback technologies such as in-home displays (IHDs), consumers can easily access real-time electricity consumption and time-use pricing inside their home. Smart meters linked to IHDs offer the promise of reducing energy usage by giving consumers control over their usage levels and supporting time-of-use tariffs that can help to spread the demand for electricity more evenly throughout the day.

Although the features of smart meters seem promising for energy conservation, the empirical literature reveals mixed results ranging from no change to a 17% reduction in energy consumption. There have been a considerable number of studies measuring the impact of smart meters, but the earlier studies lacked rigour. More recent studies using (quasi-)experimental approaches with meaningful sample sizes have found that results are context-specific with limited external validity. There remains a lack of clarity on how real-time feedback affects electricity consumption.

To contribute to the gap in the literature, the study shares findings from a robust quasi-experimental field experiment in Ontario, Canada. With an exemplary sample size of 7 000 households, researchers found that during the two-year period of the study, real-time feedback led to an overall 3% decrease in consumption but no significant shifts in consumption patterns. In addition, the study shows adjustments were made as one-time decisions rather than on a continuous basis as reactions to time-of-use prices. These results challenged their theoretical model, which was based on a “rational” consumer who made more optimal energy consumption decisions when given higher quality and quantity of information.

These findings reflect the advantages of experimentation and need for further long-term monitoring. Although in theory, smart meters should encourage optimal decision-making for consumers, the field experiment shows the value of testing assumptions. Only through experimentation did researchers learn that consumption patterns of Ontario households deviate from optimal electricity use when receiving real-time feedback. Further, although researchers found a reduction in electricity usage, they question the sustainability after five months. More longitudinal field experiments are necessary to understand the long-term effects of real-time feedback on energy use.

Digital consumer policy

Chapter 3 analyses the challenges of online consumer protection through a behavioural lens. In this digital age, companies can provide more timely and relevant content for online consumers and, at the same time, easily mislead or misinform when consumers purchase products or agree to terms on line. This chapter uses BI as an analytical tool and provides practical implications for consumer authorities in three areas: online advertising, disclosure agreements and personalised pricing.

The main contribution of this study is that it analyses each policy problem through a behavioural angle from the start. Through this approach, the study leverages relevant scientific relevant literature to identify potential biases that could apply to the context of the specific policy question under investigation. For online advertising and disclosure agreements, the chapter points to behavioural biases such as anchoring and defaults that put online consumers at risk.

Subsequently, the chapter provides practical next steps for policymakers. For online advertising, they recommend potential behavioural experiments to understand the most effective ways to protect consumers from online advertising that may leverage behavioural biases. For disclosure agreements, in addition to experimentation, the chapter provides a list of behaviourally-informed recommendations that prioritise simple, clear and timely ways to improve consumer understanding of online disclosures.

Finally, the chapter explores new territory by analysing the behavioural biases and potential implications of personalised pricing. Although there is still limited empirical evidence, there is growing interest in how online vendors tailor prices to individuals based on personal data. After exploring potential biases relevant to personalised pricing such as framing and overconfidence, the chapter poses two hypotheses to test these biases through e-commerce simulations for a future lab experiment. This behaviourally-driven angle on problem scoping and emphasis on experimentation is a model approach to understand how best to design consumer protection policies for the online market.

Cartel behaviour

Chapter 4 seeks to compare two alternative models of cartel deterrence through a BI lens. On the one hand, a number of antitrust regimes punish detected cartels through fines which are revenue-based and levied on the firm level (i.e. corporate fines). This model has been prevalent in many European countries and is reflected in the formulation of the antitrust regulations of the European Union. On the other hand, North American jurisdictions, and notably the United States system, punish cartels by means of salary-based fines that function at the level of companies’ managers (i.e. individual fines). Comparing the two regimes helps examine whether and how manager incentives differ under competing regulatory frameworks. In addition, the comparison sheds light on how different antitrust regimes impact the labour market for managers, in terms of the type and level of contracts offered by firm owners and shareholders.

In particular, the study addresses the two questions of: i) whether antitrust jurisdictions involving individual-level fines have a stronger deterrence effect than the antitrust jurisdictions based on corporate-level fines; and ii) whether different regulatory frameworks generate distortions in the labour market for managers, by making certain categories of contracts more appealing to shareholders (e.g. contracts with fixed salary versus fixed plus a variable component).

The chapter relies on a unique combination of both theoretical and experimental approaches. Interestingly, the experimental findings deviate in significant ways from the theoretical predictions. While the theoretical model predicts that the type of fines on collusion should impact managers’ decisions on whether to form a cartel, in observed experimental behaviour, there is no statistical difference in the frequency (or duration) of cartels between the EU and US antitrust frameworks. Experimental results not only depart from the theoretical model with regard to cartel prevalence but also with regard to pricing behaviour and contractual choices by shareholders.

One of the key contributions of the study is precisely to indicate a set of potential behavioural explanations for why we observe a discrepancy between the way subjects actually behave in the lab and the “as if” theoretical model, in which actors are assumed to have somewhat homogeneous preferences. Conversely, the heterogeneity in manager-specific parameters, such as risk aversion and strategies chosen, can help understand the theory-evidence gap, especially in a context in which the decision of even one single company’s manager can prevent the formation of a cartel. While these results leave scope for further discussion, they markedly signal one of the key lessons of this report, which is the value-added of incorporating BI as an ex ante rather than ex post tool.

Safety behaviour

Chapter 5 provides experimental evidence on the potential application of different behavioural principles (social norms, messenger effect and feedback/benchmarking) to strengthen different dimensions of safety culture. The study involved the participation of both regulators and regulated entities in high-risk industries and explored their perception regarding awareness of safety culture in their field as well as potential responsiveness of different organisational actors to the selected behavioural principles.

The research introduced a new paradigm in the study of safety culture by leveraging a set of behavioural scenarios and vignettes. The vignettes simulated the application of BI to common safety topics, such as the introduction of a new safety regulation and reports on lost-time injury rates. The rationale behind the selection of these scenarios was that they would be generalisable across the four participating countries – Canada, Ireland, Mexico and Oman – and across their sectors, which ranged from oil and gas to electricity regulation.

The findings indicated interesting asymmetries between how regulators and regulated entities perceived safety culture, which suggests that it is essential for regulators to take into account different views around safety when designing new policies. Equally, the results displayed asymmetries between managers and frontline workers, with the latter reporting a more negative perception of safety culture (Tear et al., forthcoming; Parand et al., 2010). This result is possibly due to frontline workers’ more tangible contact with the safety risks in the organisation and it corroborates the importance of factoring in perception and information differences to avoid unintended policy effects and to ensure policies are targeted for different audiences.

With regard to the relative effectiveness of different behavioural principles, the study suggests that feedback has a more powerful effect on safety culture than social norms and messenger effects. Interestingly, in the experiment, this is true regardless of the modality through which feedback is provided. Future policies can incorporate this finding by ensuring that workers are provided with some form of data-driven performance feedback that is aimed at correcting and preventing unsafe practices. In particular, given asymmetries between workers and managers, it would be useful to ensure that the feedback is empirical/data-driven rather than perception-based. While messenger effects and social norming were overall less effective than feedback, they still indicated a number of behavioural differences, notably when looking at the data through an international lens. In particular, social norms seemed to have their most powerful effects on the Mexican sample, while messenger effects seemed to mostly only matter in Ireland and Oman. These international differences are further discussed in the relative chapter and represent one of the main value-added of the study.

Overall, the research demonstrates that not all behavioural principles are equally effective in motivating safe behaviour and that different nationalities can report different perceptions of safety culture as well as different levels of responsiveness to behavioural principles. The guidance provided by the research is, therefore, a promising tool to further explore the framework of the multiple “unknown knowns” and “known unknowns” which characterises safety risks.