Framed field experiments implement interventions or treatments in in a real-world setting, aiming to observe their effects on specific outcomes (Harrison and List, 2004[27]). This research design is particularly valuable when studying complex phenomena, difficult to isolate and manipulate in a laboratory setting, such as social norms, attitudes and decision-making. Moreover, through placing an experiment in a real-world setting, interventions or treatments are likely to be more relevant and meaningful for participants.
In the area of public integrity, framed field experiments have been used, amongst others, to study the effects of monitoring and punishment in corruption (Armantier and Boly, 2011[28]), the use of information on voters and the effects of vote-buying on clientelism (Serra and Wantchekon, 2012[29]), and the effects of increasing wages in bribe taking scenarios (Armantier and Boly, 2013[30]).
In this report, four interventions based in behavioural insights have been proposed:
Redesign the risk registers to include intermediate indicators for intervention measures.
Design a user guide for the adoption of corruption risk methodology in Decision 599/2018.
Develop a web-based application to guide the corruption risks management.
Create a dedicated unit or person within each ministry to assist working groups in the management of corruption risk.
The first two strategies are simple, non-expensive and could be tested in a field experiment. The following paragraphs detail the main characteristic of such a potential experiment testing a user’s guide:
Strategy design: A user’s guide for the corruption risk management could be developed following Chapter 2 recommendations.
Experimental task design: The experiment could measure participants’ capacity to identify corruption risks, design a control measure and register it in a corruption risk registry. A case study of a corruption risk would need to be drafted to provide the context for the task. This case could be associated with activities outside the central authority's primary mission and that are relevant to (almost) all organisations. Examples of such transversally relevant risks could include corruption in the human resources selection process or in the public procurement of office supplies.
Romanian anti-corruption authorities could design this case study of a corruption risk and then fill the corruption risk registry, including the risk identification, likelihood and impact assessment and an intervention measure to control the risk. Public officials’ responses will be graded on their similarity to the corruption registry prepared by the anti-corruption authorities following a qualification grid that would need to be developed to guide the grading of the responses.
Base line collection and pilot: To test the logistics of conducting an experiment, a random sample of 20 public officials from the central level could be recruited to participate in a 2 to 4 hour-long in-person session. Participants could be selected from a “convenient” sample of public officials. Participants would be divided randomly into two groups. In one group, participants would be asked to read the Decision 599/2018 and the corruption risk study case. In the other group, participants would be asked to read the User Guide and the corruption risk case. After reading both texts, they will be asked to complete a corruption risk registry (Decision 599/2018, Annex 4). The demographics of the participants should be selected to be as close as possible to those of public officials appointed in the working groups. The results from this baseline will be used to assess the test score sample average. At the end of this pilot, participants from both groups could provide feedback on the experimental task and the User Guide through interviews or a focus group.
Experimental design: Two groups of public officials from a sample population would be randomly assigned to a control group or an experimental group. The size of the sample needs to be determined by a power calculation. In the control group, participants will read Decision 599/2018 document and its annexes, the corruption risk case study and be asked to fill the corruption risk registry. In the experimental group, they will go through the same process, but instead of reading Decision 599/2018, they will read the user’s guide. Both sessions will be identical except for the use of Decision 599/2018 or the user’s guide.
Expected impact: The experiment would test if the new user’s guide for the corruption risk methodology increases participants' scores in the experimental tasks. A higher grade would mean that the participant is more able to identify risk and design a thorough control measure.