Introduction

Behavioural science has become an increasingly important tool for improving public policy outcomes, including in health, education and taxation. A large body of research in psychology and behavioural economics demonstrates that individual decision-making is often shaped by heuristics, biases and contextual cues rather than by fully rational calculation (Kahneman, Reference Kahneman2011). Building on these insights, persuasion research has identified a set of influence principles – such as reciprocity, commitment and consistency, social proof, authority and scarcity – that systematically affect compliance and agreement when applied ethically (Cialdini, Reference Cialdini2009). In parallel, nudge theory has shown that small changes in choice architecture or message framing can produce substantial behavioural effects without restricting choice (Thaler and Sunstein, Reference Thaler and Sunstein2008). Together, these approaches offer a powerful toolkit for influencing behaviour in socially desirable directions.

In the domain of public finance, behavioural insights have increasingly been applied to tax compliance and debt collection, complementing a long-standing literature on deterrence and enforcement. As emphasized in Slemrod’s (Reference Slemrod2019) comprehensive overview of tax compliance and enforcement, traditional economic models – most notably the Becker–Allingham–Sandmo framework – treat tax evasion as a risky choice in which detection probabilities and penalties play a central deterrent role. Empirical evidence strongly supports this view, particularly through the effectiveness of third-party information reporting and withholding systems. At the same time, Slemrod (Reference Slemrod2019) underscores that compliance is shaped not only by formal enforcement parameters but also by taxpayers’ perceptions, beliefs and interactions with tax authorities. This recognition has motivated a growing body of research examining how communication, information and behavioural signals affect compliance decisions.

Within this broader enforcement context, behaviourally informed interventions – typically low-cost and communication-based – have emerged as a promising complement to audits and sanctions. Revenue authorities worldwide face persistent challenges in reducing arrears and encouraging timely payment, often relying on resource-intensive enforcement mechanisms with diminishing marginal returns. Behavioural interventions have therefore been deployed to influence compliance through messages that leverage social norms, salience and perceived monitoring (Behavioural Insights Team, 2012; Kettle et al., Reference Kettle, Hernandez, Sanders, Hauser and Shafir2016). Evidence from high-income countries shows that carefully designed messages can substantially increase compliance. In the United Kingdom, for example, social norm messages embedded in tax reminders significantly increased timely payment, generating additional revenue of £1.9 million within weeks (Hallsworth et al., Reference Hallsworth, List, Metcalfe and Vlaev2017). Similar effects have been documented in Belgium through message simplification and deterrence framing (De Neve et al., Reference De Neve, Imbert, Spinnewijn, Tsankova and Luts2019).

Related evidence from middle- and lower-income settings highlights both the potential and the constraints of behavioural approaches. Studies from Eastern Europe show that behaviourally informed communications improved compliance in Latvia and Poland (Hernandez et al., Reference Hernandez, Jamison, Korczyc, Mazar and Sormani2017; Jamison et al., Reference Jamison, Mazar and Sen2021), while evidence from Bulgaria indicates that audit probability messages and reciprocity-based framing can influence compliance and agreement on delayed payments (Dimitrov and Vlaev, Reference Dimitrov and Vlaev2022; Doerrenberg et al., Reference Doerrenberg, Pfrang and Schmitz2022). More broadly, research on value-added tax systems in developing countries emphasizes the importance of enforcement capacity, information trails and institutional design in shaping compliance outcomes (Gerard and Naritomi, Reference Gerard and Naritomi2018). These studies show that while behavioural incentives can reinforce compliance, their effectiveness depends critically on how enforcement is implemented and communicated at the operational level.

Despite the rapid growth of behavioural interventions in tax compliance and debt collection, a significant gap remains in our understanding of how behavioural insights operate in direct, live, person-to-person communication. Most behavioural approaches focus on written or digital channels – letters, emails or text messages – implicitly treating compliance as the outcome of centrally designed messages delivered to passive recipients. Telephone-based interactions, by contrast, remain comparatively understudied, despite the fact that call centres constitute a primary interface between citizens and revenue authorities in many tax systems. This omission is particularly consequential because frontline staff do not merely transmit information – through their behaviour, communication style and enforcement approach, they actively shape how taxpayers perceive authority, fairness and compliance expectations.

A growing body of evidence suggests that tax officials’ behaviour plays a crucial role in shaping taxpayer compliance through multiple channels, including service quality, interaction style and perceived enforcement power. Research consistently shows that clear, competent and supportive assistance from tax officers reduces confusion and compliance costs, thereby increasing both filing and truthful reporting (Alm et al., Reference Alm, Cherry, Jones and McKee2010; Yusril and Awaluddin, Reference Yusril and Awaluddin2015; Wahyuni and Setiyawati, Reference Wahyuni and Setiyawati2023; Manap et al., Reference Manap, Mustangin, Sasmiyati, Edy and Zainuddin2024). Laboratory and field studies demonstrate that when tax administrations provide calculation assistance and personalized guidance, compliance improves (Alm et al., Reference Alm, Cherry, Jones and McKee2010), while survey-based evidence across individual, corporate and local taxpayers links higher compliance to officers’ clarity, responsiveness, friendliness and expertise (Yusril and Awaluddin, Reference Yusril and Awaluddin2015; Putra et al., Reference Putra, Loist, Butarbutar, Efendi and Sudirman2023; Manap et al., Reference Manap, Mustangin, Sasmiyati, Edy and Zainuddin2024; Robbi et al., Reference Robbi, Putra, Elyta and Mahdi2024; Pradisa and Ferdiansyah, Reference Pradisa and Ferdiansyah2025; Widyari and Supadmi, Reference Widyari and Supadmi2025). These findings underscore that compliance is not solely a function of enforcement intensity but is also shaped by how taxpayers experience their interactions with the tax authority.

Interaction style and respect appear to be particularly important. Evidence from Indonesia shows that respectful treatment by tax officials increases compliance, whereas authoritarian procedures reduce it, largely through their effects on trust (Prastiwi and Diamastuti, Reference Prastiwi and Diamastuti2023). Similarly, mixed-methods research from Rwanda indicates that facilitative, educational interactions enhance trust and perceived professionalism, while purely enforcement-focused encounters can undermine tax morale, especially among larger taxpayers (Mascagni et al., Reference Mascagni, Scarpini, Mukama, Santoro and Hakizimana2025). Qualitative studies further suggest that taxpayers expect proactive, fast and competent service, and that such interactions are viewed by both taxpayers and officials as central to voluntary compliance (Arisanti and Sujana, Reference Arisanti and Sujana2024). These findings emphasize that compliance emerges from the joint influence of enforcement power and trust – both of which are embodied in the behaviour of frontline tax employees.

Within this context, emerging evidence indicates that live, interpersonal communication may be particularly powerful. A field experiment with Colombia’s national tax agency found that replacing letters with scripted phone calls increased tax payments fivefold, yielding a 25 percentage-point improvement in compliance (Mogollon et al., Reference Mogollon, Ortega and Scartascini2021). Complementary qualitative and observational studies highlight the importance of conversational dynamics – such as rapport, emotional tone and the balance between authority and cooperation – in shaping repayment outcomes (Harrington, Reference Harrington2018; Liao et al., Reference Liao, Du, Yang and Huang2021). Harrington (Reference Harrington2018), for example, shows how debt collection agents in a UK utility company navigate tensions between institutional demands for enforcement and the need to maintain cooperative, face-saving interactions, tensions that can constrain effectiveness. Likewise, Liao et al. (Reference Liao, Du, Yang and Huang2021), analysing more than 3,000 recorded debt collection calls from a large consumer finance firm in East Asia, find that emotional cues such as reassurance or fear predict repayment behaviour, whereas excessively punitive messages provoke anger and reduce compliance.

Despite these insights, systematic evaluations of behaviourally informed training for public-sector tax authority employees – particularly at the national level – remain rare. This study addresses that gap by examining whether a behavioural training intervention can improve the effectiveness of phone-based tax debt collection in a public revenue agency. In collaboration with the Bulgarian National Revenue Agency (NRA), we designed and implemented a behaviourally informed intervention centred on a two-day soft-skills training programme for call centre agents. Drawing on principles from behavioural science, persuasion and negotiation theory, the training aimed to enhance agents’ ability to balance authority and cooperation, build rapport and secure payment commitments during outbound calls. The two-day training was attended by all 42 agents employed in the NRA call centre. To ensure full participation, the training was delivered in two separate groups held in two different weeks. This scheduling served two purposes: first, it allowed for smaller group sizes that facilitated greater interactivity and more in-depth discussion during the sessions; second, it ensured that agents who were on sick leave or had pre-planned absences could attend the session that best suited their availability. In addition to the 42 agents, three supervisors also participated in the training groups, which helped ensure consistent implementation and oversight of the newly introduced communication practices.

The empirical analysis relies on descriptive comparisons of monthly aggregated administrative data on debt-collection outcomes before and after the intervention. While the design does not permit definitive causal inference, it allows us to examine whether performance patterns following the training differ meaningfully from both prior periods and historical seasonal trends.

We find that, following the intervention, debt collection outcomes improved substantially across multiple indicators, including higher rates of full repayment, increased conversion among contacted debtors and a marked increase in total debt collected. These outcome measures follow the standard operational definitions of the Bulgarian NRA. These outcome measures follow the standard operational definitions of the Bulgarian NRA. Although call agents request that payment be made within seven days during the phone conversation, a full repayment is recorded when the outstanding amount is paid within this seven-day period following the initial call. In line with the NRA’s debt-collection process, each debtor case is typically associated with a single outbound call by a phone operator. If no payment or payment agreement is secured within approximately one week after the call, the case is escalated and transferred to a separate ‘special collection’ pathway (e.g., legal or court-based enforcement), which lies outside the scope of this intervention and the available data. At that point, both the debtor and the corresponding debt are removed from the call centre’s active calling lists, and there is generally no repeated calling of the same debtor within this collection stage. Notably, these improvements coincide with a period that historically exhibits a seasonal decline in collections, strengthening the plausibility of an intervention-related effect.

The study contributes to the literature in three main ways. First, it extends behavioural public policy research beyond written nudges to live, interpersonal communication, highlighting the importance of delivery channel in shaping behavioural outcomes. Second, it provides rare field evidence on the role of behavioural training for frontline public servants, showing how behavioural insights can be embedded in everyday administrative practice. Third, it offers policy-relevant evidence from a real-world institutional setting, contributing to the growing body of Findings from the Field that document how behavioural interventions perform under operational constraints.

The objective of the intervention was to equip call centre staff with practical skills grounded in persuasion and nudge theories, thereby improving the effectiveness of phone-based debt collection. We hypothesized that, relative to business-as-usual practice, calls conducted by behaviourally trained agents would be associated with higher repayment rates and greater amounts of debt recovered, reflecting the use of more effective influence and communication strategies.