By Claire A. Dunlop, Professor of Politics and Public Policy at University of Exeter, UK and Claudio M. Radaelli, Professor of Public Policy at University College London, UK

What is the exact causal relationship between corruption in the public sector and regulation? Hundreds of studies have scrutinized this relationship. We end up with not just one, but three causal narratives: that regulation causes corruption but under certain conditions; that it is the quality of regulation to hinder corruption; and that anti-corruption regulation can aggravate the problem of corruption.

The first narrative is by far the most popular. It is corroborated by studies carried out mostly by economists – regulation of private market activities may not only be inefficient, but push companies and small business entities to pay bribes to avoid either compliance or administrative costs – or simply to get a permit that depends on the discretion of public authorities. Does it follow that de-regulation is always a good idea to curb corruption? It depends: for a start, we have an efficiency loss if we scrap regulation that generates net social benefits. Then in some cases even what apparently looks like the most benign form of de-regulation, such as de-regulating business starts-up, can facilitate corruption. This is the case when de-regulation facilitates the process of rent-extraction by ruling elites. It also depends on whether we are looking at small-scale corruption in rule-making or grand-scale regulation-induced corruption such as nationwide privatization plans or the attribution of licences to broadcast television.

Here the second narrative has a role to play. It is not regulation in general that causes corruption, but those regulations that are particularly bad. Bad for what, in what ways? Bad because of irritating burdens, the presence of discriminatory consequences between types of firms and economic operators, and also bad for governance – consultation opportunities may be in practice restricted to the usual suspects and inspection plans may de facto encourage corruption. This is where we encounter an important lesson: to be effective, the use of regulatory policy to hinder corruption has to be targeted. Instead of bonfires of regulation, we need to rely on policy instruments that allow sifting through regulation in granular ways. This explains the success of policy instruments such as regulatory impact assessment.

And yet – the third narrative claims – shouldn’t we tackle corruption by drawing on regulation directly, that is, with specific anti-corruption regulatory architectures? In the European Union public anxieties about corruption in Eastern and Southern Europe are rife. These anxieties are particularly acute when it comes from European Union funds for development and cohesion. In Brussels, the European Commission may think they are funding projects, instead they are fuelling corruption. To avoid this, the EU’s development and cohesion policies have kind of mutated into a problem of fighting corruption and protecting the monies of the European Union. So, we have a complete re-categorization of EU policy – from development policy to anti-corruption. But, the story goes on, the heavy regulatory framework raises the cost of applying for EU funding – small firms are excluded and only the best organised firms can access EU funding for projects. Other firms hire specialist consultants that translate the language of EU formalities and regulations into the language that economic operators understand. The whole system becomes heavily regulated and makes non-compliance harder to detect because the bar is raised: more specialism, more ‘entrepreneurs’ that allow local actors to fund their activities want even if these activities do not really produce ‘development’ and ‘cohesion’.

Where do these narratives leave us when it comes to designing solutions? An original way to re-cast the debate is to focus on the design of the instrumentation that governs the production and usage of rules. In a sense this is meta-regulation, that is the procedural rules (found in administrative law) that apply to the life cycle of regulation. Regulations do not come from nowhere. In the OECD experience, governments have adopted administrative procedures through which proposed regulations are appraised (via impact assessment), then designed (via drafting rules), then again notified to stakeholders for comments (via consultation in Europe, or notice and comment in the USA). These procedures are complemented by other procedures that allow regulations to enter into force, be delivered, accessed (via Freedom of Information Acts) and re-calibrated through judicial review – and in many countries via the Ombudsman. There is a whole-of-government meta-regulatory design concerning the life-cycle of regulation. By measuring the characteristics of the meta-regulatory design across countries we can infer whether the opportunities for corruption during rule-making are increased or reduced. This implies the existence of a set of coherent measures on impact assessment, consultation, judicial review, freedom of information and the Ombusdman. For this reason we are collecting a new data set where we follow the same template to gather information across the EU-28 on how impact assessment, consultation, freedom of information, judicial review and the Ombusdman procedures work in terms of openness, access, transparency and scope. This novel data set will allow to empirically identify country by country the accountability relationship that the meta-regulatory design generates, and to associate with the outcome of interest, in our case corruption.

Conceptually, the research question is whether the design that appears once we have measured the five procedures mentioned above triggers social mechanisms that effectively make the bureaucracy accountable to the general public – in some countries the design may well be flawed if it tilts the scale towards special interests. Our work in the project Protego (Procedural Tools for Effective Governance), carried out with Professor Alessia Damonte at the University of Milan, suggests that, first, the procedures work together, hence we should model the causal effect of the whole design on corruption as outcome. Second, this causal effect works through mechanisms that make the bureaucracy accountable to a narrow or wide set of actors. Third, it is the accountability relationship at the stage of rulemaking that has a special place in the production or mitigation of public sector corruption. In fact, it is when a rule is formulated or delivered that bribes are paid and corrupt exchanges take place. By working together, the instrumentation of meta-regulation can either make a special interest in control of rulemaking (this can also be the public managers themselves) or make the bureaucracy respond to a plurality of interests. We can see now that the issue is no longer one of constraining bureaucracy or giving it discretion. Neither is it an issue of the political principals being able to force on the bureaucracy their preferences. It is an issue of whether rulemaking generates accountability to a pluralist constellation of interests. If this is correct, we would expect cross-national variation on corruption depending on the combination of policy instruments affecting rulemaking. This brings us to our to final question: to examine this variation, how what types of data are available? A vast literature has raised pretty fundamental doubts on the validity of the most common cross-country indicators like Transparency International’s CPI – Corruption Perception Index. Unfortunately, this is one of the indicators most often used in cross-country research.

Are we condemned to use datasets we have little confidence in – only because they are the only ones that allow for cross-country analysis? Others have looked instead for alternatives, especially objective measures. By their very nature, objective measures do not have the bias of perceptions-based indicators. To illustrate, we can use the gap in physical infrastructure (a given level of infrastructure should exist given the level of capital outlay, it is not there it must be because of corruption), the number of public managers involved in cross-border corruption cases, or single bidding in competitive markets as proxy of favouritism in public procurement. Further, it should not be taken for granted that cross-country empirical analyses should be carried out at the national level. The regional level can provide valuable lessons.

Although the debate on measuring corruption is lively, there is no convergence. To establish the exact meaning of corruption in a given social setting and achieve strong construct validity, one has to go in the field and look at how communities socially construct corruption. However, practice and meaning-tracing has obvious limits when one is interested in cross-national research. Then one needs data that are available in time-series and for many countries. For us, the most coherent respond is to generate new data that respond both to the research questions about corruption and regulation we want to address, and reflect faithfully the understanding of causality embodied in the causal narrative we want to test.