Corruption, money laundering, and other types of financial crime remain significant problems in the European Union. Even the bloc’s government institutions themselves are not immune to the problem, especially in areas like contract procurement.
But tech can provide answers to these kinds of foul play, according to experts from the Government Transparency Institute (GTI) – a non-partisan think tank dedicated to good governance.
Public procurement and public grand programmes constitute a substantial portion of GDP in both high- and low-income economies. Across the globe, public procurement represents 15% to 30% of the GDP of many countries. This huge volume of public spending plays a crucial role in economic and social progress if allocated efficiently. However, it is also one of the government activities most vulnerable to fraud and corruption. The process is particularly vulnerable to malpractice because that is the point in the investment cycle where public money changes hands. According to some estimates, around 10% to 25% of a public contract’s overall value may be lost to corrupt practices Furthermore, fraud in public grant programs diverts taxpayers’ money away from essential services and reduces benefits for well-meaning recipients. When individual beneficiaries, private providers, or government officials corrupt grant programs, they not only undermine the integrity of the program itself, but they also risk eroding trust in government.
In the wake of the COVID-19 pandemic, marked by a high volume of accelerated spending, fraud risks have become an even more pressing concern for governments worldwide. In this environment, public control and audit bodies play a vital role to ensure money is well spent and vulnerabilities are spotted and addressed quickly.
Fortunately, there is a growing number of analytical tools to measure corruption in procurement. This is what Government Transparency Institute’s (GTI) work is all about. GTI is a non-partisan think tank researching and advocating good governance. Born from the research and civil society activism of its founder Mihály Fazekas, the Institute was founded in Budapest, Hungary in 2015 to provide an independent, research-driven voice to the causes of transparency, anti-corruption, and good governance in Europe and beyond. The aim of the Institute is to better understand the causes, characteristics, and consequences of low-quality governance with interdisciplinary analysis, drawing on political science, economics, law, and data science. We believe that the combination of a thorough qualitative understanding and precise quantitative measurement of the state is the foundation of good governance.
Public procurement corruption can be measured through risk indicators, so-called “red flags” (i.e., single bidding), which can be used to detect institutionalised forms of corruption. Advanced methods such as supervised machine learning can identify and validate these proxy indicators of corruption. One of our recently published studies uses machine learning methods to trace the organisation of corruption in public procurement, by theoretically and empirically assessing the contribution of extra-legal governance organisations (EGO) to supporting it.
The goal of the empirical analysis was to identify the best explanatory model for mafia-like EGO presence within municipal public procurement contracts. The dependent variable of the analysis was therefore EGO presence on the contract-level which results in the extra-legal provision of governance services in public procurement, including, as an extreme case, mafias in public procurement. We compiled Italian administrative data on public tenders and dissolved municipalities to assemble our original dataset. This training dataset contains procurement contracts awarded by Italian municipalities that are proven to be infiltrated by the mafia. This empirical set-up uniquely gave us a set of proven positive and negative cases to train our models on. Then we expanded our dataset to all municipal contracts in Italy, while finally we extrapolated using data on all European municipal contacts taking data from the EU’s central register, Tenders Electronic Daily.
The predictive models included both traditional regression analysis and machine learning models such as binary logistic regression, random forest, and gradient boosting classifier. Drawing on traditional regression methods as well as tree-based machine learning algorithms, we developed high precision predictive models with which we were able to identify and validate proxy indicators for EGO presence in public procurement, such as single bidding or municipal spending concentration. Internal validity of our models is very high, 85% of unseen contracts are correctly classified. External validity is moderate, our predicted EGO presence score correlates with established indicators of organised criminality across the whole of Italy and Europe with a linear correlation coefficient of about 0.4.
The below example illustrates that our institution strives to develop tools that are relevant for the 21st century and accessible to everyone. We have developed several handy tools for users to search and analyse public procurement data:
1. ProAct (Procurement Anti-corruption and Transparency) Platform
The platform has been developed through a collaboration between the World Bank and the Government Transparency Institute. The methodology for the platform is based in part on the methodology developed for the Opentender project (Opentender, developed by the GTI part of the DIGIWHIST project, is a website where you can analyse several countries’ public procurement data to see how governments are buying goods, works and services).
The tool provides access to open data from national electronic procurement systems from 46 countries and to open data on World Bank and IDB financed contracts for over 100 countries. Most of the available data cover the period from 2006 to early 2020 (in some cases to 2021), but some sources date back to 1961. The prototype enables the analysis of over 21 million contracts, more than 5 million suppliers and almost 1 million buyers across 120 countries, representing a total estimated 2% of annual world GDP.
The ProAct platform has three main goals:
- Provide easy access to public procurement data for national and cross-country analysis.
- Enable users to monitor and analyse the performance of public procurement systems, their transparency and to identify the presence of possible integrity risks to inform preventive actions, improve procedures, and design policies on the basis of evidence.
- Help promote transparency and integrity in public expenditures on goods, works and services.
Tender-X is a for-profit venture developed by the Government Transparency Institute for commercial use. The target clients of Tender-X are banks and development finance institutions that finance companies engaged in public procurement, as well as private companies and their legal and financial advisors who compete for public contracts.
Tender-X aims to save time for its clients by providing an analytical report based on quantitative data. Such data are not part of the usual qualitative and watchlist-based due diligence processes performed by in-house teams or external consultants, and can take up to several weeks. The comprehensive organisational reports of Tender-X can cover both the companies that participate in tenders and the public institutions that organise them. Tender-X has compiled one of the most comprehensive contract level procurement datasets in the world, containing well-structured data on more than 50 million contracts from 4 continents, including a wide array of risk indicators.
Tender-X integrity risk reports aim to complement compliance assessments with analysis of contracting patterns equally relevant for commercial, legal and compliance functions. The Tender-X website includes a user-friendly search functionality to select companies and organisations of interest and order reports immediately.
Tender-X is also open to deliver on-demand risk analytics tailored to client needs, making use of the wide-ranging functionalities of its dataset and drawing on data and insights from interviews and fieldwork.
