Spanish regulation prevents public administrations from splitting up large contracts into minor contracts with the intention to avoid competitive tendering and thus award directly to a single company. We exposed that, in the first seven months of 2019 alone, more than 6,500 public contracts failed to comply with this regulation. That is more than 5% of minor contracts in national, regional and local administrations published on the Platform for Public Procurement, amounting to more than €53 million.
This investigation registered more than 30m media impacts and follow ups by other newsrooms, mostly local-focused. As a result, it significantly impacted the local politics in some Spanish municipalities, such as the Ourense County Council (Galicia), Ames (Asturias), and others.
We identified Orense County Council as Spain’s public authority with the highest number of suspicious contracts to its credit, some 55 contracts. This finding became a major political issue for months after the opposition parties demanded an inquiry. A process that was eventually blocked by the president of this regional council.
On a national level, political party Ciudadanos registered in Congress a battery of questions to ask the Government to estimate the cost for public administrations of “malpractices and possible illegalities” in public procurement, based on our findings.
Moreover, we shared our database and guide of the main issues detected to some twenty public procurement and anti-corruption control bodies at all levels national, regional and local. Our goal was to help them improve their control and even sanction those responsible, saving them precious time with our list of suspicious contracts.
We downloaded all minor contracts published on the Platform for Public Procurement from January 1, 2018, until July 31, 2019, the date when we started to analyze them in depth. A total of 346,726 contracts. First, in order to create the dataset, we scraped the national procurement portal, using Ruby to gather and parse a collection of XML and HTML files. The results were placed in a PostgreSQL database, where we ran queries to detect outliers, identify gaps in the data and extract samples to be reviewed manually. The main analysis of the investigation (i.e. identifying clusters of public bodies and companies evading legal procedures via contract splitting) was done in PostgreSQL using SQL. Relevant subsets of the data were extracted files to be cleaned semi-automatically by journalists, using Open Refine and Excel spreadsheets.
The visual design of the visualizations was done using Adobe Illustrator and parts exported as SVG to be used in the visualizations. Illustrator was also used to create a number of additional infographics that complement the articles.
What was the hardest part of this project?
The hardest part was creating a database to detect possible suspicious ‘fragmentation’ of contracts. The first step was the hardest: the team cleaned the data, amended mistakes (duplicated contracts, contracts mistakenly classified as minor) and reclassified those that were under the wrong category (goods, services and works contracts). Following this, in a second phase of analysis, we looked into contracts tendered to the same company by the same administration, which when cumulated together exceeded the legal threshold allowed for minor contracts. Later, we checked those contracts that met the previous requirements and were tendered on the same day. Additionally, we also reviewed those contracts under the same issue, tendered the same year, that on accumulation exceeded the legal threshold. It must be clarified that not all of the contracts reviewed might be illegal, but they are all suspicious and have to be analysed to see if they were chopped up with the intention of evading the law.
Civio conducted an in depth research on the topic, studying the law and consulting existing independent control agencies like OIReScon and Agencia Antifrau. We have also consulted Civio community members, experts on the topic, and other specialists.
What can others learn from this project?
First of all, the journalist can learn from this project how specific risks, red flags, may indicate a potential non-compliance with public procurement regulations. We focused on one or two of them -splitting public contracts up to avoid competitive tendering-, but there are many more.
When working with a huge procurement dataset containing tens of thousands of contracts, looking at each of them’s fine print is not an option. However, journalists can use red flags to surface those tenders with suspicious features and then study those in further detail, as we did. A red flag does not mean a contract is corrupt or illegal, just a hint that a closer look may reveal irregularities. And the more red flags, the stronger the suspicion.
Unfortunately, sometimes the necessary information to detect these malpractices is not publicly available or not with high-enough quality, a prerequisite for automating the red flag analysis. To answer the most complex questions, journalists would need to go beyond the procurement datasets and cross-reference them with other sources, such as the company register -an essential one- or a list of donors to political parties. If enough information is available, we will monitor more complex scenarios. It is just a matter of having extra datasets and putting some extra effort