GAME-THEORETICAL LEARNINGS FOR AUTHORITIES TO DISCOURAGE CHEATING IN PUBLIC TENDERS
French transport company, Keolis, won a public tender in 2019 for the bus transportation rights in the Eastern provinces of the Netherlands. The contract, valued at 900 million Euros, was revoked in the summer of 2020 as reports showed that Keolis had signed fraudulent agreements with their subcontractors. With the contract already due in December, the government had to award Keolis a two-year emergency concession to limit welfare losses. This article will demonstrate, through a game-theoretical analysis, how this tender rewarded fraudulent behaviour at the expense of society, potentially setting a bad precedent. We will also explain how the government should redesign future tender processes, to incentivise ethical behaviour and discourage fraud.
In September 2019, the province of Overijssel (Netherlands) issued a European tender procedure for public bus transport services for large parts of the provinces of Overijssel, Gelderland and Flevoland. The ten-year contract gave the winner the exclusive right to transport 17 million travellers on an annual basis. Part of the deal was that the competing suppliers would replace the existing bus fleet with new zero-emission vehicles. In addition, suppliers would earn points if, for instance, they could guarantee the on-time delivery of the new buses. Such a point-based scoring method is often used to objectively compare suppliers on a total business case (an article on why this approach in general is sub-optimal can be found here). Earning more points translates into a competitive advantage during the awarding process (especially in a situation where other elements are expected to result in similar scorings across competitors).
French transport company Keolis secured the concession at a value of roughly 900 million Euros. Their winning move seemed to be based on the deals they had made with Chinese production company BYD and, at the time, Chinese-owned Dutch company Ginaf, to build their buses. These deals were heavily criticised by both Dutch production company VDL and local politicians. They claimed that by placing the order outside of the Netherlands, Keolis would severely damage employment rates in the Dutch production industry. However, the bomb did not explode until May 2020, when Keolis itself reported to the Dutch public prosecution office that they made possibly fraudulent agreements with their subcontractors that were deliberately kept out of the official tender documentation. BYD and Ginaf refused to give a legal guarantee for the on-time delivery, so Keolis signed the so-called ‘side letters’ that stated the guarantees had no legal status, no rights can be derived of them, and that they were only in place to earn the maximum number of points.
The contract was withdrawn by the province of Overijssel in August 2020, but it was originally scheduled to take effect in December. As none of the competitors were able to replace Keolis on such short notice, and the government did not want to engage in capital destruction, Keolis was allowed to provide an ‘emergency concession’ for the upcoming two years. With the ‘emergency concession’ starting on December 13th, Ginaf ironically failed to meet the delivery deadline for the intended 87 community vans. In the meantime, the government are to align the new tender procedure for the remaining eight years, but the big question is whether Keolis will be allowed to participate.
Foul play delivers a higher payoff than fair play
As game theory is all about the analysis of strategic interactions, it would make a lot of sense for the Overijssel province to look for some game-theoretical insights to determine what course of action is best to pursue with respect to Keolis’ participation in the next tender. We are happy to support them by means of this article, as we are all in favour of optimising the usage of public funds and resources.
To explain why the current situation is a problematic one from a government perspective, we will first illustrate how the chain of events in this specific case rewarded cheating. In addition, we will also explain how it has potentially created an undesirable incentive for suppliers to commit fraud in other large public tenders, especially if these suppliers take learnings from this case. Obviously, this does not mean that companies will readily respond to fraudulent incentives, as many organisations will be keen to conduct business in an ethical manner. However, the Keolis case proves that, given the right incentives, at some point people and organisations are bound to act on them.
Before we dive into the details that have corrupted the potential for a fair follow-up tender process, we will draw a simplified game-theoretical outline of actions and their payoffs.
As a starting point, we will consider a tender process in which Keolis has two actions to play: fair play or foul play (by signing fraudulent side letters). If we consider symmetric suppliers who all bid truthfully, Keolis has an equal chance of winning the ten-year contract. This chance decreases whenever competition increases. However, the option to commit fraud will give them a significantly higher payoff, as it drastically increases their chance of winning the tender.
The crux lies in the fact that foul play and getting caught does not result in a payoff that is worse than fair play. Even though there has been no dispute that Keolis cheated in the process, their only punishment has been to reduce their ten-year profit to a two-year profit. Even if we assume that Keolis will always be caught, they will only play fair if the expected payoff of potentially winning a ten-year profit exceeds a nearly certain two-year profit. As Keolis is also the incumbent, losing the concession even leads to a negative payoff as it also means losing a quarter of their annual revenue. If we add that consideration to the equation, a significantly increased probability of a two-year profit is the optimal choice in almost all competitive settings.
Mathematically speaking, the inequality we are describing here is as follows:
E(Vfair) < E(Vfraud),
where E(Vfair) = pfair * π10-year – (1 – pfair) * closs
and E(Vfraud) = pfraud * π2-year – (1 – pfraud) * closs
where E(V) is the expected payoff, p is the probability of winning the tender, π is the profit, and closs is the cost of losing the tender. In the calculation example above we assume all values to be non-negative, pfraud to be ~100%, and pfair to decrease with increasing competition as explained in the text.
The expected value of committing fraud quickly exceeds the expected value of playing fair when competition increases, as not only the probability of gains decreases, but the probability of incurring costs associated with losing a significant share of revenue increases simultaneously.
Such an incentive misalignment problem is usually solved through fines or lawsuits. However, we have seen on multiple occasions that these punishments become part of a risk assessment and paying a fine turns out to be the better business case. Simply letting Keolis pay a fine would therefore only be a temporary and ineffective punishment, unless it can be a large enough deterrent. Strictly speaking a fine is not even possible, as the tender documentation does not provide any clause related to fines. The only remaining punishment here seems to be to exclude Keolis from the next tender.
In part 2 of this article series we will analyse the dynamics of a potential next tender process and how this might also have some undesirable outcomes.