The Metro do Porto swap
How scenario analysis could have revealed the risk of a catastrophic swap deal
Metro do Porto (MdP), a Portuguese public transport company, entered into a 14-year interest rate swap with Santander on January 11, 2007, to reduce its 4.76% per annum funding costs on a €89 million loan (see Hull 2023, chap. 24). Under the swap, Santander paid MdP’s funding cost, and in return MdP paid just 1.76% plus a spread — saving 3 percentage points per year for the first two years while the spread was zero. After that, the spread was recalculated each quarter using a path-dependent cumulative formula that penalized MdP whenever the three-month Euribor rate moved outside a 2%–6% corridor.
At the time, Euribor was about 4% — safely in the middle of the corridor. But interest rates fell sharply during the financial crisis of 2008–2009 and remained at historically low levels for over a decade. The spread ratcheted up relentlessly: by May 2011, accumulated losses had reached €217 million; by 2017, the effective interest rate had surpassed 100%, and potential losses exceeded €500 million on the original €89 million loan. The deal has been described as a contender for the worst trade of all time. The Portuguese State ultimately settled with Santander, accepting contracts with rates above 100% in exchange for €2.3 billion in financing at below-market rates — a settlement that still cost the public over €1 billion.
This page shows that a straightforward Monte Carlo simulation — using only information available at the time — would have revealed the catastrophic risk embedded in this deal.
The spread formula
For the first two years of the deal, the spread was zero and MdP’s effective funding rate was just 1.76%. From year 3 onward, the spread was updated each quarter as:
\[ \text{Spread}_t = \max\!\Big[0,\;\text{Spread}_{t-1} + 2\max(2\% - R_t,\, 0) + 2\max(R_t - 6\%,\, 0) - D_t\Big] \]
where \(R_t\) is the three-month Euribor rate and \(D_t\) (the “DigiCoupon”) equals 0.5% when \(R_t \in [2\%, 6\%]\) and zero otherwise. MdP’s effective funding rate each quarter is \(1.76\% + \text{Spread}_t\).
Note
Why this formula is dangerous
The spread has a critical asymmetry. When Euribor is inside the [2%, 6%] corridor, the spread decreases by at most 0.5 percentage points per quarter (and cannot go below zero). But when Euribor drops below 2% — say to 1% — the spread increases by \(2 \times (2\% - 1\%) = 2\) percentage points per quarter. The spread accumulates four times faster than it recovers. Even a temporary excursion outside the corridor can permanently ratchet up MdP’s costs.
Spread evolution for a constant rate
Choose a hypothetical constant Euribor rate and see how the spread evolves over the 48 quarters (years 3–14) when the formula is active.
Tip
How to experiment
Try setting Euribor to 1.5% to see how quickly the spread accumulates when rates drop below the 2% floor. Then set it to 3% to confirm the spread stays at zero when rates are inside the corridor. Finally, try 7% to see the effect of rates above the 6% ceiling.
Interest rates at the time of the deal
The chart below shows quarterly observations of 3-month Euribor from 1999 to 2021 — the rate that directly drives the swap formula. The shaded band marks the 2%–6% corridor inside which MdP’s spread would remain stable.
Even before the deal, 3-month Euribor had been barely above 2% throughout 2003–2005 and nearly reached 5% in late 2000. Quarterly changes were substantial, as the histogram below shows.
Quarterly change statistics
The 31 quarterly changes observed from 1999 Q1 to 2006 Q4 are the data a risk analyst would have used to calibrate a simulation in late 2006.
Starting from about 3.75% (the 3-month Euribor rate when the deal was signed), this level of quarterly volatility implies a meaningful probability of Euribor drifting below 2% within a few years.
What scenario analysis would have shown
Using the historical statistics as defaults, the simulation below generates many possible Euribor paths over the 14-year deal and computes MdP’s effective funding rate for each path. Adjust the mean and standard deviation of quarterly changes, the starting Euribor rate, and the number of trials to explore different assumptions.
The histogram shows the distribution of MdP’s average effective funding rate over the 14-year deal across all simulated trials. The vertical dashed line at 4.76% marks the break-even point: to the right, MdP is worse off than if it had never entered the deal.
Simulated Euribor paths over the 14-year deal. The green band marks the 2%–6% corridor. When a path drops below 2%, MdP’s spread begins to accumulate.
The corresponding spread evolution for the simulated paths. Observe how even a single path dropping below 2% leads to rapid, persistent spread accumulation.
What actually happened
The charts below show the actual evolution of the spread computed from quarterly observations of 3-month Euribor over the deal period.
The outcome was catastrophic. Euribor dropped below the 2% floor within two years and stayed there — eventually going negative — for over a decade. The spread accumulated relentlessly, driving MdP’s effective rate to astronomical levels. The initial saving of 3 percentage points during years 1–2 was dwarfed by the subsequent losses.
Note
The enterprise risk management lesson
This case illustrates a core principle of enterprise risk management: potential adverse events must be identified and their consequences quantified before committing to a strategy. A simple Monte Carlo simulation using only the historical interest rate data available in 2006 would have revealed a substantial probability of catastrophic losses. The deal offered a modest guaranteed saving (3% per year for two years) in exchange for an open-ended, path-dependent tail risk. No risk appetite framework that valued the institution’s financial stability would have deemed this an acceptable trade-off.
Notably, when Santander presented the swap proposal, its historical analysis for this product only showed data from 1999 onward — omitting decades of earlier interest rate history that would have revealed rates frequently outside the 2%–6% corridor. A thorough due diligence process would have demanded a longer historical perspective and run stress tests accordingly.
Simulating in R
The Monte Carlo simulation can also be performed in R. The code below runs directly in the browser — you can edit the parameters and re-run each block.
Load and prepare data
Read the 3-month Euribor data from the ECB and extract quarterly observations (March, June, September, December).
Pre-deal statistics
Compute the mean and standard deviation of quarterly Euribor changes using data available before the deal (1999 Q1 to 2006 Q4).
Simulation parameters
Set up the simulation inputs. Edit these to explore different scenarios.
Run Monte Carlo simulation
Each trial generates a 14-year Euribor path using normal quarterly changes, then computes the spread and MdP’s effective rate. The spread formula activates from quarter 9 (year 3) onward.
Results
Summary statistics for MdP’s average effective funding rate over the 14-year deal.
Further reading
A história do pior swap de sempre. E de como ele podia ter sido evitado — ECO, 2017. Detailed account of the Metro do Porto swap, including the deal’s terms, how losses accumulated, and the failed restructuring attempts.
Estado aceita pagar swaps com juros acima de 100% no acordo com Santander — ECO, 2017. Coverage of the settlement between the Portuguese State and Santander, including contracts with interest rates exceeding 100%.
Dinheiro público continua a pagar swaps de empresas públicas de transportes: mais de 1,5 mil milhões em cinco anos — Expresso, 2020. Overview of the ongoing costs of swap contracts to Portuguese public transport companies, totalling over €1.5 billion in five years.
References
Hull, John. 2023. Risk Management and Financial Institutions. 6th ed. New Jersey: John Wiley & Sons.