Built, and Empty

What predicts a ghost airport? The question has been worked on for thirty years. Bent Flyvbjerg has been mining the cost and demand data on infrastructure megaprojects since the 1990s, anchored in his 2014 paper on the Iron Law of Megaprojects. James Robinson and Ragnar Torvik built the political-economy theory in a 2005 paper on white elephants. Don Pickrell, Martin Wachs and John Kain documented the systematic optimism in transport forecasting fifteen years before that. The American Planning Association in 2005, the OECD in 2016, the US National Academies in 2012, and the team of Suh and Ryerson in 2019 have all converged on the same correction: a new airport's most likely outcome is whatever the comparable airports actually delivered.

Six predictors recur across this literature. They are: whether the project's demand forecast can survive comparison with the actual record of comparable airports already built; whether the chosen site overrode an independent technical recommendation; whether the project's inefficiency is itself the political signal that gets it built; whether the financing leaves the state on the hook for the loss; whether momentum makes stopping more costly than continuing once construction is underway; and whether another airport already serves the same area. Each predictor has academic backing and a set of physical examples on the ground. This article measures the Palmar Sur proposal against each.

The Iron Law

The 2014 paper is the one a 2020 Surcos essay by María José Guillén and José Antonio Mora first brought into the Costa Rican airport debate. Its central finding, after three decades of infrastructure data, was what Flyvbjerg called the Iron Law of Megaprojects: "over budget, over time, over and over again." Nine out of ten megaprojects overrun their budget. Rail projects average a 44.7% cost overrun combined with a 51.4% demand shortfall. If one in ten megaprojects is on budget, one in ten on schedule, and one in ten on benefits, then roughly one in a thousand meets all three targets. Seventy years of data show no improvement.

Why does this keep happening? Flyvbjerg points to four "sublimes," the technological, political, economic and aesthetic appetites that form coalitions in favour of more megaprojects. Robinson and Torvik in 2005 added a sharper political-economy claim that this article returns to later: white-elephant projects are built because their inefficiency is what makes the political promise credible. Both bodies of work converge on the same observation. Projects get selected for how they look on paper, and the projects that look best on paper rarely perform best in operation.

The fix, named in April 2005 by the American Planning Association after Flyvbjerg, Holm and Buhl's JAPA paper of that year, is reference-class forecasting. The idea is plain: rather than trust the project sponsor's spreadsheet, ask what the comparable projects already built actually delivered. Use that as the forecast. Suh and Ryerson applied the method to airports specifically in their 2019 paper, and found that grounding airport forecasts in the actual records of similar airports improves accuracy. The OECD in 2016 and the US National Academies in 2012 reach the same conclusion. The fix is also the central question of the literature: what would a comparable airport actually deliver? The rest of this article applies that question to six predictors that recur in airport-specific cases, then measures Palmar Sur against each.

The forecast does not survive a reality check

The first predictor is whether the demand forecast can survive a comparison with what similar airports already built actually delivered. Airport demand forecasts have a poor track record. The most common error is assuming that traveler numbers will rise faster than the local economy supports. A forecast that ignores what similar airports actually did tends to overshoot reality by a wide margin. The OECD and the US National Academies have both built guidance documents around the same finding.

The forecasts that produced ghost airports were predictable failures of this kind before construction began. Sri Lanka's Mattala Rajapaksa was designed for one million passengers a year on a site with no nearby city or local travel demand to draw from, while the country's main airport at Bandaranaike already served the demand the new facility was meant to capture. The forecast required Sri Lankan air traffic to roughly double from existing levels and to migrate from the existing main airport to the new secondary one. Neither had any operational basis. The 2021 total was 25,767 passengers, less than three per cent of the design figure. Saint Helena's airport, with first scheduled commercial service inaugurated in October 2017 to a remote British overseas territory, was designed against a long-range projection of up to thirty thousand tourists a year. The island had never received anything close to that figure at any earlier travel price, and the airport's opening did not bring any new carrier or itinerary that would have plausibly multiplied it. The 2018 visitor count was 3,800.

Nicaragua's Punta Huete is the most blatant case. The China CAMC contract designed the airport for 3.5 million passengers a year. The total number of passengers disembarked anywhere in Nicaragua in 2022 was 634,807. The new airport is being built for roughly five times the country's entire current air-travel market. Mexico's AIFA, opened in 2022 on a twenty-million-passenger-a-year design, handled 6.3 million in 2024, about 3.4 per cent of Mexico's total air traffic. Canada's Mirabel, opened in 1975 on a seventeen-million-passenger projection, peaked under three million; passenger service ended in 2004 and the terminal was demolished between 2014 and 2016.

The OECD's case study of Lambert-St. Louis is documented in airport-planning textbooks. The FAA's official forecast projected 20 million passengers by 2006 and 25 million by 2012. Actual passenger volumes declined 56 per cent between 2000 and 2004 to 6.7 million, and traffic continued to fall thereafter. A third runway was built anyway. The OECD report's own commentary: "it was decided to continue development" despite uncertainty about the hub airline's future.

Site selection overrode the technical recommendation

The second predictor is whether the chosen site survived independent technical review. The infrastructure-forecasting literature is direct on this point. Martin Wachs in 1990 wrote that infrastructure forecasts often work as advocacy, not analysis. John Kain that same year traced the Dallas DART rail-transit forecasts to "extensive use of clearly unrealistic land-use forecasts and optimistic ridership forecasts" used to push the project through, with internal counter-analyses kept off the table. Flyvbjerg, in 2007, called "pervasive misinformation about the costs, benefits, and risks involved" the main problem in large-infrastructure decisions and put the fix in governance rather than in better spreadsheets. When the agencies that are supposed to vet a site advise against it and the project moves ahead anyway, the project has stopped pretending to be a technical exercise.

Sri Lanka's Strategic Enterprises Management Agency advised against the Mattala site before construction. The agency recommended expanding the existing Bandaranaike International instead. The government overrode SEMA. The result is in the operating numbers.

UNESCO opened a 2019 reactive monitoring procedure on Peru's Chinchero airport on the grounds that construction posed a heritage risk to the Sacred Valley and Machu Picchu. The Peruvian state continued construction. As of early 2024 the airport was 12 per cent complete; the Korean-led consortium had threatened to halt work over cost overruns and missing state disbursements; UNESCO opposition remained on the record.

Spain's Castellón–Costa Azahar airport sits between two existing international airports at Valencia and Barcelona, both of which already had spare capacity for the region's tourism. The province's own technical justification rested on tourism projections that the operating company Aerocas spent approximately €26 million sponsoring regional sports teams to try to generate. Castellón's principal political sponsor Carlos Fabra was sentenced to four years in prison for tax fraud in December 2014, several months after the technical case for a third international airport in the region had quietly stopped being defended.

Inefficiency is the political signal

Robinson and Torvik's 2005 paper in the Journal of Public Economics argues that white-elephant projects, the kind that cost the public more than they ever return, are built precisely because of their inefficiency. The argument turns on who pays and who benefits. The losses of a money-losing airport fall on the national treasury. The benefits stay in the home region: terminal jobs, construction contracts, maintenance contracts, ongoing political attention. Local voters want the airport to keep operating regardless of whether it ever turns a profit. A profitable airport would survive any administration. A money-losing airport survives only under a politician personally tied to keeping it open. Voting for that politician is voting for the airport to stay open, and for the local benefits that go with it. The inefficiency is what makes the patron's promise believable; no one else would pay the bill. "The politician ends up picking losers rather than winners," Robinson and Torvik write, and add: "this misallocation takes place even when its implications are understood. Thus it is not due to incompetence." The airports that are at their most economically marginal turn out to be the airports where the political incentives are at their strongest.

Spain's Castellón airport was officially declared "open" in March 2011, two months before regional elections, despite having no airline signed up to land there and no government approval to operate. The first commercial flight did not arrive until September 2015, four and a half years later. The opening served the political calendar; the operating reality lagged it by years.

Sri Lanka's Mattala Rajapaksa was named for, and built in the home province of, then-President Mahinda Rajapaksa. The site selection was a constituency targeting decision before it was anything else. The logic is direct: the airport's economic inefficiency is what makes its sponsor's commitment to the home province credible.

Mexico's AIFA, opened in 2022, was the political legacy project of the López Obrador administration; PAN senator Kenia López Rabadán has repeatedly labelled it an elefante blanco. Ecuador's Jumandy airport, opened in 2011 by Rafael Correa to serve Amazon tourism with typical flights operating with five passengers, was converted to an airbase for the Ecuadorian Air Force after a caricaturist labelled it a white elephant. The political-cycle story is the same in each case: the airport opens on the leader's calendar, and demand arrives, if at all, on a different one.

The financing leaves the state holding the loss

The fourth predictor is how the project is financed. The arithmetic of Latin American transport concessions, worked out by Engel, Fischer and Galetovic and by the World Bank's Guasch in 2004, is unflattering. More than half of these concessions get renegotiated after the contract is signed. When they get renegotiated, tariffs go up two times out of three, and the contract term gets extended one time out of three. In Colombia, the total renegotiated amounts have reached 85 per cent of what was originally invested. In Chile, 73 per cent of those renegotiations happen during construction, before the airport even opens. The contractor knew the model was wrong from the start. Across China, Ansar, Flyvbjerg, Budzier and Lunn found in 2016 that more than half of the infrastructure built in the last three decades has cost the country more than it returned. The financing instrument decides who eats the loss when demand fails to arrive.

Sri Lanka's Mattala was financed with a $190 million high-interest loan from China's Eximbank at approximately 6.3 per cent. SriLankan Airlines stopped flying to Mattala in 2015 and later said it saved $18 million a year by not flying there. In April 2024, amid Sri Lanka's ongoing $46 billion sovereign-debt restructuring following its 2022 default, Sri Lanka leased the airport to an Indian-Russian consortium. The loan and the loss stayed with the public; the operating revenue went elsewhere.

Nicaragua's Punta Huete is being reconstructed by China CAMC Engineering under a $491 million contract signed during the tenth anniversary of the Belt and Road Initiative in October 2023. The contract was hand-picked and did not go through international bidding, as Nicaragua's own Ley 323 de Contrataciones del Estado requires. The cost will sit on the Nicaraguan public balance sheet regardless of whether the projected 3.5 million passengers a year materialize.

Mozambique's Nacala airport is bound up in the corporate web of Odebrecht, the Brazilian contractor whose 2016 plea agreement with the US Department of Justice acknowledged $780 million in bribes paid to government officials across Latin America and the Caribbean to obtain $3.34 billion in contracts. Those contracts covered power plants, railroads, ports and airports, including Miami International Airport. Honduras's Palmerola airport opened in 2021 under a concession in which the state put in roughly $125 million and the private partner, Lenir Pérez's EMCO Holding, put in $87 million; under the contract canon structure the State receives no royalty until international departures exceed 600,000 a year — a threshold the Consejo Nacional Anticorrupción projects would not be reached until around year 28 of the 30-year concession. The state is also contractually obliged to maintain the access roads. Seven months after inauguration, a May 2022 storm caused the terminal roof to flood after a rainwater pipe failed. Brazil's Natal airport, the country's first private-sector airport concession, was also the first to be terminated. Inframerica requested indemnification for non-amortized capital investment after repeatedly arguing that traffic forecasts had been "out of kilter with reality." Zurich Airport Brasil took over operations in February 2024.

Lock-in once construction starts

The fifth predictor is lock-in. Cantarelli, Flyvbjerg, van Wee and Molin gave it that name in their 2010 paper. The mechanism is familiar: once a government has spent enough on a project, stopping looks worse than continuing, and the project starts to fund itself by inertia. Their additional finding was that the inertia is often deliberate. "Some decision makers create over-commitment, or exclude alternatives, in order to lock in their preferred projects," they write. The fix is to release money against construction milestones, not against electoral calendars.

Berlin Brandenburg Airport, BER, was planned to open on 30 October 2011. Actual opening: 31 October 2020, a delay of nine years. Budgeted at €2.83 billion in 2009; final reported cost approximately €6 to 10 billion, two to three times the original estimate. A 2015 whistleblower alleged that Imtech, which had filed for bankruptcy that year, had paid bribes to airport officials in exchange for inflated payments; in 2016 it emerged that the whistleblower had been poisoned and survived after a three-month illness. In October 2020 FBB announced it needed €375 million for 2021 alone to cover BER's current costs, plus €552 million as a stabilization fund "for missing passengers."

Mexico's NAICM project at Texcoco shows the lock-in cost in reverse. Cancelled in 2018 by the incoming López Obrador administration after a public consultation, the cancellation alone cost approximately 332 billion pesos according to the Superior Audit of the Federation, 232 per cent higher than the new administration's own estimate. The Sheinbaum administration will pay out more than 19 billion pesos tied to NAICM cancellation bonds, including $562 million for the 2026 tranche and $592 million for the 2028 tranche.

Willie Brown, the former mayor of San Francisco, gave the political logic plainly in a 2013 column: "We always knew the initial estimate was way under the real cost. In the world of civic projects, the first budget is really just a down payment. If people knew the real cost from the start, nothing would ever be approved. The idea is to get going. Start digging a hole and make it so big, there's no alternative to coming up with the money to fill it in." The lock-in is a design feature.

Another airport already serves the same area

The sixth predictor is the simplest. If the area the new airport is supposed to serve is already served by another airport, the travelers the new airport is built to carry already have a way to fly. The new facility has to convince them to switch. Pickrell's 1992 paper traced eight US rail-transit projects whose forecasts were used to argue for rail over keeping the bus. Actual ridership in Baltimore and Portland came in below half the forecast. In the other six cases it came in below a third. The same dynamic governs airports built where another airport already covers the route.

Spain's Castellón sits between Valencia and Barcelona. Both already had international service. The combined effect is a third airport competing for the same area that the existing two did not fill.

Peru's Pisco airport had an investment of S/153 million. In one reported month it moved eight passengers. The regional chamber of tourism estimates Pisco's inoperativity costs the regional economy about $20 million per year, a figure that rests on assuming Pisco could capture demand that Lima already serves. Capatur's president called Pisco a white elephant. Peru's Chinchero, the proposed Cusco replacement airport, is the same case in advance: Cusco's existing airport could be upgraded, but the new build at Chinchero presumes a mass of incremental demand the existing site cannot accommodate.

Honduras's Copán Ruinas airport at Río Amarillo is the closest Latin American parallel to Palmar Sur. Built in 2015 at roughly $13 to 15 million to serve the UNESCO archaeological zone, with the Honduran government forecasting that the airport's opening would draw 180,000 tourists to Copán Ruinas in 2015 alone, it registered twenty-seven flights between 2015 and 2021. Four per year. In February 2026 the president of the Copán Ruinas chamber of commerce told La Prensa that 90 per cent of foreign tourists still enter Copán overland from Guatemala. The tourists the airport was built for already had a road to Copán, and they kept using it. A note on the Costa Rican debate: the 2020 Surcos essay that brought this case into the conversation attributed Copán's failure to UNESCO objections over aircraft vibration. That framing does not hold up. The World Heritage Committee had already found "no direct impact on the Outstanding Universal Value" of the property in 2012. The real reason Copán's airport is empty is that nobody needed it. The road was already there.

How Palmar Sur stacks up

Each of the six predictors lines up with a specific, documented feature of the Palmar Sur project as currently proposed. Taken together, the six measurements describe a project that the academic literature would expect to underperform on every available axis, before the first construction crew arrives.

The forecast does not survive a comparison with what comparable airports actually delivered. The existing Palmar Sur aerodrome at MRPM is a small domestic strip, still open and still operating. In 2022, a pandemic-affected year, it handled 1,938 passengers; pre-pandemic traffic was in the low thousands. The proposal builds a new airport four kilometres from the existing aerodrome, on the closed Finca 10 strip, with a 2,600-metre runway sized for 180- to 200-passenger international jets. The domestic comparison makes the mismatch clearer. Quepos (XQP) sits adjacent to Manuel Antonio National Park, Costa Rica's most-visited protected area at 488,629 visitors in 2024, and is surrounded by mature tourism infrastructure: hotels, tour operators, a marina, transport networks, an established inbound corridor from SJO. Quepos handled 33,616 passengers in 2022 on regional service, about seventeen times Palmar Sur's traffic. Quepos is what an airport can look like when Costa Rica's strongest tourism draw is next door and the supporting infrastructure is already in place. Its scheduled service is still regional. Palmar Sur's principal draw is the UNESCO Finca 6 stone-spheres site, with no comparable tourism infrastructure built around it. The category shift the proposal asks for, from regional commuter service to scheduled international operations, is the gap the forecast has to close. Every airport in the catalogue this article has just walked through failed to close a gap of that size.

The site selection has overridden the technical record. SETENA rejected the Palmar Sur airport proposal in 2015. The current proposal is for the same Finca 10 site SETENA already turned down. The proposed runway sits 500 metres to a kilometre from the UNESCO Finca 6 stone-spheres site, less than a kilometre from the boundary of the Térraba-Sierpe national wetland.

The project is more political signal than funded plan. The Palmar Sur airport sits on the 2025 Plan Nacional de Desarrollo with a budget allocation of $42 to $105 million. Progress against the PND target is at zero per cent. What the state is actually paying for is visible in its other aviation-budget lines. The Juan Santamaría expansion has $390 million actively underway. Orotina went through a 2,829-page COCESNA-Mott MacDonald master plan with 24 technical studies and a $1.932 billion Phase 1 costing before being paused. Palmar Sur shows no equivalent planning footprint in the public record: no multi-volume master plan, no commissioned technical studies at that depth, no costed phasing. A PND line item at zero funded progress, with no visible engineering work underneath it, has the shape of a political promise.

The financing structure already encodes the downside, before construction. The Aeris Holding Costa Rica concession at Juan Santamaría contains clause 18.2. The clause grants the concessionaire a right to claim financial rebalancing if a new international airport enters operation elsewhere in the country during the concession term. Adenda 4, signed between CETAC and Aeris on 13 October 2022, extended that concession to 5 May 2036. A Palmar Sur international airport opening before 2036 would automatically trigger an Aeris compensation claim against the state. That is the central fact of this predictor. Whether the state would actually pay the claim is not a hypothetical question. The pattern of recent years has been the state finding ways to pay Aeris through whatever channel is available. The Contraloría General de la República, whose formal endorsement is legally required for major public contracts, refused to endorse Adenda 4 (resolution DCA-0232, March 2023); the DGAC, the civil-aviation administration that is party to the concession, endorsed it internally. CGR audit DFOE-CIU-IAA-00001-2025 found that CETAC had reclassified ₡8,615 million of Aeris exit-tax revenue to evade the Regla Fiscal. MOPT minister Luis Amador, who signed the Aeris extension, is under Fiscalía General criminal investigation for reassigning ₡3,209 million of CONAVI road-maintenance funds to pay Aeris on the same day he requested the transfer. A Palmar Sur concession would sit on top of the existing 18.2 exposure, inside the Latin American distribution where 54.4 per cent of transport concessions get renegotiated and 73 per cent of Chilean renegotiations happen during construction.

Lock-in has not yet been assembled, but it is the pattern to watch for. Palmar Sur is still at the early proposal stage. No earthworks, no signed construction contracts, no funded master plan at the scale of Orotina's 2,829-page, $1.2 million ICAO exercise. The lock-in predictor does not apply to a project that has not broken ground. It is the dynamic that would set in if the project moved from PND line item to funded construction, and at that point the other five predictors would be locked in with it.

The area is already served. The point applies at two levels. First: Costa Rica already has two airports with "internacional" in their names that handle zero scheduled international flights. Limón International, reopened in 2006, registered 3,107 domestic passengers and no international traffic in 2022; the DGAC director's own internal recommendation in January 2024 was to suspend its international category for insufficient fire-and-rescue equipment. Tobías Bolaños in Pavas ran 50,629 operations in 2022, of which 30,036 were flight-school training. Neither has the international traffic the designation implies. Second: the Pacific South is already reached, mostly by road. Manuel Antonio National Park, Costa Rica's most-visited at 488,629 visits in 2024, sits adjacent to Quepos (XQP), but only a small fraction of its visitors arrive by air. The rest drove, often on multi-destination tourist loops that thread Arenal, Monteverde, the central Pacific and the South Pacific. Drake (DRK), Puerto Jiménez (PJM) and Golfito (GLF) cover the Osa Peninsula on regional aircraft for travelers who want to skip the road legs. Sierpe has $20 boat service to Drake. The destinations airport proponents cite most often have hard ceilings on how much visitation they can absorb. Corcovado is legally capped at 560 visitors per day under the Sala Constitucional's August 2024 reinstatement of the prior limit; 2024 actual visitation was 102,764, less than half the cap. Piedras Blancas received 606 visitors in all of 2024. The area is covered at every scale, and the destinations that would fill an international airport at Palmar Sur do not exist at the volume an international airport needs.

The reference class

Reference-class forecasting does not tell you that a single project will fail. It tells you the range of outcomes its peers actually produced. The peers for Palmar Sur are the airports this article has just walked through: the ones whose forecasts did not match what comparable airports delivered, whose sites went ahead over technical objection, whose financing left the public holding the loss, whose construction created lock-in before the passengers arrived, and whose passengers already had another airport to use. Ciudad Real was sold at auction in 2016 for a twentieth of its build cost. Mattala was leased to an Indian-Russian consortium in 2024 during Sri Lanka's sovereign default. Mirabel's terminal was demolished. Copán Ruinas registers four flights a year. Palmar Sur, on the proposal as currently written, is being designed to join that group.