No Service Economy Without a Classroom: The Costa Rica Lesson
STORY INLINE POST
Nothing in this series happens without human capital, and the mandatory example — close, comparable, uncomfortable — is Costa Rica. A country that abolished its army in 1948, redirected spending to education for decades, and then enshrined the commitment in its constitution: Article 78 today mandates public education spending of at least 8% of GDP. When Intel chose Costa Rica in the 1990s — a decision studied ever since, starting with Debora Spar's World Bank case — it did not choose cheap labor; Costa Rica never had it. It chose bilingual technicians and engineers produced by a consistent education system, institutional stability, and a promotion agency, CINDE, that understood investment attraction is a professional service, like ours in Spyral, and not a bureaucratic errand.
The instructive part came later. When Intel moved its assembly operations to Asia in 2014, the doomsayers announced collapse. It never came: Costa Rica had already mutated, keeping and expanding Intel's engineering and global services functions and building an ecosystem around them of medical devices, corporate service centers, and software. Today, according to PROCOMER's own figures, services account for more than a third of Costa Rica's total exports — US$12.5 billion of US$35.2 billion in 2025, not counting tourism — and business services alone are the country's second-largest export, behind only medical devices. Its income per capita comfortably exceeds Mexico's, from a resource base and scale that are not remotely comparable.
The Rule, Not the Exception
Skeptics will say Costa Rica is a 5-million-person laboratory. Fair — but the pattern is the rule everywhere the leap has happened. India built its software industry on the Indian Institutes of Technology, founded from 1951 onward. Ireland did it on its regional institutes of technology and a promotion agency, the IDA, that operated with decades of continuity; Arora and Gambardella's comparative study of software industries in Brazil, China, India, Ireland, and Israel finds the same sequence in every case — technical human capital first, industry after. The Philippines converted its language advantage into a global services industry with revenue approaching US$38 billion and 1.8 million direct jobs. Nobody stumbles into a service economy. It is preceded, every time, by a deliberate, sustained educational decision.
And here the comparison demands honesty on our side. Costa Rica invests more in education, as a share of GDP, than the OECD average — and even so its learning results are middling; Mexico, with also-elevated spending ratios, gets results that are worse still. The problem is not only how much we spend, but on what, how, and for how long. Nothing compounds when every six-year administration uproots what the last one planted.
Three Imperatives, One Policy
The Mexican translation comes down to three moves that are really one. First, treat English as infrastructure, not as a school subject. A country that wants to sell services to North America with only marginal functional bilingualism competes with one hand tied behind its back. Technical English should be managed the way we manage a highway or a port: a twenty-year infrastructure investment, measurable, and shielded from political cycles — mandatory and certifiable across technical upper-secondary education.
Second, dual technical education at scale, wired into the clusters we already have. CONALEP and the technological universities should be to servindustria what the Fachhochschulen were to the German Mittelstand: training on the plant floor, curricula written jointly with industry, and clear routes from the lathe to the software. Querataro with the UNAQ and Nuevo Leon with the Tecnológico de Monterrey "Hub de Innovación y Emprendimiento" are already piloting dual models; the challenge is graduating from pilot to system.
Third — the most socially urgent — requalify the manufacturing worker we already have. Today's line operator is tomorrow's mechatronics technician, PLC programmer, or quality analyst, if someone trains him. The manufacturing rent should finance that transition: tax credits for certified training, not only for machinery, individual learning accounts, and recognition of skills acquired on the floor. It is also the only socially viable answer to automation — not protecting the job robotics will eliminate, but equipping the person for the job servindustria will create.
One last note on why this matters more than any tariff. Brazil's Embraer — the third-largest aircraft maker on Earth — is usually read as a triumph of protection. Read more closely: the real foundation was the Instituto Tecnológico de Aeronáutica, founded in 1950, which trained engineers for two decades before the company existed. Education with a purpose, twenty years before the first aircraft. That is the Costa Rican argument with a Portuguese accent — and the arithmetic is generational: the engineer who will orchestrate North America's supply chains in 2046 entered primary school this year. The 20-year horizon is not decided in 2046. It is decided in this year's classroom.









