The Template Trap: Why Borrowed Policies Fail

In October 1777, news reached Britain that General Burgoyne had surrendered his army at Saratoga. Thirteen colonies were lost and the war was going badly. Sir John Sinclair wrote to Adam Smith in a state of alarm: “If we go on at this rate, the nation must be ruined.”

Smith, then in his fifties and the most famous economist alive, replied with the calmest sentence in the language:

“Be assured, my young friend, that there is a great deal of ruin in a nation.”

It is worth being precise about what he meant, because the line gets quoted loosely. Smith was not saying that decline can continue forever. He was saying that a nation absorbs shocks that look fatal from the inside. Britain lost the American colonies - an empire’s first instalment - and went on to industrialize the world.

A nation can carry a great deal of ruin. It cannot carry unlimited ruin. That distinction is the whole point of Smith’s reassurance, and it is the point of this article.

The same mistake, made with policies

Here is a sentence you have heard in every development debate for forty years.

“It worked in [country]. Why not here?”

It is a reasonable question. It is also, on the evidence, the wrong question.

The great experiment in transferring policies wholesale was the Washington Consensus - the package of liberalization, privatization, and stabilization that the international institutions promoted from the late 1980s. These were not stupid policies. Most had worked somewhere. The theory was that a reform package proven in one place would produce growth in another, and the institutions had the authority to insist on it.

The results were a disappointment so consistent that it reshaped the field. Chile, Poland, and Estonia did well. Argentina, Russia, and much of sub-Saharan Africa did not, and Russia’s experience in the 1990s was a catastrophe that set the country back a generation.

The academics who studied the failure reached an uncomfortable conclusion. The individual reforms were sound, but the package did not travel. What worked in Santiago did nothing in Moscow - or worse.

The evidence got better, and the problem got worse

For a while, the hope was that better measurement would settle things. Randomized controlled trials arrived in development economics with a promise: stop arguing about theory, test what actually works, and count the results.

The promise half-delivered. Thousands of careful trials were run. And the finding that emerged is that even the tightest results often do not survive the trip to a new setting.

Deworming is the case everyone learned from. A study published in 2004 found that treating children for parasitic worms - a cheap, safe intervention - improved health and school attendance for millions. It became the flagship of evidence-based aid, cited in fundraising appeals and adopted at scale.

Then it was re-tested. Follow-up work, including large replication efforts, found effects far smaller than the original claimed, and in some settings close to nothing. The intervention was not worthless. It was, however, considerably less magical than one influential trial suggested - and the gap between those two statements is where development budgets go to die.

Microcredit followed the same arc. It was going to end poverty by lending small sums to the very poor. Rigorous evaluations found modest and inconsistent effects. Some studies detected small benefits for some households. Others detected none.

The general lesson now has a name. Economists speak of the external validity problem - whether a result found in one population tells you anything about another. Researchers write papers with titles like “Can results from rural Kenya predict outcomes in urban India?” The answer is more often no than anyone wanted.

What the trials delivered was not a toolkit. It was proof that a toolkit was the wrong model.

What actually travels

If neither the reform package nor the tested intervention travels reliably, what does?

There is one approach that has done better, and it starts by refusing to answer the question directly.

Growth diagnostics, developed by Ricardo Hausmann, Dani Rodrik, and Andres Velasco, begins with a simple observation: in any economy, at any moment, there is usually one constraint that binds. Capital is not scarce and labor is plentiful and roads are fine - but the power grid cannot keep the factories running, so nothing else matters. Remove that constraint and growth resumes until the next one binds.

Their framework is an attempt to identify, country by country, what the binding constraint actually is, so that reform spending goes where it produces “the biggest bang for the reform buck.”

The critical phrase is country by country. The method is universal. The findings are not.

This is where the growth-diagnostics literature diverges sharply from the earlier generation. It is not that the Washington Consensus reforms were wrong. It is that their ranking was. A reform list tells you what is generally good. A diagnostic tells you what is binding here, now, and therefore worth the political capital.

The rule that falls out of the evidence

Draw the line between policies that travel and policies that do not, and a pattern appears.

The pattern is not about left and right, and you can see that from the table: conditional cash transfers sit on the transferable side and social trust sits on the other, and neither maps onto a political tribe.

The pattern is about dependency. The more a policy depends on how people behave and what institutions already exist, the less of it survives the journey.

Four reasons, and each one is fatal on its own.

Complementarity. The parts only work together. Denmark’s immigration and welfare settlement is not one policy. It is restrictive entry, aggressive integration requirements, and universal welfare - and the third may be what makes the first two politically survivable. Import the restriction without the universality and you get resentment without cohesion. Import the universality without the restriction and the arithmetic changes. Take one piece and you do not get a third of the result. You get something else entirely.

Behavioral response. People adapt to rules, which means the rule stops producing the behavior it was written for. When Britain taxed investment income at 98% in the mid-1970s, the wealthy did not pay 98%. They restructured, relocated, and shifted how they took income - which is why, as we have argued elsewhere, the effective rate on the top 1% in that era was nothing like the headline number. You cannot import a policy without importing the adaptation.

Path dependence. What a country can do next depends on what it did before. Korea’s 1960s reforms are not replicable today even in Korea, because the Korea that could do them is gone. History is not a constraint on policy. It is a determinant of which policies exist as options.

Unobserved preconditions. This is the one that kills reforms quietly. The thing that made a policy work is often the thing you cannot see, measure, or buy - trust being the obvious example. A policy that relies on people believing the tax authority is fair will fail in a country where they do not, and no amount of implementation capacity fixes it.

The knowledge problem, again

There is a reason this pattern is familiar to anyone who has read Hayek.

His argument in “The Use of Knowledge in Society” was not that central planners were stupid. It was that the knowledge required to allocate resources does not exist in any single place. It is dispersed - held by millions of people who each know their own circumstances and cannot articulate the whole. The price system works because it aggregates that knowledge without anyone needing to understand it.

We made a version of this argument about central planning in The Information Problem. The same logic applies with equal force to policy transfer, and this is the part that most reform advocates miss.

When a consultant arrives from the IMF carrying a reform list, they are carrying an answer that was computed elsewhere, from knowledge they do not have. What they lack is not intelligence or good faith. It is the specific, local, unarticulated knowledge of how this economy actually works - which regulation is enforced and which is a formality, which official must be paid, why the previous reform failed. That knowledge exists. It is just not in the consultant’s briefcase.

The version of the consultant’s job is not to bring the answer. It is to help the country find its own.

Three countries, three constraints

Vietnam grew at about 8% in 2025, one of the fastest rates on earth. The temptation is to explain it with tax and regulation, and to recommend that others copy the recipe. But the growth has a different engine: integration into global supply chains, with Samsung, Intel, LG, and Foxconn expanding capacity in Bac Ninh, Binh Duong, and Dong Nai as manufacturing shifts out of China.

The binding constraint is not tax either. The constraint is procedural.

Consider the numbers on public investment. For 2024 the government set a target of 95% disbursement of its public investment budget. Late in the year the Ministry of Finance was still reporting that the money was not moving. The causes named were not lack of funds but slow site clearance and slow bidding - the administrative machinery between a decision and a shovel in the ground.

For private projects the equivalent obstacle is the Environmental Impact Assessment. Vietnamese legal practice describes it as “the approval most frequently underestimated by foreign investors, and the one most likely to blow out the project timeline.” An entire consultancy industry exists to steer investors through it.

Behind both sits a harder problem. Vietnam’s anti-corruption campaign, launched in 2013 and known as the “blazing furnace,” has prosecuted thousands of officials, many of them senior. The campaign is popular and has done real work. It has also been described by the international bar as a double-edged sword: when signing an approval carries personal risk, the rational response is to sign nothing. Officials delay, refer upward, and wait. Projects stall not because anyone blocked them, but because nobody will move them.

That is what a binding constraint looks like in practice, and note what kind of thing it is. It is not a tax rate, and it cannot be fixed by cutting a budget or liberalizing a market. It is a question of whether the people who must approve things believe they will be safe if they do. Which is the same “hard to read government” problem from the other end: the investor cannot predict the rules, and neither, it turns out, can the official applying them.

Land has teeth here too, and the absence of freehold limits what a business can pledge as collateral. But Vietnam rewrote its land law in 2024 - Law No. 31/2024/QH15, in force since August that year - overhauling land pricing and use rights. Land is being addressed. Approvals are the thing still stalling.

The power constraint is real as well: serious shortages in 2023 halted factories, and grid capacity remains a live question as manufacturing expands. But energy is the tractable one. A power shortage can be attacked with capital, contracts, and construction. A chilling effect on official decision-making cannot, and no amount of infrastructure spending fixes it.

Britain is the mirror image, and its argument about immigration is aimed at the wrong variable.

Nobody objects to people who arrive to work. The complaint is about those who arrive and do not contribute - and that is a question of design before it is a question of numbers, because a rule forbidding work manufactures the outcome it is then blamed for.

The evidence on this is unusually direct. Fasani, Frattini and Minale studied what happens when countries bar asylum seekers from working, using almost thirty years of data on refugees arriving in Europe. Exposure to a ban at arrival reduced employment probability by 15% in the years that followed, an effect driven by lower labor market participation, growing with the length of the ban, and persisting for up to ten years.

It was not mechanical - they excluded everyone still under a ban. People barred from working did not go on to work.

The cost of that design is measurable. The same study put the output lost to the bans imposed on asylum seekers who arrived during the 2015 crisis at EUR 37.6 billion.

That is the complementarity problem in miniature. It is not the number of arrivals that determines whether they contribute. It is whether the rules let them.

What the data points at is a regional economic problem that deindustrialization and austerity deepened, compounded by a welfare design that marks recipients as claimants rather than members - the opposite of the universal design that appears to build trust rather than erode it.

America is the same story with more money. In 2009 the Obama administration sold its stimulus as a program of “shovel-ready” projects - work that could begin the moment the funds arrived. Within six weeks it had announced funding for about 2,000 transportation projects. By October 2010 the president had abandoned his own phrase: “there is no such thing as a shovel-ready project.”

The accounting explains why. Of the $787 billion package, the Department of Transportation received $48.1 billion - roughly 6% of the total, to cover every road, bridge, and transit project in the country. Through the end of 2009, the White House’s own Council of Economic Advisers estimated the act had produced or saved 262,000 construction jobs. The headline figure was always “3.5 million jobs saved or created” - a claim that cannot be tested, because a job that never existed cannot be observed to have been saved.

The charging stations make the point more cleanly. When Obama took office there were roughly 500 public charging stations in the United States. His administration set a goal of 100,000 by 2015. By the time he left office there were about 16,000. That is a thirty-fold increase and 16% of a promise at the same time, and both facts are true.

No one would describe America as corrupt in the Vietnamese sense. The obstruction is nevertheless the same kind of thing: permitting, environmental review, local veto points, and officials who cannot be blamed for delay but can be blamed for error. Announcements are cheap. Shovels are not.

It can be reversed, though rarely

The diagnosis should not end in fatalism. Procedural paralysis and corruption are not permanent national characteristics, and the counter-examples are worth keeping in view.

Singapore went from one of Asia’s most corrupt countries to among the cleanest in roughly fifty years, through four things: strong laws, independent adjudication, consistent enforcement, and a civil service paid enough that bribery is not worth the risk. Hong Kong’s ICAC has done comparable work since 1974. Both required sustained political will at the very top, which is exactly what makes the success rare rather than routine. It is also notable that the countries that managed it - Singapore, Hong Kong, Denmark - were all small enough for a single determined leadership to reach every part of the system.

So it can be done. The constraint is not whether reform is possible. It is whether anyone with authority has diagnosed the actual problem and is prepared to spend a decade on it.

What Smith was really saying

Return to Saratoga.

Smith told a frightened young man that nations absorb shocks. He was right, and the following two centuries proved it. But note what his reassurance requires. It requires that someone eventually stops doing the thing that is causing the damage - that the ruin, however large, is finite.

Which brings us back to the transfer problem, and to why it matters that we get this right.

If a nation’s growth depends on finding its own binding constraint, then the most damaging thing a reform program can do is import an answer that does not fit, spend a decade and a country’s political capital on it, and conclude that reform is impossible when it fails.

That is how ruin accumulates. Not through one catastrophe - nations absorb those - but through a sequence of confident, well-funded, imported mistakes, each one making the next reform harder to sell.

The method travels. The answer does not. A country that knows how to diagnose its own constraint can learn from everywhere, because it knows what it is looking for. A country that only knows how to import answers has to keep guessing - and blaming the evidence when the guesses fail.