# Before a State Breaks, Which Layer Goes First? Notes on Odd Lots with James Robinson

Source: Realpha Blog (blog.getrealpha.com)
Original article and charts: https://blog.getrealpha.com/en/blog/oddlots-2026-10-21-james-robinson-on-what-it-looks-like-when-a-state-/

> Listening notes on the Bloomberg Odd Lots live episode from Chicago (21 Oct 2026) with Nobel laureate James Robinson: extractive institutions, the failure of inclusion, and how industrial policy avoids capture. Educational notes on institutions and investing frameworks — not investment advice, no stock picks or price targets.

Published: 2026-10-21
Locale: en
Tags: institutional economics, Odd Lots, James Robinson, industrial policy, political risk

TL;DR: Robinson says the US is 'not yet' failing, but Americans without a college degree earn no more than they did 50 years ago — he calls that a failure of inclusion; he also expects China's unreformed politics plus opened economy to go off the rails, while admitting he cannot say how or when.


![An empty legislative corridor at dusk, one lit desk far at the end, a thin crack in the floor running toward the viewer](/covers/oddlots-2026-10-21-james-robinson-on-what-it-looks-like-when-a-state--cover.png)

> When timber snaps, the worm went through it first; when a wall gives way, the crack was already there. Yet a wormed beam will not break without a hard wind, and a cracked wall will not fall without heavy rain.
>
> —— 《韓非子・亡徵》 (Warring States period; my own translation)

On 21 October 2026, the Bloomberg Odd Lots live episode taped in Chicago featured James Robinson — Nobel laureate, co-author of *Why Nations Fail*, professor at the University of Chicago. Tracy Alloway's second question was whether the US is a failing nation; he answered "not yet," then named the wound he does see: in America, people without a college degree — what he called the narrow majority of the country — earn no more today than they did fifty years ago. He calls that a failure of inclusion. In the same hour he made a claim worth recording about China: an unreformed political system plus an opened economy is "extractive growth," and it will go off the rails — while admitting he cannot say how or when. The range of all this is decades of equilibrium, so it offers nothing about next year. Worth stating that limit up front.

## "Good institutions" used to be a label I stuck on an outcome

I had read *Why Nations Fail* and come away convinced, with one splinter I could never remove: Britain and the US are rich, so we say their institutions are good; Venezuela and Argentina are poor, so we say theirs are bad. Doesn't that story always win? I treated "institutions" as an after-the-fact label — enormous explanatory reach, zero forecasting value — and so it never entered any sheet I built when looking at companies.

![Three nodes form a closed loop that runs from outcome to label and back to explanation, with "zero predictive power" written in the middle — the shape itself is a circle with no way out.](/figures/institution-as-label-loop-en.svg)

Joe Weisenthal asked the splinter better than I could: suppose you knew nothing about a country's GDP. Would you have any other way to tell whether its institutions were good or bad? That question is about judging in advance, and I had only ever grumbled about it without asking how I would measure anything.

## Jamestown got an assembly only after two-thirds of them starved

Robinson's answer came in two layers, and both pushed back on what I thought.

The first layer is measurement. Look past income per head, he said: life expectancy, infant mortality. Countries failing economically tend to do badly across most socioeconomic outcomes, and those numbers don't wait on a statistics bureau. Then the example I keep coming back to: the most famous photograph in his book is the Korean peninsula at night, bright in the south, black in the north. Joe joked that the north looks a little brighter now, and Robinson replied, "the presidential palace in Pyongyang — there's a wave pool." Something as abstract as institutions is visible from orbit.

![A peninsula-shaped landmass is split horizontally into two halves: the top is almost entirely dark with a single bright point at its center, while the bottom is densely covered in bright points, so the difference in brightness is visible at a glance.](/figures/night-lights-two-halves-en.svg)

The second layer is mechanism, and this is where my old view fell over. He told the Jamestown story: the English colonists arrived with a playbook copied from the Spanish conquest of Latin America, planning to live off coerced indigenous labour. Virginia had no such population to exploit, and because they were busy looking for people to exploit instead of planting crops, two-thirds of them starved to death. Only after that model crashed did they experiment — to get English people to come, they had to offer terms worth coming for, so Virginia created a legislative assembly with universal male suffrage, an unprecedented innovation at the time. That is when I understood what he means by institutions: the equilibrium left standing after a group of people, given their actual distribution of power, fight to a stalemate. Bill Gates reviewed the book and complained that it explains why strong institutions matter without explaining how to get them; Robinson's reply sits on the same line — Gates thinks like an engineer, but social systems are not computer programs, and there is no lever to pull in Syria.

So institutions are a set of pressure relations that leave marks. Marks can be read in advance.

![On the left, two of three human figures are darkened to show that two thirds starved, and a bent arrow leads to the right, where an upward staircase reads: grant terms, set up an assembly, open up voting.](/figures/jamestown-two-thirds-en.svg)

## Three things I now watch, and the China question still on my desk

After the reversal I rewrote "institutions" into three questions I can research myself without waiting for a GDP print.

First, distribution — and look at it politically. Robinson said the biggest problem with inequality is the way it spills over into the political system, where the rules themselves get rewritten. His line on Elon Musk drew a clean distinction: when people say oligarch about Russia or Latin America they mean fortunes made by currying favour with government or through monopolies, whereas Musk made his money through innovation, and that should be acknowledged — yet it does society no good that he can pour hundreds of millions into elections. That gives me a ruler for companies: tell a technical moat from a relationship moat by how the money was made, not by how big it is.

Second, the question I have every time a government announces support for an industry: should I follow? Tracy raised the US taking stakes in chip companies and asked how you keep the process from going corrupt. He answered with one word — institutions — and when pressed, went on: it turns on how contracts are allocated and whether that process is accountable. His contrast case was Korean shipbuilding. Hyundai hadn't built a rowing boat in the late 1960s and went on to build a world-class industry; the US could do the same, he said, but "you need to be immensely serious. You need to plan. You need to have strategy," then added, "I don't see that." I've made that my opening move on policy news: look for evidence of a plan and of accountability, and absent both, treat it as a one-off subsidy rather than a decade-long industry. He also noted Korea's version lasted because governance became democratically accountable in the 1980s — and that the Soviet Union grew fast for fifty years too.

![A flow of money enters and passes through two gates; what gets through continues to the right as a long channel, while what does not leaks downward just in front of the gates as a short stub.](/figures/policy-two-gates-en.svg)

Third, the one I have no answer for: China. He calls it extractive growth that will go off the rails, on the reasoning that power tends to corrupt and absolute power corrupts absolutely. When Tracy asked what would make him think his creative-destruction argument was wrong, he joked that he thinks he's wrong all the time, he just doesn't admit it in public — then offered the counter-case himself: China has no tradition of accountability going back two and a half thousand years, yet also a millennium-old examination system and the social mobility he sees first-hand teaching Chinese students. Could that cultural combination be the exception that holds? He left the door open. (In passing, he described the British elite as "just a bunch of demobilized warlords," which made me laugh for a while.) I give more weight to someone willing to name where he might be wrong — and my own conclusion stops here: I accept the direction, I won't use the timing.

The AI section lands on the same distribution problem. He drew the analogy to the Industrial Revolution: once cotton spinning and weaving were mechanised in Britain, weavers' wages fell by two-thirds, and the reaction was the Luddites taking hammers to the machinery. He said the biggest factor explaining stagnant incomes for Americans without a college education is the diffusion of computers displacing less-skilled work — and we are still recovering from that round while AI arrives. Productivity gains he does not doubt; how the benefits get spread is the critical part, and there he said plainly, "I don't know." His current research uses AI too: digitising large bodies of political philosophy to study how the arguments of Hobbes and Locke shaped how people acted in 17th-century British politics — work he says was impossible before.

## The One Thing to Take With You

One idea: the outcomes you see today are what past arrangements about whose voice counts left behind, so to read where something is heading, look at who sets the rules and who bears the cost when it goes wrong.

One thing to try today: pick an arrangement near you that bothers you — who decides how the family spends the holidays, which company process blocks everyone, some rule your building committee passed — and do just one thing with it. On a single sheet of paper, write down who has the final say, and who carries the consequences when it goes badly. I tried this once and found two different names in those two boxes, and after that I stopped wondering why the thing never gets fixed.

![A sheet is divided into two columns, with the human-figure marker in the left column placed high and the one in the right column placed low, and a diagonal line connecting the two offset positions.](/figures/who-decides-who-bears-en.svg)