Two countries can start with similar resources and technology and, fifty years on, differ many times over in income per head. The Solow model says capital and technology drive that gap, but takes saving and innovation as given — it never says why one country accumulates and another stalls. That question is development economics, the topic that most often brings intermediate students to an economics tutor in Glasgow.
1 · What “institutions” actually means
Institutions are not buildings or organisations. They are the rules of the game — the constraints, formal and informal, that decide what people expect when they invest and produce.
Three do most of the work. Property rights: can you keep what you build? Contract enforcement: if a deal breaks, will a court enforce it? Constraints on the executive: can power expropriate and rewrite the rules at will, or is it checked?
Secure those and the return on investment is predictable, so people invest. Leave them insecure and the smart move is to consume now or send wealth abroad. Institutions set the incentive to accumulate and innovate — what the growth models, from Solow to the endogenous-growth theories built on ideas, take as given.
2 · The correlation — and why it proves nothing on its own
Plot income per head against any index of institutional quality and you get a clear upward cloud: richer countries score better. The scatter below shows the shape — stylised data, not real countries, but the pattern the real data show. Do not read it as institutions cause growth: one cloud fits three stories.
One — institutions cause income. Secure property rights raise investment, so income rises. The causal claim.
Two — income causes institutions (reverse causality). Good courts are expensive, so a rich country can afford them and a poor one cannot. Same cloud, no causal effect.
Three — something causes both (omitted variables). Climate, schooling, or resources could lift income and good institutions at once, loading onto the institutions coefficient.
The reading lesson examiners lean on: a raw slope measures association, the sum of all three — an upper bound on the causal effect, not the effect.
3 · The natural experiment: settler mortality
To pick out story one, you need variation in institutions that income and the omitted variables did not produce. Acemoglu, Johnson and Robinson found a source in colonial history.
Where colonisers faced high mortality — diseases they had no resistance to — they did not settle, and built extractive institutions: power held by a few, property insecure, entry blocked. Where they could settle, they built inclusive ones: broad secure property, impartial courts, a constrained executive — and effort pays off only when the returns are safe. These institutions persisted long after independence.
The identifying idea: past mortality plausibly affects income today only through the institutions it seeded, not directly. If so, it isolates the institutional variation income and geography did not cause. Putting a number on it is the job of instrumental variables, a separate page; the logic is the point — a lever that moves institutions and nothing else.
The cleanest case shares everything the omitted-variable worry names: the two Koreas — one language, one geography, one history — then opposite institutions after division, and incomes that diverged past comparison. The rules changed; nothing else did.
The instrument is debated, honestly: the mortality data and the exclusion assumption have both been challenged — patchy figures, and colonial channels beyond institutions such as human capital. Not settled, and this page does not settle it.
4 · Rival explanations, and what reform can do
The institutional story has real rivals. Geography is the strongest: climate, disease, soil, and sea access shape productivity directly — a malarial lowland is poor for reasons unrelated to its courts. But much of that may run through institutions, since the colonisers’ disease environment shaped how they governed. Culture — trust, work norms — plausibly matters too, but co-evolves with institutions, hitting the same reverse-causality wall.
Two sobering points for policy. Reform is slow: rules that lasted centuries do not yield to a five-year plan, because those who profit from extraction defend them. And transplants fail: a copied constitution rarely reproduces the outcome, because the informal norms and balance of power beneath it did not come too.
Worked example — reading the stylised scatter
The diagram is a stylised cloud: quality across, log income per head up.
Step 1 — Fit. OLS gives ŷ = 7.57 + 0.435x (slope rounds to 0.43), x the quality index, ŷ predicted log income. The slope is positive.
Step 2 — Judge it. R² = 0.68: quality alone explains about two-thirds of the variation — strong, but noisy. A third is left over.
Step 3 — Predict. Compare x = 3 with x = 7. The log-income gap is 0.435 × 4 = 1.74; the axis is logarithmic, so that is a ratio: e^1.74 ≈ 5.7 — about 5.7 times the income per head.
Step 4 — Re-read for reverse causality. If half that link runs backwards, the causal slope falls to 0.5 × 0.435 ≈ 0.22, and the same gap predicts e^0.87 ≈ 2.4 — under half the raw figure. The scatter did not move; your reading did.
Step 5 — Re-read for omitted variables. The leftover third is where geography, human capital, and culture sit. If any lifts both, part of the 0.435 is theirs and the slope shrinks again. The raw slope is a ceiling.
Step 6 — Interpret. The scatter is honest about association and silent on cause. A causal number needs variation income and geography did not make — the settler-mortality experiment is one try. It poses the question; it cannot answer it.
Shown an upward-sloping cloud of institutions against income, could you name the three stories behind it — and say which one a natural experiment is built to isolate? Reading a correlation as a ceiling, not a cause, is exactly what development-economics questions reward. A one-on-one economics tutor works the reverse-causality and omitted-variable traps with you until the identification logic is an instinct, not a definition you recite. Book a trial session.
Practice
Q1. Using ŷ = 7.57 + 0.435x, predict the log-income difference between economies scoring 8 and 5 on the index, then convert it to an income ratio.
Q2. Suppose 40% of the estimated institutions–income association runs from income to institutions. What causal slope does that imply, and does raw OLS over- or under-state the effect?
Q3. The fit has R² = 0.68. What share of the variation is left unexplained, and name two things it might reflect?
Answers. Q1: 0.435 × (8 − 5) = 1.30 log points; ratio e^1.30 ≈ 3.7, about 3.7× the income per head. Q2: implied slope 0.6 × 0.435 ≈ 0.26; part of the link is reverse, so raw OLS over-states the effect. Q3: 1 − 0.68 = 0.32 — about a third unexplained, leaving room for geography, human capital, culture, or measurement error.
Key takeaways
- Institutions are the rules of the game — property rights, contract enforcement, constraints on the executive — not buildings or geography.
- A raw correlation is a ceiling, not a cause: it fits reverse causality and omitted variables as well as causation.
- Natural experiments are the way out: isolate institutional variation income and geography did not cause. The data are debated; the logic is not.
- Reform is slow and transplants fail: the powerful defend the rules that favour them.
Why Glasgow University students choose our economics tutoring
- Identification as a habit: sessions drill the reflex examiners reward — read a correlation, then name the reverse-causality and omitted-variable stories behind it.
- Models interrogated, not memorised: you learn to fit and cross-examine a relationship, not recite a result, so you can defend it under pressure.
- One-on-one, matched to your course: a PhD tutor works from your notation and past papers, whether your module follows Acemoglu and Robinson, Ray, or Banerjee and Duflo.
FAQ
Q: What do economists mean by “institutions”?
A: The rules of the game — property rights, contract enforcement, and constraints on power — the constraints that decide whether investment pays off.
Q: Why doesn’t the correlation prove institutions cause growth?
A: The same correlation appears if income causes institutions, or if a third factor like geography raises both. A raw slope mixes all three.
Q: What is the settler-mortality natural experiment?
A: Where colonisers faced high mortality they built extractive institutions that persisted. Since past mortality plausibly affects income today only through institutions, it can isolate their effect. The data are debated.
Q: What is the difference between extractive and inclusive institutions?
A: Extractive institutions concentrate power and expropriate the many; inclusive ones spread power and protect property broadly, which makes investment pay.
Q: If institutions matter, why not just copy a rich country’s laws?
A: Formal rules sit on informal norms and a balance of power that do not transplant, so a copied constitution rarely reproduces the outcome.
Book an economics tutor in Glasgow or online
Development economics rewards students who read a correlation critically — who spot the reverse-causality and omitted-variable traps before an examiner does. One-on-one sessions build that instinct on your own past papers. Tell us your course and exam date, and we will match you with the right tutor this week.
