Skip to content

Lesson M19.L04: Institutions, Geography, and Development

Module: Economic Growth Part II Level: intermediate Duration: 30 minutes Learning Objective: Evaluate the relative importance of institutions versus geography in determining long-run income levels. Data as of: 2024 Provenance: World Bank Open Data | NBER Working Papers

Explanation

Why are some countries rich and others poor? Two major competing explanations dominate the development literature:

1. Institutions hypothesis (Acemoglu, Johnson & Robinson 2001 โ€” "AJR"). Secure property rights, rule of law, and constraints on executive power are the fundamental cause of long-run prosperity. Countries with "inclusive" institutions attract investment, enable contracting, and protect innovators. "Extractive" institutions โ€” designed to concentrate rents in the hands of an elite โ€” suppress growth. AJR's key contribution is causal identification: they use European settler mortality as an instrumental variable (IV). Where settlers faced high mortality (malaria, yellow fever โ€” e.g., West Africa), they established extractive institutions and did not migrate in large numbers. Where mortality was low (e.g., Australia, New Zealand, Canada), they settled and replicated European property rights. Because settler mortality affects current income only through the institutions it produced, it is a valid instrument. The 2SLS estimates show strong causal effects of institutions on GDP per capita.

2. Geography hypothesis (Sachs; Diamond 1997). Sachs: tropical geography increases disease burden (malaria, helminthiasis) which directly reduces worker productivity and life expectancy โ€” a poverty trap independent of institutions. Diamond (Guns, Germs and Steel): biogeography โ†’ agricultural productivity differences โ†’ population density differences โ†’ institutional differences. In Diamond's view, institutions are endogenous to geography.

The Australian natural experiment. Australia and Pacific island nations (PNG, Fiji, Vanuatu) share broadly similar tropical/sub-tropical geographies in their northern regions. Yet Australia's GDP per capita (~USD 65,000) dwarfs PNG's (~USD 3,200). AJR argue this is explained by the British institutional legacy โ€” common law, secure property rights, parliamentary democracy โ€” transplanted via settlement. Geography alone cannot explain a 20:1 income gap between near-neighbours.

Key terms: - Inclusive institutions: broad property rights, rule of law, access to markets for all - Extractive institutions: designed to extract resources for a small elite, discouraging investment - Instrumental variable (IV): a variable that affects the outcome only through its effect on the endogenous variable (here: settler mortality โ†’ institutions โ†’ income) - 2SLS: Two-stage least squares, an IV regression technique

Worked Example

The AJR settler mortality IV โ€” logic of the natural experiment.

Problem: Regressing current income on current institutional quality (e.g., "protection against expropriation") gives a biased estimate because richer countries may simply afford better institutions (reverse causality), and omitted variables (culture, geography) affect both.

AJR's solution โ€” two stages:

Stage 1 โ€” Regress current institutional quality (INST) on settler mortality (MORT):

\[\widehat{INST}_i = \hat{\gamma}_0 + \hat{\gamma}_1 \ln(MORT_i) + \hat{\varepsilon}_i\]

Empirical finding: ฮณฬ‚โ‚ < 0 (higher settler mortality โ†’ worse institutions today). F-statistic >> 10 (strong instrument).

Stage 2 โ€” Regress log income per capita (ln Y) on predicted institutional quality:

\[\ln(Y_i) = \hat{\beta}_0 + \hat{\beta}_1 \widehat{INST}_i + \hat{u}_i\]

Empirical finding: ฮฒฬ‚โ‚ โ‰ˆ 0.94 (one-unit improvement in institutional score โ†’ ~94% higher per-capita income). This is substantially larger than the OLS estimate, suggesting OLS underestimates the true institutional effect due to measurement error in institutions.

Numerical illustration:

Suppose two hypothetical colonies: - Colony A: ln(settler mortality) = 3.0 (low mortality); predicted INST = 8.5 / 10 - Colony B: ln(settler mortality) = 6.0 (high mortality); predicted INST = 5.0 / 10

Using ฮฒฬ‚โ‚ = 0.94:

\[\ln(Y_A) - \ln(Y_B) = 0.94 \times (8.5 - 5.0) = 0.94 \times 3.5 = 3.29\]
\[\frac{Y_A}{Y_B} = e^{3.29} \approx 26.8\]

The institutions difference alone โ€” instrumented via settler mortality โ€” predicts a 27-fold income gap. This is close in magnitude to observed gaps between high- and low-institution former colonies.

Common Misconception

Misconception: "Geography determines institutions, so controlling for geography eliminates any independent effect of institutions on income."

Correction: AJR's IV strategy is specifically designed to address this. Settler mortality is largely determined by disease ecology (a geographic factor), but its effect on current income runs through institutions, not through direct productivity effects on today's workers. When AJR include latitude and other geographic controls alongside their IV, the institutional effect remains large and statistically significant. This suggests institutions have causal force beyond their geographic origins โ€” an economy can improve institutions even without changing its geography, with substantial income effects.

Practice Prompts

  1. Conceptual: Why is "settler mortality" a useful instrumental variable for institutional quality rather than simply including geography directly?

โ†’ Answer: An IV must satisfy two conditions: (i) relevance โ€” it must strongly predict the endogenous variable (settler mortality strongly predicts institutional quality because high-mortality environments discouraged European settlement and hence replication of European property-rights institutions); (ii) exclusion restriction โ€” it must affect the outcome only through the endogenous variable. Settler mortality in the 17thโ€“19th centuries no longer directly affects productivity (diseases have been controlled), so its effect on current income plausibly runs only through the institutions it produced. Direct geographic controls (latitude, temperature) cannot serve this role because they may directly affect agricultural productivity, disease burden, and labour productivity today โ€” violating the exclusion restriction.

  1. Numerical: Using AJR's estimated coefficient ฮฒฬ‚โ‚ = 0.94, if a country improves its "average protection against expropriation" score from 4.0 to 7.0 (on a 0โ€“10 scale), by what percentage does predicted per-capita income increase?

โ†’ Answer: - ฮ”INST = 7.0 โˆ’ 4.0 = 3.0 points - ฮ”ln(Y) = 0.94 ร— 3.0 = 2.82 - % change in Y = e^2.82 โˆ’ 1 = 16.78 โˆ’ 1 โ‰ˆ 1,578% increase - Equivalently, Y rises by a factor of e^2.82 โ‰ˆ 16.8ร—

A three-point improvement in institutional quality is associated with roughly a 17-fold increase in per-capita income โ€” highlighting the enormous predicted returns to institutional reform in developing countries.

  1. Application: Papua New Guinea and Australia share a geographically proximate region (Cape York Peninsula/PNG border). How would AJR, Diamond, and Sachs each explain the ~20:1 income gap between Australia and PNG?

โ†’ Answer: - AJR: British colonisation of Australia transplanted inclusive institutions (common law, parliamentary democracy, property rights), while PNG's colonial history (German then Australian administration, independence 1975) produced weaker institutions. Settler mortality was high in PNG (malaria, tropical disease) โ†’ few European settlers โ†’ extractive governance structures persisted. Institutions are the fundamental cause. - Diamond: Geographic and biogeographic differences shaped early productive capacity โ€” Australia's temperate south had better conditions for domesticable crops and animals, enabling denser settlement and more complex institutions. PNG's mountainous terrain fragmented polities. - Sachs: PNG's high tropical disease burden (malaria is endemic, especially in lowland PNG) directly reduces worker productivity and life expectancy today, creating a poverty trap independent of institutions. Better bed-net programs, malaria eradication, and health infrastructure could raise PNG incomes even without institutional reform.

Further Resources