Institutions & Growth: An IV Approach

Research

Econometrics

Causal Inference

Does institutional quality cause economic growth?

Overview

This bachelor’s thesis investigates the “deep determinants” debate in development economics — why some countries are rich and others poor. The paper sits between two competing literatures:

AJR / RST (institutions-first): settler mortality IV shows institutions drive income per capita
GLLS (human capital-first): human capital predates institutions; Rule of Law proxies are endogenous outcome variables

The contribution is a 3-endogenous-variable 2SLS instrumenting institutions, trade, and human capital simultaneously.

Model

Log(Income) = α + δ·Institutions + η·Human Capital + φ·Trade + θ·Geography + ε

Endogenous variables: Institutions (ROLA / Democracy), Trade, Human Capital

Instruments: Settler Mortality, Malaria Ecology, Constructed Trade, ENGFRAC, EURFRAC

Exogenous control: Tropicar (% landmass in tropics)

Results

Human capital (years of schooling) is the only statistically significant variable across all 2SLS specifications. Institutions lose significance once endogeneity is controlled for via IV.

Specification	N	HC Coef	Sig
ROLA · Full sample	110	0.603	*
ROLA · Colonial subsample	51	0.460	**
Democracy · Full sample	109	0.638	**
Democracy · Colonial subsample	50	0.630	***

Colonial subsample: SettlerMortality > 0. Robust SE used.

A 1-year increase in average schooling → 46–64% increase in income per capita (semi-elasticity, varies by specification).

Limitations

Cragg-Donald statistic falls below all Stock-Yogo critical values (2SLS bias > 30%). Results are suggestive but not conclusive. LIML estimation confirms weak identification — consistent with GLLS (2004), which encounters the same problem.

Code & Data

Python Notebook — statsmodels OLS + linearmodels.IV2SLS
Stata .do files — original ivregress 2sls, Sargan, Cragg-Donald, LIML
Full Report

Key References

Acemoglu, D., Johnson, S. & Robinson, J. (2001). The Colonial Origins of Comparative Development. NBER WP 7771.

Glaeser, E. et al. (2004). Do Institutions Cause Growth? NBER WP 10568.

Rodrik, D., Subramanian, A. & Trebbi, F. (2004). Institutions Rule. Journal of Economic Growth, 9(2), 131–165.

--- title: "Institutions & Growth: An IV Approach" description: "Does institutional quality cause economic growth?" categories: [Research, Econometrics, Causal Inference] --- ## Overview This bachelor's thesis investigates the "deep determinants" debate in development economics — why some countries are rich and others poor. The paper sits between two competing literatures: - **AJR / RST (institutions-first):** settler mortality IV shows institutions drive income per capita - **GLLS (human capital-first):** human capital predates institutions; Rule of Law proxies are endogenous outcome variables The contribution is a **3-endogenous-variable 2SLS** instrumenting institutions, trade, and human capital simultaneously. ## Model ``` Log(Income) = α + δ·Institutions + η·Human Capital + φ·Trade + θ·Geography + ε ``` **Endogenous variables:** Institutions (ROLA / Democracy), Trade, Human Capital **Instruments:** Settler Mortality, Malaria Ecology, Constructed Trade, ENGFRAC, EURFRAC **Exogenous control:** Tropicar (% landmass in tropics) ## Results ::: {.result-box} Human capital (years of schooling) is the **only statistically significant variable** across all 2SLS specifications. Institutions lose significance once endogeneity is controlled for via IV. ::: | Specification | N | HC Coef | Sig | |---|---|---|---| | ROLA · Full sample | 110 | 0.603 | * | | ROLA · Colonial subsample | 51 | 0.460 | ** | | Democracy · Full sample | 109 | 0.638 | ** | | Democracy · Colonial subsample | 50 | 0.630 | *** | *Colonial subsample: SettlerMortality > 0. Robust SE used.* A 1-year increase in average schooling → **46–64% increase in income per capita** (semi-elasticity, varies by specification). ## Limitations Cragg-Donald statistic falls below all Stock-Yogo critical values (2SLS bias > 30%). Results are suggestive but not conclusive. LIML estimation confirms weak identification — consistent with GLLS (2004), which encounters the same problem. ## Code & Data - [Python Notebook](https://github.com/tkay305/iv-regression-institutions-growth) — `statsmodels` OLS + `linearmodels.IV2SLS` - [Stata .do files](https://github.com/tkay305/iv-regression-institutions-growth) — original `ivregress 2sls`, Sargan, Cragg-Donald, LIML - [Full Report](https://github.com/tkay305/iv-regression-institutions-growth/blob/main/docs/thesis_report.docx) ## Key References Acemoglu, D., Johnson, S. & Robinson, J. (2001). *The Colonial Origins of Comparative Development.* NBER WP 7771. Glaeser, E. et al. (2004). *Do Institutions Cause Growth?* NBER WP 10568. Rodrik, D., Subramanian, A. & Trebbi, F. (2004). *Institutions Rule.* Journal of Economic Growth, 9(2), 131–165.