The Colonial Origins of Comparative Development:

An Empirical Investigation Abstract

Demetria Parisi,
Francesca Ventimiglia,
Sila Sahin.

A summary of the paper by Daron Acemoglu, Simon Johnson, and James A. Robinson

Abstract

How can the vast income disparities across the different countries of the world be explained? What are the causes of such phenomenon? In the following document, we will reproduce the work of Acemoglu D. et al. and estimate how the differences in institutions affect the economic performance of different countries measured by income (GDP) per capita. For this purpose, the authors use the disparity in mortality rates in countries colonized by Europeans as an instrumental variable to identify the exogenous variation in current institutions across countries. Indeed, the diverse colonization policies adopted by European colonizers resulted in differing institution types which persist to the present day. In countries where the mortality rate was high, for instance, the colonizers preferred to set up extractive institutions instead of settling. Hence, a two-stage least squares model is applied in which the differences in mortality rates are utilized as a source of exogenous variation in institution quality. Former colonies with high quality institutions are expected to have a better current economic performance as compared to the ones with merely extractive or low quality institutions since good institutions lead to higher investment in capital and more efficient use of resources. These conditions, in turn, are presumed to produce a higher level of income. The reproduced paper seeks to establish this relationship.

Research question

Using the instrumental variable approach with mortality rates in European colonies as an instrumental variable, what is the estimated effect of current quality of institutions on income per capita and therefore economic performance? In this context, our research question arises as, “What are the fundamental causes of the large differences in income per capita across countries?”.

Motivation

The Colonial Origins of Comparative Development: An Empirical Investigation by Acemoglu et al. is one of the most essential and influential scientific works in its field. It has been subject to numerous follow-up analysis and the basis of further research. Our motivation behind choosing this work was to illustrate a significant economic paper using the python program. An evaluation of the effect of institutions on economic performance impacts current policy decisions, which could possibly yield to an improvement in the level of income per capita in countries struggling with their economic performance. We try to establish this coherence by showing that improving institutions in low income countries will benefit the region's GDP per capita by a large and statistically significant amount. Moreover, this study allows to further research on the most appropriate ways to improve institutions by correcting for specific aspects such as property rights enforcement, rule of law and other institutional features.

Assumptions

The model we are about to develop is based on three main assumptions:

The colonization policies implemented by Europeans were numerous. However, we will focus on the extreme cases: on one hand, we find the “extractive” strategy, in which institutions were mainly directed towards the transfer of colonies’ resources to the colonizers without accounting for protection of property rights or protection against government expropriation; on the other hand, Europeans who decided to settle in the colonies created the so-called “Neo-Europes”, whose institutions were inspired by the European model.
The above mentioned policies were influenced by the mortality and disease rates of the colonies. In colonies with a high mortality rate, colonists implemented the “extractive” strategy, whereas colonizers were settling in the colonies and opting for a “Neo-Europes” strategy when the chances of death by yellow fever or malaria were low.
The colonial state and the types of institution prevailed after independence up to current time.

To determine the variable current institutions, the authors use the protection against risk of expropriation index from Political Risk Services.

Method

The choice of using an instrumental variable approach stems from the idea that early institutions can be related to a number of variables which could influence the estimated effect. Countries may differ in their level of income per capita for a variety of reasons, including cultural and geographical factors. By instrumentalizing mortality rates, we ensure that variation in GDP across countries results solely from the countries' institutional characteristics. The idea behind the presented model can be summarized as follows: the settlers' mortality rates influence the settlement decision and the colonization strategy; the implemented policy will then result in low or high quality of early institutions, which will be reflected in current institutions and economic performance. The first stage of the model will regress the level of current institutions, which is the treatment variable, on settler mortality rates, the instrumental variable. In the second stage, the outcome of interest, here the economic performance, will be regressed on the resulting exogenous variation in the level of current institutions.

Descriptive Statistics

Before conducting the regression analysis, we take a look at the data and its properties. The sample contains data for every country in the world, comprising information on settler mortality, protection against expropriation risk, and PPP adjusted income (GDP) per capita in 1995. In the paper, the authors base their analysis on a subsample limited to 64 ex-colonies. Since we do not have access to detailed information concerning the subsample, we conduct our regression analysis on the given worldwide sample. The fact that the sample used by Acemoglu et al. is drawn from the data that we utilize, promises valid regression results approximating the findings of the paper. The variable describing the protection against expropriation risk serves to pick up institutional differences in the various countries. The graphical illustration of the variable suggests that the majority of countries tend to have higher quality institutions with more property rights. Regarding the income levels across countries according to the World Bank for the year 1995, the PPP adjusted GDP distribution nearly follows a normal distribution with most countries having an average income level, few countries having an extremely low GDP and some countries having an extremely high GDP level. Note that further information on the data and its sources can be found in Appendix Table A1, a table retrieved from the original paper.

In [2]:

import math
import numpy as np
import pandas as pd
from scipy import arange, optimize
import matplotlib.pyplot as plt
%matplotlib inline

In [3]:

Data_descriptive_stats = pd.read_stata ('descriptive-statistics.dta')

In [4]:

Data_descriptive_stats

Out[4]:

	shortnam	euro1900	excolony	avexpr	logpgp95	cons1	cons90	democ00a	cons00a	extmort4	logem4	loghjypl	baseco
0	AFG	0.000000	1.0	NaN	NaN	1.0	2.0	1.0	1.0	93.699997	4.540098	NaN	NaN
1	AGO	8.000000	1.0	5.363636	7.770645	3.0	3.0	0.0	1.0	280.000000	5.634789	-3.411248	1.0
2	ARE	0.000000	1.0	7.181818	9.804219	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
3	ARG	60.000004	1.0	6.386364	9.133459	1.0	6.0	3.0	3.0	68.900002	4.232656	-0.872274	1.0
4	ARM	0.000000	0.0	NaN	7.682482	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
5	AUS	98.000000	1.0	9.318182	9.897972	7.0	7.0	10.0	7.0	8.550000	2.145931	-0.170788	1.0
6	AUT	100.000000	0.0	9.727273	9.974877	NaN	NaN	NaN	NaN	NaN	NaN	-0.343900	NaN
7	AZE	0.000000	0.0	NaN	7.306531	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
8	BDI	0.000000	1.0	NaN	6.565265	5.0	1.0	0.0	1.0	280.000000	5.634789	-3.506558	NaN
9	BEL	100.000000	0.0	9.681818	9.992871	NaN	NaN	NaN	NaN	NaN	NaN	-0.179127	NaN
10	BEN	0.000000	1.0	NaN	7.090077	3.0	1.0	0.0	1.0	266.519989	5.585449	-2.830218	NaN
11	BFA	0.000000	1.0	4.454545	6.845880	3.0	1.0	0.0	1.0	280.000000	5.634789	-3.540459	1.0
12	BGD	0.000000	1.0	5.136364	6.877296	7.0	2.0	0.0	1.0	71.410004	4.268438	-2.063568	1.0
13	BGR	100.000000	0.0	8.909091	8.457443	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
14	BHR	0.000000	1.0	8.000000	9.685953	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
15	BHS	10.000000	1.0	7.500000	9.285448	NaN	NaN	NaN	NaN	85.000000	4.442651	NaN	1.0
16	BIH	100.000000	0.0	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
17	BLR	100.000000	0.0	NaN	8.340456	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
18	BLZ	20.000000	1.0	NaN	8.377932	NaN	NaN	NaN	1.0	163.300003	5.095589	NaN	NaN
19	BOL	30.000002	1.0	5.636364	7.926602	3.0	7.0	4.0	3.0	71.000000	4.262680	-1.966113	1.0
20	BRA	40.000000	1.0	7.909091	8.727454	1.0	7.0	1.0	3.0	71.000000	4.262680	-1.142564	1.0
21	BRB	20.000000	1.0	NaN	9.266721	NaN	NaN	NaN	1.0	85.000000	4.442651	-0.906340	NaN
22	BTN	0.000000	1.0	NaN	NaN	1.0	2.0	0.0	1.0	NaN	NaN	NaN	NaN
23	BWA	0.000000	1.0	7.727273	8.855093	7.0	7.0	0.0	1.0	NaN	NaN	-2.364460	NaN
24	CAF	0.000000	1.0	NaN	7.192934	1.0	1.0	0.0	1.0	280.000000	5.634789	-3.411248	NaN
25	CAN	99.000000	1.0	9.727273	9.986449	7.0	7.0	9.0	7.0	16.100000	2.778819	-0.060812	1.0
26	CHE	100.000000	0.0	10.000000	10.120211	NaN	NaN	NaN	NaN	NaN	NaN	-0.134675	NaN
27	CHL	50.000000	1.0	7.818182	9.336092	1.0	7.0	5.0	7.0	68.900002	4.232656	-1.335601	1.0
28	CHN	0.000000	0.0	7.772727	7.886081	NaN	NaN	NaN	NaN	118.000000	4.770685	-2.813411	NaN
29	CIV	0.000000	1.0	7.000000	7.444249	1.0	2.0	0.0	1.0	668.000000	6.504288	-2.333044	1.0
...	...	...	...	...	...	...	...	...	...	...	...	...	...
133	STP	100.000000	1.0	8.409091	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
134	SUR	1.000000	NaN	4.681818	8.010000	NaN	NaN	NaN	1.0	32.180000	3.471345	-1.370421	NaN
135	SVK	100.000000	0.0	NaN	8.852236	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
136	SVN	NaN	0.0	NaN	9.300181	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
137	SWE	100.000000	0.0	9.500000	9.866304	NaN	NaN	NaN	NaN	NaN	NaN	-0.239527	NaN
138	SWZ	0.000000	1.0	NaN	8.107720	2.0	1.0	0.0	1.0	NaN	NaN	-1.807889	NaN
139	SYR	0.000000	1.0	5.818182	8.029433	NaN	NaN	NaN	NaN	NaN	NaN	-0.825536	NaN
140	TCD	0.000000	1.0	NaN	6.835185	1.0	1.0	0.0	1.0	280.000000	5.634789	-3.442019	NaN
141	TGO	0.000000	1.0	6.909091	7.222566	3.0	1.0	0.0	1.0	668.000000	6.504288	-3.218876	1.0
142	THA	0.000000	0.0	7.636364	8.774931	NaN	NaN	NaN	NaN	140.000000	4.941642	-1.851509	NaN
143	TJK	NaN	0.0	NaN	6.887553	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
144	TKM	NaN	0.0	NaN	7.640123	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
145	TTO	40.000000	1.0	7.454545	8.768730	7.0	7.0	0.0	1.0	85.000000	4.442651	-0.697155	1.0
146	TUN	3.000000	1.0	6.454545	8.482602	1.0	3.0	0.0	1.0	63.000000	4.143135	-1.527858	1.0
147	TUR	0.000000	0.0	7.454545	8.641179	NaN	NaN	NaN	NaN	NaN	NaN	-1.523260	NaN
148	TWN	0.000000	0.0	9.227273	NaN	NaN	NaN	NaN	NaN	NaN	NaN	-0.809681	NaN
149	TZA	0.000000	1.0	6.636364	6.253829	3.0	3.0	0.0	1.0	145.000000	5.634789	-3.442019	1.0
150	UGA	0.000000	1.0	4.454545	6.966024	7.0	3.0	0.0	1.0	280.000000	5.634789	-3.442019	1.0
151	UKR	100.000000	0.0	NaN	7.811974	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
152	URY	60.000004	1.0	7.000000	9.031214	1.0	3.0	1.0	1.0	71.000000	4.262680	-1.078810	1.0
153	USA	87.500000	1.0	10.000000	10.215740	7.0	7.0	10.0	7.0	15.000000	2.708050	0.000000	1.0
154	UZB	NaN	0.0	NaN	7.807917	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
155	VEN	20.000000	1.0	7.136364	9.071078	1.0	3.0	1.0	3.0	78.099998	4.357990	-0.703197	1.0
156	VNM	0.000000	1.0	6.409091	7.279319	1.0	3.0	0.0	1.0	140.000000	4.941642	NaN	1.0
157	YEM	0.000000	1.0	6.363636	6.646390	NaN	NaN	NaN	NaN	NaN	NaN	-1.551169	NaN
158	YUG	100.000000	0.0	6.318182	NaN	NaN	NaN	NaN	NaN	NaN	NaN	-1.203973	NaN
159	ZAF	22.000000	1.0	6.863636	8.885994	3.0	7.0	3.0	3.0	15.500000	2.740840	-1.386294	1.0
160	ZAR	8.000000	1.0	3.500000	6.866933	1.0	1.0	0.0	1.0	240.000000	5.480639	-3.411248	1.0
161	ZMB	3.000000	1.0	6.636364	6.813445	3.0	1.0	0.0	1.0	NaN	NaN	-2.975930	NaN
162	ZWE	7.200000	1.0	6.000000	7.696213	7.0	3.0	0.0	1.0	NaN	NaN	-2.733368	NaN

163 rows × 13 columns

In [5]:

import pylab
s = pd.Series(Data_descriptive_stats.avexpr)
p = s.plot(kind='hist', color='orange')
pylab.rc("axes", linewidth=1.0)
pylab.rc("lines", markeredgewidth=1.0) 
pylab.xticks(fontsize=10)
pylab.yticks(fontsize=10)
plt.xlabel ('Avg. Protection against Expropriation Risk', fontsize=11)
plt.ylabel ('Frequency', fontsize=11)
plt.title ('Quality of Institutions', fontsize=12)
plt.show()

In [6]:

s = pd.Series(Data_descriptive_stats.logpgp95)
p = s.plot(kind='hist', color='orange')
pylab.rc("axes", linewidth=1.0)
pylab.rc("lines", markeredgewidth=1.0) 
pylab.xticks(fontsize=10)
pylab.yticks(fontsize=10)
plt.xlabel ('Log PPP GDP', fontsize=11)
plt.ylabel ('Frequency', fontsize=11)
plt.title ('GDP Distribution in 1995 (World Bank)', fontsize=12)
plt.show()

OLS Regression

Before applying the instrumental variable approach, the log per capita income is first regressed on the quality of institutions represented by the protection against expropriation variable in a ordinary least-squares regression. Additionally, we control for several factors such as geographical region and conditions. In mathematical terms, we evaluate $$ log~y_{i} = \mu + \alpha R_{i} + X^{'}_{i} \gamma + \epsilon_{i} $$ where:

$ log~y_{i} $ is the logarithm of the GDP per capita ( logpgp95 ) between 1975 and 1995;
$ R_{i} $ is the average protection against expropriation ( avexpr ) between 1985 and 1995;
$ X^{'}_{i} $ is a vector of covariates that Acemoglu controls for, e.g. africa, lat_abst , asia , other ;
$\epsilon_{i} $ is the random error term.

The results on our coefficient of interest, $ \alpha $, indicate that there is, in fact, a positive and statistically significant correlation between institutional quality and GDP. However, we cannot infer any causal relationship from this result: are rich countries the ones with better institutions or are good institutions the key to a higher per capita GDP? Moreover, geographical characteristics seem to influence the GDP as well. For instance, the regression results show that the GDP of a country is lower when said country is located in Africa or Asia.

In [7]:

Ols_data = pd.read_stata ('OLS.dta')

In [8]:

Ols_data

Out[8]:

	shortnam	africa	lat_abst	avexpr	logpgp95	other	asia	loghjypl	baseco
0	AFG	0.0	0.366667	NaN	NaN	0.0	1.0	NaN	NaN
1	AGO	1.0	0.136667	5.363636	7.770645	0.0	0.0	-3.411248	1.0
2	ARE	0.0	0.266667	7.181818	9.804219	0.0	1.0	NaN	NaN
3	ARG	0.0	0.377778	6.386364	9.133459	0.0	0.0	-0.872274	1.0
4	ARM	0.0	0.444444	NaN	7.682482	0.0	1.0	NaN	NaN
5	AUS	0.0	0.300000	9.318182	9.897972	1.0	0.0	-0.170788	1.0
6	AUT	0.0	0.524444	9.727273	9.974877	0.0	0.0	-0.343900	NaN
7	AZE	0.0	0.447778	NaN	7.306531	0.0	1.0	NaN	NaN
8	BDI	1.0	0.036667	NaN	6.565265	0.0	0.0	-3.506558	NaN
9	BEL	0.0	0.561111	9.681818	9.992871	0.0	0.0	-0.179127	NaN
10	BEN	1.0	0.103333	NaN	7.090077	0.0	0.0	-2.830218	NaN
11	BFA	1.0	0.144444	4.454545	6.845880	0.0	0.0	-3.540459	1.0
12	BGD	0.0	0.266667	5.136364	6.877296	0.0	1.0	-2.063568	1.0
13	BGR	0.0	0.477778	8.909091	8.457443	0.0	0.0	NaN	NaN
14	BHR	0.0	0.288889	8.000000	9.685953	0.0	1.0	NaN	NaN
15	BHS	0.0	0.268333	7.500000	9.285448	0.0	0.0	NaN	1.0
16	BIH	0.0	0.488889	NaN	NaN	0.0	0.0	NaN	NaN
17	BLR	0.0	0.588889	NaN	8.340456	0.0	0.0	NaN	NaN
18	BLZ	0.0	0.190556	NaN	8.377932	0.0	0.0	NaN	NaN
19	BOL	0.0	0.188889	5.636364	7.926602	0.0	0.0	-1.966113	1.0
20	BRA	0.0	0.111111	7.909091	8.727454	0.0	0.0	-1.142564	1.0
21	BRB	0.0	0.145556	NaN	9.266721	0.0	0.0	-0.906340	NaN
22	BTN	0.0	0.303333	NaN	NaN	0.0	0.0	NaN	NaN
23	BWA	1.0	0.244444	7.727273	8.855093	0.0	0.0	-2.364460	NaN
24	CAF	1.0	0.077778	NaN	7.192934	0.0	0.0	-3.411248	NaN
25	CAN	0.0	0.666667	9.727273	9.986449	0.0	0.0	-0.060812	1.0
26	CHE	0.0	0.522222	10.000000	10.120211	0.0	0.0	-0.134675	NaN
27	CHL	0.0	0.333333	7.818182	9.336092	0.0	0.0	-1.335601	1.0
28	CHN	0.0	0.388889	7.772727	7.886081	0.0	1.0	-2.813411	NaN
29	CIV	1.0	0.088889	7.000000	7.444249	0.0	0.0	-2.333044	1.0
...	...	...	...	...	...	...	...	...	...
133	STP	0.0	0.011111	8.409091	NaN	0.0	0.0	NaN	NaN
134	SUR	0.0	0.044444	4.681818	8.010000	0.0	0.0	-1.370421	NaN
135	SVK	0.0	0.537778	NaN	8.852236	0.0	0.0	NaN	NaN
136	SVN	0.0	0.511111	NaN	9.300181	0.0	0.0	NaN	NaN
137	SWE	0.0	0.688889	9.500000	9.866304	0.0	0.0	-0.239527	NaN
138	SWZ	1.0	0.292222	NaN	8.107720	0.0	0.0	-1.807889	NaN
139	SYR	0.0	0.388889	5.818182	8.029433	0.0	1.0	-0.825536	NaN
140	TCD	1.0	0.166667	NaN	6.835185	0.0	0.0	-3.442019	NaN
141	TGO	1.0	0.088889	6.909091	7.222566	0.0	0.0	-3.218876	1.0
142	THA	0.0	0.166667	7.636364	8.774931	0.0	1.0	-1.851509	NaN
143	TJK	0.0	0.433333	NaN	6.887553	0.0	1.0	NaN	NaN
144	TKM	0.0	0.444444	NaN	7.640123	0.0	1.0	NaN	NaN
145	TTO	0.0	0.122222	7.454545	8.768730	0.0	0.0	-0.697155	1.0
146	TUN	1.0	0.377778	6.454545	8.482602	0.0	0.0	-1.527858	1.0
147	TUR	0.0	0.433333	7.454545	8.641179	0.0	1.0	-1.523260	NaN
148	TWN	0.0	NaN	9.227273	NaN	0.0	1.0	-0.809681	NaN
149	TZA	1.0	0.066667	6.636364	6.253829	0.0	0.0	-3.442019	1.0
150	UGA	1.0	0.011111	4.454545	6.966024	0.0	0.0	-3.442019	1.0
151	UKR	0.0	0.544444	NaN	7.811974	0.0	0.0	NaN	NaN
152	URY	0.0	0.366667	7.000000	9.031214	0.0	0.0	-1.078810	1.0
153	USA	0.0	0.422222	10.000000	10.215740	0.0	0.0	0.000000	1.0
154	UZB	0.0	0.455556	NaN	7.807917	0.0	1.0	NaN	NaN
155	VEN	0.0	0.088889	7.136364	9.071078	0.0	0.0	-0.703197	1.0
156	VNM	0.0	0.177778	6.409091	7.279319	0.0	1.0	NaN	1.0
157	YEM	0.0	0.166667	6.363636	6.646390	0.0	1.0	-1.551169	NaN
158	YUG	0.0	0.488889	6.318182	NaN	0.0	0.0	-1.203973	NaN
159	ZAF	1.0	0.322222	6.863636	8.885994	0.0	0.0	-1.386294	1.0
160	ZAR	1.0	0.000000	3.500000	6.866933	0.0	0.0	-3.411248	1.0
161	ZMB	1.0	0.166667	6.636364	6.813445	0.0	0.0	-2.975930	NaN
162	ZWE	1.0	0.222222	6.000000	7.696213	0.0	0.0	-2.733368	NaN

163 rows × 9 columns

In [9]:

import statsmodels.api as sm
import statsmodels.formula.api as smf

In [10]:

results = smf.ols('logpgp95 ~ avexpr + africa + asia + lat_abst + other', data=Ols_data).fit()

In [11]:

results.summary()

Out[11]:

OLS Regression Results
Dep. Variable:	logpgp95	R-squared:	0.715
Model:	OLS	Adj. R-squared:	0.702
Method:	Least Squares	F-statistic:	52.74
Date:	Tue, 31 Jan 2017	Prob (F-statistic):	4.18e-27
Time:	18:22:16	Log-Likelihood:	-102.45
No. Observations:	111	AIC:	216.9
Df Residuals:	105	BIC:	233.2
Df Model:	5
Covariance Type:	nonrobust

	coef	std err	t	P>\|t\|	[95.0% Conf. Int.]
Intercept	5.8511	0.340	17.230	0.000	5.178 6.524
avexpr	0.3896	0.051	7.691	0.000	0.289 0.490
africa	-0.9164	0.166	-5.511	0.000	-1.246 -0.587
asia	-0.1531	0.155	-0.989	0.325	-0.460 0.154
lat_abst	0.3326	0.445	0.747	0.457	-0.551 1.216
other	0.3035	0.375	0.810	0.420	-0.440 1.047

Omnibus:	4.342	Durbin-Watson:	1.865
Prob(Omnibus):	0.114	Jarque-Bera (JB):	3.936
Skew:	-0.457	Prob(JB):	0.140
Kurtosis:	3.126	Cond. No.	58.2

First-stage

Now, we turn to the analytical method that promises to yield valid insights regarding the research question. In the first stage of the instrumental variable (IV) analysis, the treatment variable is regressed on the instrumental variable. Here, the selected instrumental variable are the mortality rates faced by European settlers in colonial times since they are easy to measure and arguably exogenous - a necessary condition in the isolation of the effect the treatment variable has on the outcome variable. In particular we want to regress $$ R_i = \zeta + \beta log M_i + X^{'}_{i} \delta + u_i $$ where $log M_i$ is the logaritm of European settlers' mortality (logem4) and $u_{i} $ is the random error term. Note that $log M_i$ is the instrument used to predict the average protection against expropriation risk which will be employed in the second stage to calculate our coefficient of interest, $\alpha$. When regressing the institution quality on the colonists' mortality rates between the XVII and XIX century, we obtain a strong first stage, meaning that there is a significantly strong negative correlation between the two variables (taken as logs). In particular, a higher settler mortality rate is correlated with a lower institution quality. As a result, the pattern in the relationship between the two factors seems to be following the path theorised by Acemoglu, Johnson and Robinson:

N.	Correlation
1.	`potential settler mortality rates` major determinant in `settlements`
2.	`settlements` major determinant in `institutions in 1900`
3.	`institutions in 1900` major determinant in `institutions today`

In [12]:

Data_first_stage = pd.read_stata ('1st-stage.dta')

In [13]:

Data_first_stage

Out[13]:

	lat_abst	euro1900	excolony	avexpr	logpgp95	cons1	indtime	democ00a	cons00a	extmort4	logem4
0	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
1	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
2	NaN	28.000000	0.0	5.000000	NaN	3.0	154.0	1.0	3.0	78.099998	4.357990
3	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
4	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
5	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
6	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
7	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
8	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
9	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
10	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
11	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
12	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
13	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
14	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
15	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
16	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
17	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
18	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
19	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
20	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
21	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
22	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
23	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
24	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
25	NaN	NaN	NaN	4.431818	NaN	NaN	NaN	NaN	NaN	NaN	NaN
26	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
27	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
28	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
29	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
...	...	...	...	...	...	...	...	...	...	...	...
346	NaN	0.000000	0.0	9.227273	NaN	NaN	NaN	NaN	NaN	NaN	NaN
347	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
348	0.066667	0.000000	1.0	6.636364	6.253829	3.0	33.0	0.0	1.0	145.000000	5.634789
349	0.011111	0.000000	1.0	4.454545	6.966024	7.0	33.0	0.0	1.0	280.000000	5.634789
350	0.544444	100.000000	0.0	NaN	7.811974	NaN	NaN	NaN	NaN	NaN	NaN
351	0.366667	60.000004	1.0	7.000000	9.031214	1.0	165.0	1.0	1.0	71.000000	4.262680
352	0.422222	87.500000	1.0	10.000000	10.215740	7.0	195.0	10.0	7.0	15.000000	2.708050
353	0.455556	NaN	0.0	NaN	7.807917	NaN	NaN	NaN	NaN	NaN	NaN
354	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
355	0.146100	10.000000	1.0	NaN	8.318742	NaN	NaN	NaN	NaN	85.000000	NaN
356	0.088889	20.000000	1.0	7.136364	9.071078	1.0	165.0	1.0	3.0	78.099998	4.357990
357	NaN	NaN	0.0	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
358	0.177778	0.000000	1.0	6.409091	7.279319	1.0	41.0	0.0	1.0	140.000000	4.941642
359	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
360	NaN	NaN	0.0	NaN	8.140316	NaN	NaN	NaN	NaN	NaN	NaN
361	NaN	NaN	0.0	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
362	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
363	NaN	NaN	0.0	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
364	NaN	NaN	0.0	NaN	7.867105	NaN	NaN	NaN	NaN	NaN	NaN
365	0.166667	0.000000	1.0	6.363636	6.646390	NaN	NaN	NaN	NaN	NaN	NaN
366	0.488889	100.000000	0.0	6.318182	NaN	NaN	NaN	NaN	NaN	NaN	NaN
367	0.488889	100.000000	0.0	6.318182	NaN	NaN	NaN	NaN	NaN	NaN	NaN
368	0.322222	22.000000	1.0	6.863636	8.885994	3.0	139.0	3.0	3.0	15.500000	2.740840
369	0.000000	8.000000	1.0	3.500000	6.866933	1.0	35.0	0.0	1.0	240.000000	5.480639
370	0.166667	3.000000	1.0	6.636364	6.813445	3.0	31.0	0.0	1.0	NaN	NaN
371	0.222222	7.200000	1.0	6.000000	7.696213	7.0	72.0	0.0	1.0	NaN	NaN
372	0.222222	7.200000	1.0	NaN	NaN	7.0	72.0	0.0	1.0	NaN	NaN
373	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
374	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
375	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN

376 rows × 11 columns

In [14]:

First_stage = smf.ols('avexpr ~ logem4', data=Data_first_stage).fit()

In [15]:

First_stage.summary()

Out[15]:

OLS Regression Results
Dep. Variable:	avexpr	R-squared:	0.298
Model:	OLS	Adj. R-squared:	0.288
Method:	Least Squares	F-statistic:	30.99
Date:	Tue, 31 Jan 2017	Prob (F-statistic):	4.08e-07
Time:	18:22:17	Log-Likelihood:	-126.60
No. Observations:	75	AIC:	257.2
Df Residuals:	73	BIC:	261.8
Df Model:	1
Covariance Type:	nonrobust

	coef	std err	t	P>\|t\|	[95.0% Conf. Int.]
Intercept	9.4925	0.550	17.263	0.000	8.397 10.588
logem4	-0.6441	0.116	-5.567	0.000	-0.875 -0.413

Omnibus:	0.907	Durbin-Watson:	1.806
Prob(Omnibus):	0.635	Jarque-Bera (JB):	0.873
Skew:	-0.021	Prob(JB):	0.646
Kurtosis:	2.473	Cond. No.	17.8

Exclusion Restriction

In order for the instrumental variable regression method to be valid, the so-called exclusion restriction has to be fulfilled. Said condition requires the instrumental variable to influence the outcome of interest only via the treatment variable. It is fairly straightforward that the colonists' expected mortality rates between the seventeenth and nineteenth century (our instrument) are in no way directly related to nowadays economic performance, the outcome variable, in each specific country. In fact, the only concern that might arise relates to current diseases: in such case, the instrument's effect on GDP might be a mere reflection of the disease environment and not a consequence of the country's quality of institutions. However, it has been shown that the colonists' mortality rates were maily due to their lack of immunity against diseases such as malaria and yellow fever, immunity that had been developed by the indigenous population over the centuries. It is therefore unlikely that such diseases take a central role in causing some former colonies to be extremely poor. Hence,

"The advantage of our approach is that conditional on the variables that we already control for, settler mortality more than 100 years ago should have no effect on output today, other than through its effect on institutions."
Acemoglu et al., 2001

Second-stage

In the second stage of the IV regression analysis, the impact of institutions on income per capita is estimated. Since it is not possible to reproduce the instrumental variable regression command that the authors apply with the python software, we conduct the second-stage analysis using the OLS regression as before. In the original paper, Acemoglu et al. replicate the second stage of the analysis with the ordinary least-squares method as well as shown in Table 5. In particular, we estimate the following equation $$ log~y_{i} = \mu + \alpha \hat{R}_{i} + X^{'}_{i}\gamma + \epsilon_{i} $$ where

$\hat{R}_{i}$ is the value of average expropriation risk obtained in the first stage;
$\alpha$ is our coefficient of interest, i.e. it represents the causal relationship running between quality of institutions and per capita GDP;
$X^{'}_{i}$ is the matrix containing covariates that the authors control for, e.g. lat_abst, f_brit, f_french, sjlofr in the following table;
$\epsilon_{i} $ is the random error term.

The regression results match the ones of the paper and show a highly significant positive correlation between the type and quality of institutions and the level of income in the investigated countries (0.4761). It can be seen that the effect is even larger than estimated in the predictive OLS regression above (0.3896). These findings support the initial expectations in which more extensive and higher quality institutions lead to an overall better economic performance.

In [16]:

Data_IV_reg = pd.read_stata ('IV-REG-additional-controls.dta')

In [17]:

Data_IV_reg

Out[17]:

	shortnam	catho80	muslim80	lat_abst	no_cpm80	f_brit	f_french	avexpr	sjlofr	logpgp95	logem4	baseco
0	AFG	0.000000	99.300003	0.366667	0.699997	1.0	0.0	NaN	1.0	NaN	4.540098	NaN
1	AGO	68.699997	0.000000	0.136667	11.500004	0.0	0.0	5.363636	1.0	7.770645	5.634789	1.0
2	ARE	0.400000	94.900002	0.266667	4.399999	1.0	0.0	7.181818	0.0	9.804219	NaN	NaN
3	ARG	91.599998	0.200000	0.377778	5.500001	0.0	0.0	6.386364	1.0	9.133459	4.232656	1.0
4	ARM	0.000000	0.000000	0.444444	100.000000	0.0	0.0	NaN	0.0	7.682482	NaN	NaN
5	AUS	29.600000	0.200000	0.300000	46.700001	1.0	0.0	9.318182	0.0	9.897972	2.145931	1.0
6	AUT	88.800003	0.600000	0.524444	4.099997	0.0	0.0	9.727273	0.0	9.974877	NaN	NaN
7	AZE	0.000000	93.400002	0.447778	6.599998	0.0	0.0	NaN	0.0	7.306531	NaN	NaN
8	BDI	78.300003	0.900000	0.036667	15.899997	0.0	0.0	NaN	1.0	6.565265	5.634789	NaN
9	BEL	90.000000	1.100000	0.561111	8.500000	0.0	0.0	9.681818	1.0	9.992871	NaN	NaN
10	BEN	18.500000	15.200000	0.103333	63.500000	0.0	1.0	NaN	1.0	7.090077	5.585449	NaN
11	BFA	9.000000	43.000000	0.144444	46.400002	0.0	1.0	4.454545	1.0	6.845880	5.634789	1.0
12	BGD	0.200000	85.900002	0.266667	13.699999	1.0	0.0	5.136364	0.0	6.877296	4.268438	1.0
13	BGR	0.500000	10.600000	0.477778	88.500000	0.0	0.0	8.909091	0.0	8.457443	NaN	NaN
14	BHR	0.800000	95.000000	0.288889	3.300000	1.0	0.0	8.000000	0.0	9.685953	NaN	NaN
15	BHS	25.500000	0.000000	0.268333	27.300003	1.0	0.0	7.500000	0.0	9.285448	4.442651	1.0
16	BIH	15.000000	40.000000	0.488889	41.000000	0.0	0.0	NaN	0.0	NaN	NaN	NaN
17	BLR	14.000000	0.000000	0.588889	86.000000	0.0	0.0	NaN	0.0	8.340456	NaN	NaN
18	BLZ	66.800003	0.000000	0.190556	19.999996	1.0	0.0	NaN	0.0	8.377932	5.095589	NaN
19	BOL	92.500000	0.000000	0.188889	5.200000	0.0	0.0	5.636364	1.0	7.926602	4.262680	1.0
20	BRA	87.800003	0.100000	0.111111	8.099997	0.0	0.0	7.909091	1.0	8.727454	4.262680	1.0
21	BRB	5.900000	0.200000	0.145556	60.700001	1.0	0.0	NaN	0.0	9.266721	4.442651	NaN
22	BTN	0.000000	5.000000	0.303333	95.000000	1.0	0.0	NaN	0.0	NaN	NaN	NaN
23	BWA	9.400000	0.000000	0.244444	63.799999	1.0	0.0	7.727273	0.0	8.855093	NaN	NaN
24	CAF	33.099998	3.200000	0.077778	13.700002	0.0	1.0	NaN	1.0	7.192934	5.634789	NaN
25	CAN	46.599998	0.600000	0.666667	23.200001	1.0	0.0	9.727273	0.0	9.986449	2.778819	1.0
26	CHE	52.799999	0.300000	0.522222	3.700000	0.0	0.0	10.000000	0.0	10.120211	NaN	NaN
27	CHL	82.099998	0.000000	0.333333	16.000002	0.0	0.0	7.818182	1.0	9.336092	4.232656	1.0
28	CHN	0.000000	2.400000	0.388889	97.599998	0.0	0.0	7.772727	0.0	7.886081	4.770685	NaN
29	CIV	18.500000	24.000000	0.088889	52.799999	0.0	1.0	7.000000	1.0	7.444249	6.504288	1.0
...	...	...	...	...	...	...	...	...	...	...	...	...
133	STP	92.400002	0.000000	0.011111	5.399999	0.0	0.0	8.409091	1.0	NaN	NaN	NaN
134	SUR	36.000000	13.000000	0.044444	14.400002	0.0	0.0	4.681818	1.0	8.010000	3.471345	NaN
135	SVK	74.000000	0.000000	0.537778	17.600000	0.0	0.0	NaN	0.0	8.852236	NaN	NaN
136	SVN	71.400002	1.500000	0.511111	27.099998	0.0	0.0	NaN	0.0	9.300181	NaN	NaN
137	SWE	1.400000	0.100000	0.688889	30.099998	0.0	0.0	9.500000	0.0	9.866304	NaN	NaN
138	SWZ	10.800000	0.100000	0.292222	55.199997	1.0	0.0	NaN	0.0	8.107720	NaN	NaN
139	SYR	1.300000	89.599998	0.388889	8.900002	0.0	1.0	5.818182	1.0	8.029433	NaN	NaN
140	TCD	21.000000	44.000000	0.166667	23.400000	0.0	1.0	NaN	1.0	6.835185	5.634789	NaN
141	TGO	29.299999	17.000000	0.088889	47.599998	0.0	1.0	6.909091	1.0	7.222566	6.504288	1.0
142	THA	0.400000	3.900000	0.166667	95.500000	0.0	0.0	7.636364	0.0	8.774931	4.941642	NaN
143	TJK	0.000000	85.000000	0.433333	15.000000	0.0	0.0	NaN	0.0	6.887553	NaN	NaN
144	TKM	0.000000	87.000000	0.444444	13.000000	0.0	0.0	NaN	0.0	7.640123	NaN	NaN
145	TTO	35.799999	6.500000	0.122222	44.500000	1.0	0.0	7.454545	0.0	8.768730	4.442651	1.0
146	TUN	0.100000	99.400002	0.377778	0.499998	0.0	1.0	6.454545	1.0	8.482602	4.143135	1.0
147	TUR	0.100000	99.199997	0.433333	0.700003	0.0	0.0	7.454545	1.0	8.641179	NaN	NaN
148	TWN	NaN	NaN	NaN	NaN	NaN	NaN	9.227273	NaN	NaN	NaN	NaN
149	TZA	28.200001	32.500000	0.066667	28.099998	0.0	0.0	6.636364	0.0	6.253829	5.634789	1.0
150	UGA	49.599998	6.600000	0.011111	41.900002	1.0	0.0	4.454545	0.0	6.966024	5.634789	1.0
151	UKR	0.000000	0.000000	0.544444	100.000000	0.0	0.0	NaN	0.0	7.811974	NaN	NaN
152	URY	59.500000	0.000000	0.366667	38.599998	0.0	0.0	7.000000	1.0	9.031214	4.262680	1.0
153	USA	30.000000	0.800000	0.422222	25.600002	1.0	0.0	10.000000	0.0	10.215740	2.708050	1.0
154	UZB	0.000000	88.000000	0.455556	12.000000	0.0	0.0	NaN	0.0	7.807917	NaN	NaN
155	VEN	94.800003	0.000000	0.088889	4.199997	0.0	0.0	7.136364	1.0	9.071078	4.357990	1.0
156	VNM	3.900000	1.000000	0.177778	94.900002	0.0	1.0	6.409091	1.0	7.279319	4.941642	1.0
157	YEM	0.000000	99.500000	0.166667	0.400000	0.0	0.0	6.363636	1.0	6.646390	NaN	NaN
158	YUG	4.000000	19.000000	0.488889	76.000000	0.0	0.0	6.318182	0.0	NaN	NaN	NaN
159	ZAF	10.400000	1.300000	0.322222	49.299999	1.0	0.0	6.863636	0.0	8.885994	2.740840	1.0
160	ZAR	48.400002	1.400000	0.000000	21.199999	0.0	0.0	3.500000	1.0	6.866933	5.480639	1.0
161	ZMB	26.200001	0.300000	0.166667	41.599998	1.0	0.0	6.636364	0.0	6.813445	NaN	NaN
162	ZWE	14.400000	0.900000	0.222222	63.299999	1.0	0.0	6.000000	0.0	7.696213	NaN	NaN

163 rows × 12 columns

In [18]:

Second_stage = smf.ols('logpgp95 ~ avexpr + lat_abst + f_brit + f_french + sjlofr', data=Data_IV_reg).fit()

In [19]:

Second_stage.summary()

Out[19]:

OLS Regression Results
Dep. Variable:	logpgp95	R-squared:	0.670
Model:	OLS	Adj. R-squared:	0.654
Method:	Least Squares	F-statistic:	42.55
Date:	Tue, 31 Jan 2017	Prob (F-statistic):	9.33e-24
Time:	18:22:22	Log-Likelihood:	-110.70
No. Observations:	111	AIC:	233.4
Df Residuals:	105	BIC:	249.7
Df Model:	5
Covariance Type:	nonrobust

	coef	std err	t	P>\|t\|	[95.0% Conf. Int.]
Intercept	4.2941	0.404	10.628	0.000	3.493 5.095
avexpr	0.4761	0.056	8.498	0.000	0.365 0.587
lat_abst	1.2293	0.487	2.525	0.013	0.264 2.195
f_brit	0.3231	0.179	1.808	0.073	-0.031 0.677
f_french	-0.3784	0.202	-1.870	0.064	-0.780 0.023
sjlofr	0.6409	0.176	3.650	0.000	0.293 0.989

Omnibus:	8.072	Durbin-Watson:	1.685
Prob(Omnibus):	0.018	Jarque-Bera (JB):	7.916
Skew:	-0.643	Prob(JB):	0.0191
Kurtosis:	3.239	Cond. No.	59.3

Conclusion

In the presented work, we reproduced the findings of Acemoglu et al. concerning the interdependence between institutions and economic performance of former European colonies. In line with the methodoloy of the original paper, we utilized the two-stage instrumental variable approach in order to isolate exogenous sources of variation in institutions to assess their impact on income per capita. This procedure is crucial in establishing a causal relationship between the two investigated variables of interest, which, in turn, allows for the inference of valid outcomes. The results we obtained in the regression analyses state a large and statistically significant effect of institutional structure on economic performance measured by GDP. Naturally, there are many questions that presented paper does not cover and which could be studied in further scientific research. However, the given findings present valuable insights into the determinants of a country's economic efficiency and therefore wealth.

Tables

Appendix Table A1 gives an in-depth description of data and sources in Acemoglu et al., 2001.

In [20]:

from IPython.display import Image, display
display(Image(filename='A1.png'))

Table 5 replicates the IV regression results in obtained by Acemoglu et al., 2001.

In [21]:

from IPython.display import Image, display
display(Image(filename='Table1.png'))