AREC 345: Global Poverty & Economic Development
Lecture 6:
The African Slave Trade
Professor: Pamela Jakiela
Department of Agricultural and Resource Economics
University of Maryland, College Park
The Long-Run Impacts of the Slave Trade
Exposure to the Slave Trade & Income in 2000
10
Scatter plot of GDP per capita by slave exports
Log GDP per capita in 2000
6
7
8
9
Mauritius
Equatorial Guinea
Seychelles
Tunisia
Botswana
Namibia
Morocco
Swaziland
South Africa
Egypt
Gabon
Algeria
Libya
Cape Verde Islands
Lesotho
Congo
Senegal
Mozambique
Ivory Coast
BeninGhana
Nigeria
Cameroon
Kenya
Mauritania
Sudan
Gambia
Somalia
Burkina
Liberia
Mali Faso
Angola
Uganda
Madagascar
Guinea-Bissau
Malawi
Central African Republic Zambia
Ethiopia
Burundi
Togo
Guinea
Tanzania
Niger
Chad
Sierra Leone
Zimbabwe
Sao Tome & Principe
Djibouti
Rwanda
Comoros
5
Democratic Republic of Congo
-2
0
2
4
6
8
Log total slave exports normalized by land area
10
Data from Nunn (2008)
Dep. Var. = GDP
E (GDP) = 7.517 − 0.117 · slave exports
↔
Slave exports
Constant
OLS
(1)
−0.117∗∗∗
(0.025)
7.517∗∗∗
(0.126)
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 5
Exposure to the Slave Trade & Income in 2000
Use regression results to calculate t-statistic:
Regression results:
The formula for the t-statistic:
Dep. Var. = GDP
Slave exports
Constant
OLS
(1)
−0.117∗∗∗
(0.025)
7.517∗∗∗
(0.126)
t-stat=
b̂
standard error of b̂
=
=
Absolute value of t-statistic is greater than 2.58,
|t-stat| > 2.58 ⇒ statistically significant at the 99 percent level
|t-stat| > 1.96 ⇒ statistically significant at the 95 percent level
|t-stat| > 1.64 ⇒ statistically significant at the 90 percent level
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 6
Exposure to the Slave Trade & Income in 2000
E (log GDP per capita) = 7.517 − 0.117 · log slave exports per square km
Interpretation:
Country
South Africa
Uganda
Malawi
Nigeria
Slave
Exports
1.67
19.30
1062.97
2188.16
Log of
Slave Exports
0.51
2.96
6.97
7.69
Predicted
Log GDP
7.46
7.17
6.70
6.61
Predicted
GDP
$1,732.62
$1,300.75
$813.74
$747.82
Actual
GDP
$4,139
$788
$ 679
$1,156
Clearly, exposure to the slave trade isn’t the whole story!
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 7
How Do We Interpret Our Regression Results?
What linear regression tells us:
• Is there a relationship between GDP and exposure to the slave trade?
• Is the association statistically significant?
This doesn’t tell us whether the relationship is causal
Other reasons for a significant association:
• Reverse causality: changes in the dependent variable cause
changes in the independent variable
• Omitted variable bias: some other factor is causing changes in
both the dependent and the independent variables
Selection bias is a form of omitted variable bias
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 8
Adding Controls to a Regression Specification
“Many of the countries that have the lowest slave exports are either small
islands or North African countries, both of which tend to be richer than
other countries in Africa.”
Log GDP per capita in 2000
6
7
8
9
10
Scatter plot of GDP per capita by slave exports
Mauritius
Equatorial Guinea
Seychelles
Tunisia
Botswana
Namibia
South Africa
Egypt
Morocco
Swaziland
Gabon
Algeria
Libya
Cape Verde Islands
Lesotho
Congo
SenegalBenin
Mozambique
Ivory Coast
Ghana
Nigeria
Cameroon
Kenya
Mauritania
Sudan
Gambia
Somalia
Burkina
Liberia
Mali Faso
Angola
Uganda
Madagascar
Guinea-Bissau
Malawi
Central African Republic Zambia
Ethiopia
Burundi
Togo
Guinea
Tanzania
Niger
Chad
Sierra Leone
Zimbabwe
Sao Tome & Principe
Djibouti
Rwanda
Comoros
5
Democratic Republic of Congo
-2
0
2
4
6
8
Log total slave exports normalized by land area
North Africa
Island nation
10
Other
Data from Nunn (2008)
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 9
Adding Controls to a Regression Specification
A multivariate regression including a control for North Africa:
E (GDP) = a + b · slave exports + c · north africa
Log GDP per capita in 2000
6
7
8
9
10
Scatter plot of GDP per capita by slave exports
Mauritius
Equatorial Guinea
Seychelles
Tunisia
Botswana
Namibia
Morocco
Swaziland
South Africa
Egypt
Cape Verde Islands
Lesotho
Gabon
Algeria
Libya
Congo
SenegalBenin
Mozambique
Ivory Coast
Ghana
Nigeria
Cameroon
Kenya
Mauritania
Sudan
Gambia
Somalia
Burkina
Liberia
Mali Faso
Angola
Uganda
Madagascar
Guinea-Bissau
Malawi
Central African Republic Zambia
Ethiopia
Burundi
Togo
Guinea
Tanzania
Niger
Chad
Sierra Leone
Zimbabwe
Sao Tome & Principe
Djibouti
Rwanda
Comoros
5
Democratic Republic of Congo
-2
0
2
4
6
8
Log total slave exports normalized by land area
North Africa
10
Sub-Saharan Africa
Data from Nunn (2008)
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 10
Adding Controls to a Regression Specification
10
Scatter plot of GDP per capita by slave exports
Log GDP per capita in 2000
6
7
8
9
Mauritius
Equatorial Guinea
Seychelles
Botswana
Namibia
South Africa
Gabon
Swaziland
Congo
Tunisia
Cape Verde Islands
Lesotho
SenegalBenin
Mozambique
Ivory Coast
Ghana
Algeria
Nigeria
Cameroon
Libya
Kenya
Mauritania
Sudan
Gambia
Somalia
Burkina
Liberia
Mali Faso
Angola
Uganda
Madagascar
Guinea-Bissau
Malawi
Central African Republic Zambia
Ethiopia
Burundi
Togo
Guinea
Tanzania
Niger
Djibouti
Chad
Sierra Leone
Zimbabwe
Morocco
Sao Tome & Principe
Rwanda
Comoros
Egypt
5
Democratic Republic of Congo
-2
0
2
4
altexports
North Africa
6
8
10
Sub-Saharan Africa
Data from Nunn (2008)
Including a control is equivalent to subtracting off the average levels of x
and y from N. African nations, so that the means equal the rest of sample
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 11
Adding Controls to a Regression Specification
A multivariate regression including a control for North Africa:
E (GDP) = a + b · slave exports + c · north africa
Mauritius
Equatorial Guinea
Seychelles
Tunisia
Botswana
Namibia
South Africa
Egypt
Morocco
Swaziland
Gabon
Algeria
Libya
Cape Verde Islands
Lesotho
Congo
SenegalBenin
Mozambique
Ivory Coast
Ghana
Nigeria
Cameroon
Kenya
Mauritania
Sudan
Gambia
Somalia
Burkina
Liberia
Mali Faso
Angola
Uganda
Madagascar
Guinea-Bissau
Malawi
Central African Republic Zambia
Ethiopia
Burundi
Togo
Guinea
Tanzania
Niger
Chad
Sierra Leone
Zimbabwe
Sao Tome & Principe
Djibouti
Rwanda
Comoros
Log GDP per capita in 2000
6
8
9
7
Log GDP per capita in 2000
9
6
7
8
10
Scatter plot of GDP per capita by slave exports
10
Scatter plot of GDP per capita by slave exports
Mauritius
Equatorial Guinea
Seychelles
Botswana
Namibia
South Africa
Gabon
Swaziland
Congo
Tunisia
Cape Verde Islands
Lesotho
Zimbabwe
Morocco
Sao Tome & Principe
Egypt
Somalia
Liberia
Uganda
Central African Republic Zambia
Burundi
Niger
Djibouti
Rwanda
Comoros
5
Democratic Republic of Congo
5
Democratic Republic of Congo
SenegalBenin
Mozambique
Ivory Coast
Ghana
Algeria
Nigeria
Cameroon
Libya
Kenya
Mauritania
Sudan
Gambia
Burkina
Mali Faso
Angola
Madagascar
Guinea-Bissau
Malawi
Ethiopia
Togo
Guinea
Tanzania
Chad
Sierra Leone
-2
0
2
4
6
8
Log total slave exports normalized by land area
North Africa
Sub-Saharan Africa
10
Fitted values
Data from Nunn (2008)
AREC 345: Global Poverty & Economic Development
-2
0
North Africa
2
4
altexports
6
Sub-Saharan Africa
Data from Nunn (2008)
Lecture 6: The African Slave Trade, Slide 12
8
Fitted values
10
Adding Controls to a Regression Specification
A multivariate regression including control for North Africa, islands
E (GDP) = a + b · slave exports + c · north africa + d · island nation
Dep. Var.: Log GDP per Capita
Slave exports
OLS
(1)
−0.117∗∗∗
(0.025)
North Africa
Island nation
Constant
7.517∗∗∗
(0.126)
OLS
(2)
−0.100∗∗∗
(0.031)
0.415
(0.337)
0.169
(0.392)
7.397∗∗∗
(0.177)
Interpretation:
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 13
Did Underdevelopment Cause Slave Exports?
“An alternative explanation for the relationship is that societies that were
initially underdeveloped may have been more likely to engage in the slave
trades, and these same societies are still relatively underdeveloped today.”
How can we explore this possibility?
• No data on GDP per capita is available from the 1400s
• In primitive societies, population density is a proxy for income
Why is this the case?
• Nunn uses data on population density in 1400 to test whether
the least developed areas were the most impacted by slave trades
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 14
Did Underdevelopment Cause Slave Exports?
Slave exports by land area
0
2
4
6
8
10
Scatter plot of slave exports by historical population density
Ghana
Togo
Guinea-Bissau
Benin
Angola
Nigeria
Senegal
Gambia
Guinea
Ethiopia
Malawi
Sierra Leone
Mozambique
Mali
Burkina Faso
Tanzania
Chad
Sudan
Democratic Republic of Congo
Congo
Madagascar
Ivory Coast
Kenya
Cameroon
Gabon
Liberia
Somalia
Zambia
Algeria
Uganda
Niger
Mauritania
Libya
Central
African Republic
Zimbabwe
South Africa
Burundi
Egypt
Equatorial Guinea
-2
Namibia
Djibouti
Cape
Comoros
Mauritius
Sao
Seychelles
Botswana
Tome
Verde& Islands
Principe
-3
-2
Swaziland
Lesotho
Morocco
Tunisia
-1
0
1
2
Estimated population density in 1400
Rwanda
3
4
Data from Nunn (2008)
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 15
Consequences of the Slave Trade
Demographic consequences:
• Loss of working age population, increased dependency ratio
• Slower population growth, delayed urbanization (?)
Political consequences:
• Undermined pre-existing political structures (e.g. Kongo Kingdom)
• Empowered those willing to enslave others
• Laid foundations of extractive (rather than productive) institutions
Social consequences:
• Erosion of trust, decreased trade
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 16
Impacts on Trust, Ethnic Fractionalization
.25
Ethnic diversity
.5
.75
1
Scatter plot of ethnic diversity by slave exports
Uganda
Liberia
Madagascar
Congo
Democratic
Republic of Congo
Cameroon
Chad
Kenya
Nigeria
Central African Republic
Ivory Coast
Sierra Leone
Somalia
Guinea-Bissau
Libya
Benin
Angola
Gambia
Zambia Gabon
South Africa
Guinea
BurkinaEthiopia
Faso
Tanzania
Sudan
Togo
Senegal
Mozambique
Mali
Malawi
Ghana
Niger
Mauritania
Djibouti
Namibia
Morocco
Mauritius
Cape
Verde Islands
Botswana
Rwanda
Zimbabwe
Equatorial Guinea
Seychelles
0
Algeria
Burundi
Lesotho
Egypt
Swaziland
Tunisia
Comoros
-2
0
2
4
6
Slave exports by land area
8
10
Data from Nunn (2008)
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 17
Impacts on Trust, Ethnic Fractionalization
Atlantic Slave Trade
AREC 345: Global Poverty & Economic Development
Indian Ocean Slave Trade
Lecture 6: The African Slave Trade, Slide 18
Impacts on Trust, Ethnic Fractionalization
Nunn and Wantchekon (2011) estimate the relationship between
exposure to the slave trade, how much people trust others
E (Trust) = a + b · slave exports
Unit of observation is the individual:
• Data is from the Afrobarometer surveys: nationally-representative
surveys of African countries, conducted in local languages
• Independent variable: how much an individual’s ethnic group was
exposed to the slave trade (throughout the course of history)
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 19
Impacts on Trust, Ethnic Fractionalization
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 20
Long-Term Consequences of the Slave Trade
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 21
Study Guide: Key Terms
• Atlantic, Indian Ocean, Red Sea, and Trans-Saharan slave trades
• ethnicity vs. shipping records
• t-statistic
• omitted variable bias
• statistical significance
• control (in a regression)
• multivariate regression
AREC 345: Global Poverty & Economic Development
Lecture 6: The African Slave Trade, Slide 22