Ilovepdf Merged

Elena Meschi
INTERNATIONAL ECONOMICS AND

DEVELOPMENT STUDIES
a.a. 2017/2018
Lecture 1
Introduction
Outline of the lecture
 Preliminary information and organization of the

course
 Materials and textbook
 Exam
 Object of development economics
 Contents and tentative course plan
 Meaning of development and the development gap
Course information
3
 Time
 Monday 12:15 - 13:45, AULA MF1, Polo didattico San Basilio
 Tuesday 12:15 - 13:45, AULA 4, San Sebastiano
 Wednesday 08:45 - 10:15 AULA 2B, Polo didattico San Basilio
 Office hours
 Monday 16.00-18.00 (subject to variations: check on my
webpage or on moodle)
 Studio n.134, S.Giobbe I piano Dipartimento Economia plesso A
 Email: elena.meschi@unive.it
References
4
 Material:
 Textbook: Gerard Roland “Development Economics”, Pearson, 2014
 Only chapters discussed in class
 Slides of the lectures

 Reading list
 Detailed course outline, slides and reading list available in

Moodle.
 The outline may be updated and additional readings may

be posted on Moodle along the way
Assessment
5
 Final grades will be awarded by weighting

participation, home assignment and a final written test
as follows:
 Participation and home assignment 10 %
 Written exam 90 %
 The final written exam consists of theoretical questions,

analytical exercises, and discussion of empirical
results.
Services for students with Disabilities and Specific Learning Disorders
Students with disabilities and/or SLDs attending this course are invited to notify
the professor and contact the Disability and SLD Service (disabilita@unive.it)
for any related issues such as: course accommodations, note-taking and study
support, specialized equipment and adaptive technologies, in order to facilitate
course attendance and exam preparation.
For general information about the services offered by our University:

Disability and SLDs Service
disabilita@unive.it
www.unive.it/disabilita
041 234 7961
7
Development Economics
8
“Development Economics” tries to provide solutions to help poor

economies grow as a way to raise living standards and reduce
hunger and disease. Economic growth is required but structural
changes are also fundamental.
 Some questions for Development Economists:
Why do some countries grow rapidly and other countries grow slowly or not
at all?
What is the relationship between economic growth and economic
development?
What are the sources of development and how do we measure
development?
Which economic policies work best, at the regional, national, urban, and
village levels, to help people escape poverty, achieve higher living standards,
receive better health care and education, and live longer?
Development Economics
9
 Development economics:
 is a branch of economics which deals with economic aspects of the
development process in low income countries
 applies economic concepts and theories to gain a better
understanding of development process
 Studies causes of underdevelopment and identifies best policies
 Uses statistical tools of policy evaluation
 Increasing use of randomized experiments
 Nature of development economics:

 Traditional economics- efficient allocation of scarce resources
 Development economics- Role of values, attitudes, and
institutions
Contents of the course
10
 The Development Gap

 Poverty and Inequality
 Population Growth
 Economic Growth
 International Trade and Exchange Rates
 Institutions and Economic Development
 Delivering Healthcare in Developing Countries
 Delivering Education in Developing Countries
Today: outline
11
 Dimensions of development
 Facts about the development gap
 Evolution of the gap
 Stories of catch-up and decline
 References: Roland, chapter 1

The development gap
Figure 1.1a. The actual land mass of countries.
The development gap
Figure 1.1b. Countries size as proportional to population.
In economics, size matters.
The development gap
Figure 1.1c. Countries size as proportional to income.
Facts about the Development Gap
 Differences in economic development between the advanced economies
of the United States, Japan, and Western Europe and the poorer
economies of Africa, Asia, Latin America, and Eastern Europe.
 Differences can be measured in terms of income, life expectancy, health,
education, and level of urbanization.
 The development gap evolves over time. Currently, it is decreasing for
some countries (e.g. recently, China and India) while increasing for others
(e.g. Democratic Republic of the Congo).
 Economic development is not irreversible. Even some rich economies have
displayed protracted decline (Argentina).
 Poor economies tend to grow faster when they do grow, but experience
shows that protracted declines occur more frequently than for rich countries.
16
 Development multifaceted phenomenon

 Income
 Poverty
 Health
 Education
 Urbanization
Income gap
17
 The first basic indicator of development is real income

per capita.
 Gross Domestic Product (GDP)
 Market value of all final goods and services produced within a
country in a given period of time
 Value of output produced in a country. It is not a proper income
measure because it does not account for net foreign income, foreign
aid, and remittances.
Y=C+I+G+NX
 Gross National Income (GNI)
 is the market value of all final goods and services produced by
permanent residents of a country in a given period of time
GNI= GDP+ net factor income from abroad
Income gap
18
 GDP per capita needs to be expressed in a common

currency (usually the dollar) for comparison across countries.
 Market exchange rate
 Purchasing power parity (PPP) exchange rates
 Under PPP, exchange rates should adjust to equalize the price of a
common basket of goods and services across countries  PPP rates are
defined so that the same basket of goods in any two countries has the
same dollar value.
 Unlike market exchange rates, PPP rates take into account non-tradable
goods and services. The price of non-tradables is tied to local wages,
so they tend to be cheaper in developing nations.
Comparison of GNP
19
GNP Per Capita (US $)

Country Exchange rate PPP
UK 24,500 23,550
USA 34,260 34,260
Zimbabwe 480 2,590
Bangladesh 380 1,650
China 840 3,940
India 460 2,390
Sri Lanka 870 3,470
PPP: Big Mac Index
20
 Big Mac Index is published by The Economist as an informal way of measuring

the purchasing power parity (PPP) between two currencies and provides a test of
the extent to which market exchange rates result in goods costing the same in
different countries.
 Basket made of a single Big Mac as sold by McDonald, chosen because:
 available in many countries around the world
 relatively standardized product
 includes input costs from a wide range of sectors in the local economy, such as
agricultural commodities (beef, bread, lettuce, cheese), labor (blue and white collar),
advertising, rent and real estate costs, transportation
 The Big Mac PPP exchange rate between two countries is obtained by dividing
the price of a Big Mac in one country (in its currency) by the price of a Big Mac in
another country (in its currency). This value is then compared with the actual
exchange rate; if it is lower, then the first currency is under-valued (according to
PPP theory) compared with the second, and conversely, if it is higher, then the first
currency is over-valued.
21
PPP: Big Mac Index
22
The income gap
Figure 1.2. GDP Per Capita in 2010 (Purchasing Power Parity in Dollars)
The income gap
Classifying countries in terms of GDP per capita (on a PPP basis) shows great
differences. For example, in 2010:
Luxemburg is richest at $86,000. Democratic Republic of the Congo is
poorest at $350.
 ~ $50,000 include the United States, Norway, and Singapore. It also
includes some oil producing countries like Qatar and the United Arab Emirates.
 Richer Europe: GDP per capita between $20,000 and $40,000
Middle income countries, with GDP per capita between $5,000 and
$15,000 include Russia, Brazil, and Mexico.
 The poorest countries have less than $1,000 in GDP per capita. Including
Burundi, Liberia, Eritrea, and Mozambique.
Income gap
25
 Defining the Developing World

 World Bank Scheme- ranks countries on GNP/capita
 Low Income Countries (LIC)
 Lower Middle Income Countries (LMC)
 Upper Middle Income Countries (UMC)
 High Income Countries (HIC)
 GNI and GDP per capita data, and income-level

aggregations are available in the World Development
Indicators database at data.worldbank.org
Classification of Economies by Region and
Income, 2013
Classification of Economies by Region and
Income, 2013 (continued)
Classification of Economies by Region
and Income, 2013 (continued)
The poverty gap
29
 The Poverty Gap

The poverty headcount ratio measures the number of the world’s
population living on less than $2 a day (in a PPP basis).
This is the measure used by the World Bank to gauge “extreme poverty.”
According to this definition, 70% of the population in South Asia (Pakistan,
India, Bangladesh, and neighbors) lives in extreme poverty.
Likewise, nearly 70% of the population of Sub-Saharan Africa fits that
definition.
InEast Asia and the Pacific, 33% of the population lives below the
poverty line.
In total, this adds up to approximately 2.53 billion people. Almost half of
the world’s population lives in poverty.
The poverty gap
Poverty Headcount Ratio in 2008 at Poverty Line of $2 a Day in 2005 Prices (PPP basis).
The health gap
The Health Gap
 There are large differences in life expectancy and infant mortality between
developed and developing countries.
 Life expectancy measures the number of years a newborn infant would live if
health and living conditions at the time of its birth remained the same throughout
its life.
 Reflects the health conditions and quality of health care
 A child born in a developed country is usually expected to live longer than a child born
in a developing country.
 In general, life expectancy is highly positively correlated with income.
 There are some exceptions. Cuba (79) is an example of a poor country with high life
expectancy.
The health gap
Table 1.1. 2010 Life Expectancy (Years) at Birth in Regions of the World and
in Selected Countries.
The health gap
 The infant mortality rate measures the probability that a child will die
before reaching the age of 1.
 It is computed as the number of children dying before age 1 per 1,000
live births in the same year.
 It is negatively correlated with income.
 It is highest in Sub-Saharan Africa and South Asia.
 Sierra Leone has an infant mortality rate of 134 per 1,000 live births.
 By comparison, the United States has an infant mortality rate of 6.5 per 1,000
live births.
 Policy interventions can make a difference in lowering child mortality. For

example, Cuba has an infant mortality rate of 4.6, which is lower than
that of the United States.
Life Expectancy
The health gap
Figure 1.4. Infant Mortality Rates in 2010 (Per 1,000 Live Births)
Infant Mortality Rates
The education gap
37
The Education Gap

 Countries that invest in near-universal, quality education can realize high
productivity gains and economic growth.
 Many poor countries cannot afford a good educational system, contributing to
the perpetuation of the development gap.
 There have been improvements in primary school enrollment in DCs, but many
lag behind developed nations in terms of secondary school enrollment.
 Secondary school enrollment is pupils enrolled in secondary education divided by the
population in the age group. (Can be greater than 100%)
The UK has a secondary school enrollment rate of 178%, while Niger’s is only 12%.
South Korea is a success story in terms of economic growth and educational attainment.
Around 97% of South Koreans between the ages of 25 and 34 have achieved secondary
education.
The education gap
Figure 1.5. Secondary School Enrollment (Gross) in 2009, by Region
Years in School
The urbanization gap
40
The Urbanization Gap

 Urbanization rate measures proportion of population living in urban areas
(>200,000 people) as opposed to rural areas.
 Development drives urbanization as workers flow to industries and services
located in cities.
 Urbanization is increasing rapidly across the world. In 2010, the global
rate exceeded 50% for the first time.
InAfrica, Ethiopia, and Rwanda urbanization rates are below 20% while Djibouti
and Gabon have urbanization rates close to 90%.
In Asia, Nepal is the least urbanized at 18.2%. South Korea, on the other hand,
has an urbanization rate of 81%, higher than Japan’s 67%.
InLatin America, Guyana is the least urbanized at 28.5%. Brazil, Argentina, and
Venezuela have urbanization rates above 90%.
Figure 1.6. World Urbanization Rates, 2010
 The recent acceleration of urbanization is not necessarily due to
attractive higher levels of income in urban areas. Many people living is
large cities live in extreme poverty with little basic infrastructure (safe
water, electricity, gas, transportation, sewage).
 Most of the world’s largest cities are in the developing world. These
include: Mexico City (Mexico), Sao Paulo (Brazil), Mumbai (India),
Shanghai (China), Djakarta (Indonesia), Kolkata (India) and Cairo
(Egypt).
 There is a great need for public policy to address how to improve
housing, infrastructure, health and education in large cities of the
developing world.
 The development gap, as measured by income, poverty, health,
education and urbanization, is a “stylized fact” of the economics of
development. It raises the important questions in the economics of
development:
Why did some countries develop earlier than others?
 To frame development issues, we need a dynamic (over time) view of
economic development:
How is the development gap is evolving over time?
 Some countries have made great progress toward closing the gap, while for
others the gap has widened.
 Countries that are falling further behind rich countries usually experience
economic stagnation and even decline.
The evolving Development Gap
Differences in Economic Growth
Notably, the highest growth rates in GDP per capita in the last 3 decades have
been in East Asia, the Pacific and South Asia.
 By contrast, the lowest growth rates in GDP per capita in the last 3 decades
have been in Sub-Saharan Africa, the Middle East, North Africa, Latin America, and
the Caribbean.
 As shown in Figure 1.8 (see next slides) there has been wide variation between
countries.
 Between 1980 and 2010, China grew on average at over 8% a year.
 Botswana, India, Thailand, and Singapore grew at over 4% a year during the same
period.
 On the other hand, Liberia, Saudi Arabia, Cote D’Ivoire, Georgia, Niger, Moldova, Togo,
Gabon, Burundi, Venezuela, and Nicaragua experience economic contraction during the
period.
Figure 1.7. Average Annual Growth Rate (1980-2010) of GDP Per Capita (PPP)
in Constant 2005 Prices.
The evolving
Development Gap
Figure 1.8. Average Annual Growth

Rate (1980-2010) of GDP Per
Capita (PPP) in Constant 2005 Prices
for Selected Countries.
47
Population Growth
 Population growth contributes to GDP growth because a larger population
increases the labor force and thus should increase economic output.
 However, if output growth is slower than population growth, GDP per capita
falls.
 Noticeably, the world’s poorest region, Sub-Saharan Africa has had the
highest population growth.
 Likewise population growth has been high in the Middle East and North
Africa, South Asia, Latin America, and the Caribbean.
If this trend continues, the proportion of the world’s population living in
poverty will increase.
Figure 1.9. Total Population in 2010 as a Multiple of the Population in 1960.
50
Definition of Economic Development:
51
 World Bank in its 1991 WDR asserted that the “challenge

of development is to improve the quality of life.”
 The improved QOL involves higher incomes, better

education, higher standards of health and nutrition, less
poverty, a cleaner environment, more equality of
opportunities, greater individual freedom, and a richer
cultural life.
Definition of Economic Development
52
 “Development is a multi dimensional process involving

changes in social structures, popular attitudes, and
national institutions, as well as the acceleration of
economic growth, the reduction of inequality, and the
eradication of poverty.” (Todaro and Smith)
 Three Objectives of Development:
 Increase availability and distribution of basic goods
 Raise levels of living
 Expand range of social and economic choices available
to individuals
Human Development
53
 A standard definition of human development (1990

HDR):
 “[…] a process of enlarging people’s choices to live lives
they have reason to value… The most critical ones are to
lead a long and healthy life, to be knowledgeable and to
enjoy a decent standard of living.”
 A broader definition (2010 HDR):
 “Human development is the expansion of people’s
freedoms to live long, healthy and creative lives; to
advance other goals they have reason to value; and to
engage actively in shaping development equitably and
sustainably on a shared planet”
Human Development Index
54
 Emphasizes that outcomes for people and their

capabilities should be the ultimate criteria for
assessing the progress of a country, not economic
growth alone.
 Accounts for average achievements in
 life expectancy (proxy for leading a long and healthy
life)
 education (proxy for being knowledgeable) and
 income per capita (proxy for command over resources to
have a decent standard of living).
55
 Initiated in 1990 and undertaken by UNDP in its

annual series of HDRs.
 The HDI consists of three equally weighted
components:
 (1) “A long and healthy life” (Health)
 (2) “Knowledge” (Education)
 (3) “A decent standard of living” (Wealth)
 HDI= 1/3(Income index)+1/3(Life expectancy

index)+1/3(education index)
Each component of the HDI is measured in the

following way:
 Health
 Measured by life expectancy at birth.
 Education
 Measured as a combination of adult literacy (with
two-thirds weight) and gross enrollment (with one-
third weight).
 Wealth
 Measured by GDP per capita.
 Data: http://hdr.undp.org/en/data
Stories of Catch-up and Decline
 The development gap emerged because some countries
developed earlier than others.
 In historical perspective the “Industrial Revolution”:
• The British economy began to expand and change rapidly in the
late 18th century.
• The American economy and much of continental Europe started in
the 19th century.
• Germany and Japan started late, but caught up very rapidly.
 The record is not all good. Examples of rich countries that
subsequently declined include China, Argentina, and the
Ottoman Empire (Turkey).
The Historical Catch-Up of Japan
At end of the Tokugawa (Shogun) period, Japan was a feudal society and
remained closed to the outside world.
 In1867, the Meiji emperor implemented a comprehensive program of social and
political reforms designed to mimic the “institutions” of industrialized countries.
 The Japanese government made large infrastructure investments and promoted
industrialization. Business conglomerates emerged and transformed the economy.
At the beginning of the “Meiji Restoration,” Japanese income per capita was
estimated at less than 30% that of the U.S. and UK. By 1940 the ratio was over
40%.
Growth after WWII was even more striking. Japan saw rapid and consistent growth
based on high-quality, low-cost manufacturing.
By the 1980s, the Japanese economy overtook the UK, and reached 80% of the U.S. per
capita income.
Figure 1.10. Japan’s Per Capita Income as a Percentage of Levels in the United
Kingdom and the United States.
The Historical Catch-Up of Germany
 Germany was established as a unified country only after 1871.
 After unification, the government under Bismarck launched an industrialization
program.
 Unification helped development of larger markets, as tariffs between German
states were abolished.
 A new innovation, the universal bank lent money to firms and, also, held equity
in industrial enterprises. Allowed the financing of large investments that
produced economies of scale (steel, rail, chemicals, etc.) helping the German
economy grow rapidly.
 By the beginning of WWI, German income was around 80% of the UK level.
 After WWII the (West) German economy also recovered rapidly, surpassing
UK income per capita by the 1960s.
Figure 1.11. German Per Capita Income as a Percentage of Levels in the United
Kingdom.
 Alexander Gerschenkron: “the advantages of (economic) backwardness”.
Latecomers can achieve a faster process of industrialization that would
allow them to catch up to early industrializers.
 Adapt existing technologies used in richer countries
 Reach economies of scale to quickly reach the industrial frontier of the most
advanced technologies in various industrial sectors
 Growth is paced by large volumes of capital investment by both the private
and public sectors
 State policy should encourage private capital formation, building
infrastructure, assuring competition in domestic markets, and shielding “infant
industries” from foreign competition.
 The data on catch-up shows that the advantage does not always apply.
Economic Decline
 When we look at history, some of the currently poor countries and regions of
the world were once the richest.
 China was once the richest country in the world with higher living standards than
Europe until around the 15th century and higher than Japan until about the 19th
century. During the 19th and 20th century Chinese growth continually lagged and
did not start growing again until recently (1980s).
 After 1453, Eastern Europe was dominated by the Ottoman Empire. But the
Empire declined throughout the 19th century and eventually collapsed during
WWI.
 Argentina was one of the richest countries in early 20th century, mainly due to
high agricultural productivity. In 1900s Argentina income levels were 80% of the
U.S. By 2000, income levels declined to about 30% of U.S.
Figure 1.12. Estimates of GDP Per Capita in China and Europe in 1990
International Dollars.
Figure 1.13. GDP Per Capita of Argentina as a Percentage of Levels in the
United Kingdom and the United States.
 There are many other examples of former world powers that have declined
and become poor countries and/or regions of the world today.
 To date, there are no examples of prolonged economic decline in
industrialized countries. However, the period of industrialization is still short
(about 250 years).
 An important question is then, why do some wealthy countries begin to
decline and ultimately become poor?
 Undoubtedly, this question will become more important as the fear that
competition of developing countries, such as China and India, who will
undermine the economies of rich countries, increases over time.
Characteristics of the Developing World
 Common characteristics among developing in

comparison with developed countries:
1. Lower levels of living and productivity
2. Lower levels of human capital (health, education, skills)
3. Higher Levels of Inequality and Absolute Poverty
4. Higher Population Growth Rates
5. Greater Social Fractionalization
6. Larger Rural Populations but Rapid Rural-to-Urban Migration
7. Lower Levels of Industrialization and Manufactured Exports
8. Underdeveloped Financial and Other markets
9. Colonial Legacy and External Dependence
Elena Meschi
INTERNATIONAL ECONOMICS
AND DEVELOPMENT STUDIES
a.a. 2017/2018
Lectures 2 and 3
Poverty and Inequality
Outline of the lectures
 Poverty and Inequality

 Measure poverty across countries using different
approaches and explain how poverty has evolved over
time
 Measure income inequality within a country and between
nations
 Explain the relationship between income inequality and
economic growth across countries and over time.
 References: Roland, chapter 2

Measuring Poverty
Poverty is the most visible feature of developing countries

 Measuring Poverty
 Poverty measurements are usually based on survey data. A
survey is a sampling of data based on questionnaires asked
to a representative subset of a population.
 How to define poverty?
 Self assessment: ask people in local communities to identify who,
among them, are poor.
 Objective measures
 What indexes can be used?
Measuring Poverty
4
 Absolute poverty: is a condition where household

income is below a necessary level to maintain basic
living standards (food, shelter, housing). This
condition makes it possible to compare between
different countries and also over time.
 Relative poverty – A condition where household

income is a certain percentage below median
incomes.
Poverty Measurements
5
 Absolute Poverty:
 In absolute sense a person is poor if his income falls below a certain
minimum level which is required to satisfy minimum basic needs. The
problem faced in measuring poverty is to specify the minimum basic
needs or what is also called minimum living standard (MLS) that
should be defined as poverty line.
 Then a person having an income below that poverty line will be
described as poor in the absolute sense.
 Poverty line: income required to purchase 2,000 calories of food,
based on the food basket typically purchased by local households.
 Calorie counting determines the minimal local cost of 2,000 calories worth of
food per day (the average daily number of calories humans need) which
becomes the level of income below which people are considered poor.
6
 One dollar a day: Using the work of Chen and Ravallion, the World
Bank established an international extreme poverty line of $1 per
day. Updated to $1.90 in 2015 a day to account for inflation
 They used existing household consumption surveys and PPP rates to
compile most accurate and internationally comparable poverty
estimates across countries and across time.
 Cutoff determined by averaging out the poverty lines (the minimal
requirements necessary to afford minimal standards of food, clothing,
health care and shelter) in local currency of the 15 poorest countries.
 Data is found in the annual World Development Report and are used for
the Millennium Development Goals (MDG), global targets established at
the 2000 Millennium Summit.
7
 Relative Poverty
 A condition where household income is a certain
percentage below median incomes. For example, the
threshold for relative poverty could be set at 50% of
median incomes (or 60%)
 Sometimes we are interested in focusing on the poorest
segment (e.g. a fifth, or two-fifths) of the population;
these are the relatively poor.
 Relative poverty is useful for showing the percentage of
the population who have been relatively left behind
 Defined relative to the members of a society and,
therefore, differs across countries.
 Poverty headcount: number of people (Q) in the

population (N) earning below the poverty line.
 Poverty headcount ratio (incidence of poverty):

proportion of the population below the poverty line
(H = Q/N).
9
10
 These indexes do not reflect the degree of

poverty. For example, a family with income
1% percent below the poverty line is counted
the same as a family 50% below.
11
 Poverty Gap
 The poverty gap takes into account the degree of poverty,
or the distance from the poverty line.
𝑄
𝑖=1(𝑧 − 𝑌𝑖 )
𝑇𝑃𝐺 =
𝑧
 z is the absolute poverty line, Yi is income of person i
 or 𝑇𝑃𝐺 = 𝐻 ∗ (𝑧 − 𝑌)/𝑧
 Average poverty gap: 𝐴𝑃𝐺 = 𝑇𝑃𝐺/𝑁

 Where N is number of persons in the economy
12
 Table 2.1. Example of Poverty Headcount Ratio and Average Poverty

Gap.
 Take 3 countries, A and B, that have total pop 200 million each with 100
million below poverty line: poverty headcount ratio 50%

Poverty measurement
13
 Country A:
 Average income (𝑌):
[(20*0.25)+(80*0.75)]/100=(5+60)/100=0.65
 APG=H (z-𝑌)/z = 50%*(1-0.65)=17.5%
 Country B
 𝑌= [(80*0.25)+(20*0.75)]/100=20+15=0.35
 APG=50%*(1-0.35)=32.5%
 Country B’s poverty is more serious

14
Poverty gap
15
 Poverty gap
Incidence and evolution of Poverty
Table 2.2. The Poverty Headcount Ratio and the Average Poverty Gap for
Selected Countries.
Table 2.3. The Poverty Headcount and Headcount Ratio Based on $1.25 a Day.
18
 Progress on Extreme Poverty

 Clear progress on $1.25-a-day headcount
 There has been a large reduction in absolute (number of
persons), especially in East Asia.
 Absolute numbers of poors increased in South Asia, sub-
Saharan Africa and Europe and Central Asia.
 Transition from socialism resulted in -- hopefully temporary
increases -- in relative poverty in Europe and Central Asia.
 Less clear progress on $2.00-per-day headcount (see
next figure)
 Incidence of extreme poverty is uneven
19
Poverty over time
20
 The Millennium Development Goal of halving 1990

poverty by 2015 was achieved in 2010.
 More than 1 billion people have been lifted out of
extreme poverty since 1990.
 In 1990, nearly half of the population in the
developing regions lived on less than $1.25 a day.
This rate dropped to 14 per cent in 2015.
 At the global level more than 800 million people
are still living in extreme poverty.
Poverty over time
21
 The world's extremely poor people are distributed very

unevenly across regions and countries. The overwhelming
majority of people living on less than $1.25 a day reside in
two regions- Southern Asia and sub-Saharan Africa- and
they account for about 80 per cent of the global total of
extremely poor people.
 Nearly 60 per cent of the world's 1 billion extremely poor
people lived in just five countries in 2011: India, Nigeria,
China, Bangladesh and the Democratic Republic of the
Congo (ranked from high to low).
Poverty across regions and over time
22
Poverty across regions and over time
23
 Most think that poverty reduction is due the economic boom of the Asian giants
and that Africa lags behind. UNDP contends that the goal of cutting in half the
proportion of people in the developing world living on less than $1 a day by
2015 remains within reach. However, this achievement will be due largely to
extraordinary economic success in most of Asia.
 In contrast, previous estimates suggest that little progress was made in reducing
extreme poverty in sub-Saharan Africa. The World Bank concurs: In 1990, 28.3
percent of the people in low and middle-income countries lived on less than $1
a day. By 1999 the share had fallen to 21.6 percent, driven mainly by strong
growth in China and India (. . . ) In Sub-Saharan, where the GDP per capita fell
by 5 percent, the extreme poverty rate rose from 47.4 percent in 1990 to 49
percent in 1999. The numbers are believed to be still rising (World Bank 2004)
 The U.N. Millennium Campaign Deputy Director for Africa says: Poverty
continues to intensify due to the exclusion of groups of people on the basis of
class, caste, gender, disability, age, race, religion and other status, (UN
Millennium Campaign 2009)
Who are the poors?
24
Poverty is still predominantly a

rural phenomenon.
Very high among people in rural

areas, among young people and
people with no education
Poverty: rural versus urban
25
Poverty: Indigenous vs non-indigenous
26
Indigenous Poverty in Latin America

Measuring inequality
27
 Measuring Inequality
 Size distributions (quintiles, deciles)
 Lorenz curves
 Gini coefficient
 Data on Inequality over time and across countries

Desirable Properties for Inequality
28
Measures
 Anonymity: measure should not depend on who has higher
income; e.g. whether we believe the rich or poor to be good or
bad people
 Scale independence: inequality measures should not depend on
size of the economy – want a measure of income dispersion
 Population independence principle: an inequality measure should
not be based on the number of income recipients
 Transfer principle - all other incomes constant, if transfer income
from a richer to a poorer person (not so much that the poorer
person is now richer than the originally rich person), resulting new
income distribution is more equal.
 Gini coefficient satisfies all four properties; so does the
coefficient of variation (CV), and some others
29
 Economic interpretation of Inequality:

 inequality measures the disparity between a
percentage of population and the percentage of
resources (such as income) received by that population.
 Inequality increases as the disparity increases.
 Ifa single person holds all of a given resource,

inequality is at a maximum. If all persons hold the
same percentage of a resource, inequality is at a
minimum.
30
 Size distributions (quintiles, deciles): useful to measure the

income distribution within countries.
 Divide population of a country into quantile groups: quartiles (4
groups), quintiles (5 groups), or deciles (10 groups).
 Compute the share of total income accruing to each quantile
 The quantile ratio is the average income in the highest quantile

over the average income in the lowest quantile. Provides no
information about people in the middle range of the income
distribution.
Typical Size Distribution of Personal Income in a Developing
Country by Income Shares - Quintiles and Deciles
32
 The Lorenz curve plots the cumulative share of

income held by the different quantiles of the
population (Figure 2.1).
 A completely equal distribution should match
the diagonal perfectly while a completely
unequal distribution should resemble a
backward L-shape.
Figure 2.1. The Lorenz Curve.
Figure 2.2. Lorenz Curve with Extremely Equal and Unequal Income Distribution.
 The Greater the Curvature of the Lorenz Line, the Greater the
Relative Degree of Inequality
36
 Curvature of Lorenz curve summarises inequality

 Perfect equality, everyone same income:
 Lorenz curve lies along a 45° ray from the origin
 Complete inequality, all income held by just one

person:
 Lorenz curve lies along the horizontal axis.
Gini Coefficient
 From the Lorenz curve we can construct the Gini coefficient, a
numerical representation of income inequality. It is defined as twice
the area between the diagonal and the Lorenz curve.
 In the case of perfect income equality, the Gini coefficient would
equal 0. In the case of complete income inequality, the Gini
coefficient would equal 1.
 Thus, G = 1 - 2S. Where S is the area under the Lorenz curve. The
ratio is between 0 and 1, and therefore can also be represented as
percentages.
38
The Lorenz Curve and the Gini Coefficient
G = A / (A + B)
Since A + B = 0.5 (axes range 0 to 1)
G=A/0.5= A x 2
B
Figure 2.3. The Lorenz Curve and the Gini Coefficient.
The Gini Coefficient
Twice the area between the Lorenz curve and the equality diagonal.
Pros Cons
 Generally regarded as  Requires comprehensive
gold standard in economic individual level data
work  Requires more sophisticated
 Incorporates all data computations
 Allows direct comparison
between units with different
size populations
 Attractive intuitive
interpretation
Other indices of inequality: Range
The range is simply the difference between the highest and lowest observations.
Number of employees Salary
2 $1,000,000
4 $200,000
6 $100,000
6 $60,000
8 $45,000
12 $24,000
In this example, the Range = $1,000,000-$24,000

= 976,000
Other indices of inequality: Range
The range is simply the difference between the highest and

lowest observations.
Pros Cons
 Easy to Understand  Ignores all but two of the
 Easy to Compute observations
 Does not weight
observations
 Affected by inflation
 Skewed by outliers
Other indices of inequality: Range Ratio
The Range Ratio is computed by dividing a value at one predetermined
percentile by the value at a lower predetermined percentile.
Number of employees Salary
95 percentile
Approx. equals 2 $1,000,000
36th person
4 $200,000
5 percentile 6 $100,000
Approx. equals
2nd person 6 $60,000
8 $45,000
12 $24,000
In this example, the Range Ratio=200,000/24,000 =8.33

Note: Any two percentiles can be used in producing a Range Ratio. In some contexts, this 95/5
ratio is referred to as the Federal Range Ratio.
Other indices of inequality: Range Ratio
The Range Ratio is computed by dividing a value at one
predetermined percentile by the value at a lower
predetermined percentile.
Pros
Cons
 Easy to understand
 Ignores all but two of the
 Easy to calculate observations
 Not skewed by severe  Does not weight observations
outliers
 Not affected by inflation
The Coefficient of Variation
The Coefficient of Variation is a distribution’s standard deviation divided by
its mean.
Both distributions above have the same mean, 1, but the standard deviation is
much smaller in the distribution on the left, resulting in a lower coefficient of
variation.
The Coefficient of Variation
The Coefficient of Variation is a distribution’s standard

deviation divided by its mean.
Pros Cons
 Fairly easy to understand  Requires comprehensive
 If data is weighted, it is individual level data
immune to outliers
 No standard for an
 Incorporates all data acceptable level of
 Not skewed by inflation inequality
Gini coefficient around the world
47
Inequality around the world
Figure 2.4. Income

Inequality in Selected
Countries.
Table 2.4. Income Inequality in Different Regions of the World (1995-2006).
50
51
Economic determinants of inequality
52
Education and Income Inequality

The variance in access to education across countries is a possible
determinant of income inequality.
 People in societies where access to education is limited will tend to
acquire less skills and, thus, earn lower wages.
 Uneducated people also tend to be excluded from groups in society that
are powerful enough to demand from their government.
 Data on the relationship between income inequality and education is
indicative, but not conclusive.
 i.e. Africa has both high income inequality and low literacy rates. South Asia: low
income inequality but literacy rates are even lower than in Africa.
 Also, the direction of causality is not clear. Income inequality may lead to
educational inequality, or vice versa.
Table 2.5. Education and Income Inequality.
54
 Land Ownership and Income Inequality

 Another possible determinant of income inequality is the
historical access to property ownership.
 The distribution of land ownership is particularly important in
developing countries, where a bigger share of the population
depends on agricultural production for subsistence.
 It is clear from the data (Figure 2.5) that income inequality
and land inequality are positively correlated (i.e., inequality
of land wealth leads to inequality of income).
 Latin America is also extremely inequitable in terms of land
ownership, while South Asia has more equitable distribution.
Figure 2.5. Income and Land Inequality.
Inequality, Growth, And Development
56
 Is inequality good or bad for economic growth?

 How does income inequality affect economic growth?
 What is the direction of causality?
 How has income inequality evolved historically?
57
The Kuznets Hypothesis

 Over the course of development, income inequality
follows a U-curve relationship. Inequality first increases but
then decreases.
 At initial stages, only small parts of society benefit, but
income eventually “trickles down” to the rest of society.
 One process that generates U effect is the shift from
agriculture to industry.
 However, there is little evidence to support this hypothesis
in both cross-sectional and panel studies.
Figure 2.6. The

Kuznets Curve.
59
The “Inverted-U” Kuznets Curve

60
Estimates of the Kuznets Curve

61
 Plot of Inequality Data for Selected Countries

62
Income Inequality and Economic Growth

Since the early 1990s there has been almost unanimous
consensus that income inequality is bad for economic growth.
 The data show a negative correlation between long-term
growth rates of GDP per capita and the Gini coefficient.

 In recent decades, Latin America has experienced both
high income inequality and poor economic growth.

 Asia, on the other hand, has relatively equitable
distribution and high rates of growth.
63
Some possible theoretical explanations include:

 Credit market imperfections: people in countries
with a more equal income distribution will have
access to more credit.
 Political economy: high inequality builds
pressure for redistribution of income, possibly

leading to high taxation of the rich.
Figure 2.7. Income Inequality and Economic Growth.

Poverty and Growth
65
 Relationship between Growth and Poverty

 Association between growth and poverty reduction
 When it is inclusive, growth reduces poverty
 Lower extreme poverty may also lead to higher growth
Poverty and Growth
66
 Poor health, nutrition, and education lowers

economic productivity of people in poverty, leading
directly and indirectly to slower growth
 Higher income for the poor raises demand for
locally produced goods
 Often, the poor lack access to credit, which
constrains entrepreneurship, children’s education,
and fertility reduction
 Social exclusion/injustice associated with poverty
also leads to bad government policies that can
reduce growth
Inequality over time
67
 Income inequality seems to have been increasing within countries

in the past few decades.
 It has increased in Central and Eastern Europe and in China as a result
of the transition from socialism to capitalism.
 It has increased in the US and many Western European countries.
 Alternatively, we can look at world-wide income distribution.
 Income inequality seems to be decreasing across countries as
living standards increase in developing economies.
 Xavier Sala-i-Martin finds evidence that the Gini coefficients across
countries have declined since around the 1980s.
 A possible explanation is the recent economic performance of China and India.
As these countries develop, hundreds of millions of people have been pulled out of
poverty.
Figure 2.8. The
Evolution of the
World Gini
Coefficient.
Figure 2.9.
Evolution of the
World Gini
Coefficient
(1970-200),
Excluding
Various
Countries.
«Within» or «between»?
70
 Two (related but different) concepts of inequality:

 Inequality between countries
 Inequality within country
 Between countries, three distinct concepts:
1) Unweighted international inequality:
compares per capita income across countries
2) Population - weighted international inequality:
compares weighted per capita income across countries
Note: inequality ranks in (1) and (2) are the same
3) World income distribution
compares incomes of all individuals in the world
Three concepts of global inequality
71
Inequality 1950-2009
72
Policy Options on Income Inequality and
Poverty: Some Basic Considerations
73
 Policy options
 Changing relative factor prices
 Progressive redistribution of asset ownership
 Progressive taxation
 Transfer payments and public provision of goods and

services
The Need for a Package of Policies
74
 Policies to correct factor price distortions

 Policies to change the distribution of assets, power,
and access to education and associated employment
opportunities
 Policies of progressive taxation and directed transfer
payments
 Policies designed to build capabilities and human and
social capital of the poor
 Some specific programs covered in later lectures
include: conditional cash transfers; and micro-finance
Elena Meschi
a.a. 2017/2018
Lectures 4, 5
Population growth
Outline
2
1. Show trends in global population growth over

time.
2. Explain the determinants of population growth
and the demographic transition.
3. Discuss determinants of fertility decisions and
motivations for family planning
4. Policy options
5. Paper Osili and Long (2008)
Why are economists interested in
3
fertility and population growth?
 The Malthusian problem: with fixed land, chances are that
population growth outweighs productivity growth (due to diminishing
returns of labor). This an underlying threat which materialized in
Europe before the industrial revolution and in Africa in the period
1950-70 (when rapid mortality decline -epidemiological transition -
was not accompanied by an as much rapid fertility decline). But in
the industrial era the Malthus trap did not snap shut in Europe,
America, Asia and Latin America
 Fast population growth rates have not implied falling per-
capita incomes as:
 labor productivity grew much faster than population behaviors
changed
 reduced fertility
Why are economists interested in
4
fertility and population growth?
 Variations in fertility would alter the age composition of
the population with consequences on saving rates (the
young save more) and the sustainability of the pension
system
 Fertility influences women participation to the labor

market, choices about children education, health care,
nutrition
 A world with finite resources cannot sustain unlimited

population growth
Population over time and the
5
demographic transition
 The last two centuries have seen unprecedented population
increase:
 World population was about 1 billion in the beginning of the 1800s
 Up to 1800s, world’s population was fairly stable through history. Wars,
famine and disease played a role in holding down growth.
 From then we have experienced a population “explosion” an increase
associated with the events associated with the industrial revolution (e.g.,
increases in agricultural productivity, income, advances in hygiene, health
science, etc.)
 6 billion in 2000, 7 billion in 2012
 About every 4 days, the world population increases by 1 million.
 In the last 60 years, the largest population increases have been
taking place in developing countries, especially in Asia and Africa.
Figure 3.1. Evolution of the World Population since 1000 BCE.
Figure 3.2. Population in 1950 and Population Increase (1950-2010) per
Continent.
World Population Growth, 1950-2050
World Population Distribution by Region,
2010 and 2050
10
 The growth rate of the population has increased dramatically
since the Industrial Revolution.
 Growth rate of the population: gt = ((pt – pt-1)/pt-1 )*100
 percentage yearly net relative change in population due to natural
increase and net international migration
 Natural increase: difference in the fertility rate and mortality rate.
 The global population growth rate increased steadily from
1800s to the 1960s, peaking at about 2%. Since then it has
decreased, to around 1.5% now.
Figure 3.3. Estimated Average Annual Population Growth Rates since 1000 BCE.
12
 A possible explanation is that the world, like individual
countries, is undergoing a demographic transition:
A shift from a stabilized population with high birth and
high mortality rates to a stabilized population
characterized by low birth and low mortality rates.
 Demographers expect the world population to stabilize
at around 8 to 12 billion.
 What are the key determinants that drive the
demographic transition?
Figure 3.4. Average Annual Population Growth Rates per Continent during
Recent Decades.
The determinants of population growth
14
 Components of population change:

 Additions:births (BT)
 Deductions: deaths (DT)
 Net In-migration (NMt)= immigration – emigration

𝑃𝑇 = 𝑃𝑇−1 + 𝐵𝑡 − 𝐷𝑡 + 𝑁𝑀𝑡
 Rate of growth:
𝑃𝑇 −𝑃𝑇−1 𝐵𝑡 𝐷𝑡 𝑁𝑀𝑡
𝑔𝑡 = = − −
𝑃𝑇 𝑃𝑇 𝑃𝑇 𝑃𝑇
𝑔𝑡 = 𝐶𝐵𝑅𝑡 − 𝐶𝐷𝑅𝑡 +𝑁𝑀𝑅𝑡
15
 Total fertility rate (TFR): average number of children born to a woman

of childbearing age (15-44).
 Birth rate: number of babies born each year per 1000 inhabitants.
 Death rate: number of deaths each year per 1000 inhabitants.
 Both the fertility and death rate can be computed age-specific.
 Population growth rate: difference between birth rate and death rate
(expressed per 1000 habitants, divided by 10 to get % growth rate).
 The net migration rate is the difference between the number of
persons entering and leaving a country per 1000 inhabitants.
 The birth, death, and fertility rates depend on the age distribution defined
as the percentage of the population belonging to different age groups.
16
Age Distribution and Population Growth

There is usually a high proportion of young people in
developing countries, while there is a usually a high
proportion of elderly people in developed countries.
Even where mortality and fertility rates are the same
the age distribution significantly affects population growth.
17
 Table 3.1. shows an example of the effect on population growth

of differences in age distribution.
 Suppose country A has 70 young and 30 old and country B has
50 young and 50 old (50% male and 50% female in each).
 Today’s young will become old in the next period and die. The
mortality rate among the young is 10%. Fertility rate is 2.2.
 A’s population in period 2 = 2.2*(70/2)+(70*.9) = 140  population
grows 40%
 B’s population in period 2 = 2.2*(50/2)+(50*.9) = 100  population
grows 0%
Table 3.1. The Effects of Age Distribution.
Population Pyramids: All Developed
and Developing Countries
Population
pyramids often
used to show the
age distribution
within a country.
Population Pyramid: Italy
20
Population Pyramid: Italy
21
Population Pyramid: China
22
Population Pyramid: Congo
23
Age structure: some concepts
24
 Dependency burden:
 Old age dependency ratio
 Ratioof retired people (over 65) over the working-age
population (15–64 years old)
 Burden on pension systems
 Youth dependency ratio
 Number of children (0–14 years old) over the working-age
population (15–64 years old)
 Gives rise to the hidden momentum of population growth.
 It is a dynamic latent process of population growth where
population continues to grow despite a fall in birth rate due to
larger number of child bearing couples.
 High birth rates cannot be altered overnight
Age structure: some concepts
25
 Demographic trends tend to have considerable inertia.

 A stationary population is defined as having zero growth over
time (i.e., stable fertility and mortality rates).
 Even temporary fertility shocks to a stationary population
creates inertia, leading to growth long after the shock (e.g.,
U.S. baby boom and baby boom echo).
 Therefore, family planning policies tend to have a delayed effect. This
must be taken into account when evaluating their effectiveness.
Population Pyramids
26
 Other data on population growth

 World Bank:
https://data.worldbank.org/indicator/SP.POP.GROW
 Eurostat: http://ec.europa.eu/eurostat/statistics-
explained/index.php/People_in_the_EU_%E2%80%9
3_population_projections
 World life expectancy project:
http://www.worldlifeexpectancy.com/italy-population-
pyramid
The Demographic Transition
 Stage I: High birthrates and death rates

 Stage II: Continued high birthrates, declining death
rates
 Stage III: Falling birthrates and death rates,
eventually stabilizing
The Demographic Transition in Western
Europe
The Demographic Transition in
Developing Countries
STAGE 1– High Stationary or Pre-Industrial
 High birth rates

 Little or no family planning
 Parents have many children because few survive
 Many children are needed to work the land
 Children are a sign of virility
 Some religious beliefs and cultural traditions encourage large families
 High death rates

 Disease and plague (e.g. bubonic, cholera, kwashiorkor)
 Famine, uncertain food supplies, and poor diet
 Poor hygiene, no piped clean water or sewage disposal
 Population Growth– Slow

STAGE 1 Pyramid
 Due to high birth rates, the pyramid shape would have a

wide base
 Due to high death rates, the pyramid would be very short
in height; concave shape indicates low life expectancy.
STAGE 2– Early Expanding or Early Industrial
(Rapid population growth)
 Ehrlich described this stage as the “population explosion”
 High Birth Rates
 As STAGE 1
 Falling death rates
 Improved medical care e.g. vaccinations , hospitals, doctors, new drugs, and scientific
inventions
 Improved sanitation and water supply
 Improvements in food production in terms of quality and quantity
 Improved transport to move food and doctors
 A decrease in child mortality
Population growth-- rapid
 As death rates are addressed, the population explosion begins
 The height of the pyramid grows to reflect the prevention of more deaths; shape becomes
less concave as life expectancy increases
 The width of the base remains large due to the ongoing high birth rates
STAGE 2 Pyramid
 As death rates are addressed, the population explosion begins;

 The height of the pyramid grows to reflect the prevention of more
deaths; shape becomes less concave as life expectancy increases;
 The width of the base remains large due to the ongoing high birth
rates
STAGE 3– Late Expanding or Late
Industrial
 Falling birth rates
 Family Planning utilized; contraceptives, abortions, sterilization, and other
government incentives adopted
 A lower infant mortality rates means less pressure to have children
 Increased mechanization and industrialization means less need for labour
 Increased desire for material possessions and less desire for large families
 Emancipation of women
 Death rates low
 As Stage 2
 Population growth– still growing but slowing

STAGE 3 Pyramid
 As birth rates begin to be addressed, the base begins to

stabilize and eventually narrow;
 Death rates are low and stabilized, meaning that the pyramid
continues to grow higher.
STAGE 4– Low Stationary or Low
Fluctuating
 Birth rates low
 Fertility rates plunge to below replacement rate (2.1 children per
woman) because:
 Valuation of women beyond childbearing and motherhood becomes
important
 Increasing value is placed on material goods over family size in modern
industrialized society
 Widespread choice of contraception by families
 Death rates low
 Child mortality reduced and life expectancy increased due to:
 Capital ($$$) investment in medical technology
 Widespread knowledge of healthy diet and lifestyle
 Population growth– slow or declining (due to aging

societies)
STAGE 4 Pyramid
 Birth rates and death rates are low; as fertility continues to

decline, an AGING SOCIETY emerges.
 Pyramid seems to “invert”…
Fertility Rate for Selected Countries, 1970
and 2012
https://data.worldbank.org/indicator/SP.DYN.TFRT.IN/
Fertility Rates by regions - 2016
40
Fertility rates - 2016

Sub-Saharan Africa
Middle East & North Africa
South Asia
Latin America & Caribbean
Europe & Central Asia (excluding high income)
North America
East Asia & Pacific
Euro area
Italy
0 1 2 3 4 5 6
The determinants of fertility rates
41
The Economics of Fertility Choices

 How can we make economic sense of fertility decisions?
 The Microeconomic Household Theory of Fertility

 Demand for Children in Developing Countries:
Cd = f (Y, Pc, Px, tx), x =1,..., n

 Cd is the demand for surviving children
 Y is the level of household income
 Pc is the “net” price of children
 Px is price of all other goods
 tx is the tastes for goods relative to children
42
Cd = f (Y, Pc, Px, tx), x =1,..., n

 The higher the household income, the greater the demand
for children.
 The higher the net price of children, the lower the quantity
demanded.
 The higher the prices of all other goods relative to
children, the greater the quantity of children demanded.
 The greater the strength of tastes for goods relative to
children, the fewer children demanded.
In symbols, these relationships may be written as:
¶Cd
>0
¶Y
44
How can we make economic sense of fertility decisions? Conventional

consumer theory assumes a utility function bundling consumer goods and
children subject to a budget constraint (Fig. 3.6).
 Shifts in relative price: A reduction in cost of a good relative to other goods
leads to an income effect and a substitution effect.
 If the income effect dominates, then it is possible to increase consumption of
all goods (panel a).
 If the substitution effect dominates, people tend to increase consumption of
cheaper good, and reduce consumption of more expensive goods (panel b).
 Evidence shows that the income effect dominates when there is a decrease in the
cost of children and the substitution effect dominates when there is a decrease in
the price of other goods.
45
 Figure 3.6 Panel (A) A Reduction in the Cost of Children

46
 Figure 3.6 Panel (B) A Reduction in the cost of consumer goods

 If there is an upward shift in income, it’s possible to afford both more
children and consumer goods (Figure 3.7).
 Inferior goods are defined as goods for which demand decreases as income
increases (potatoes, rice, millet).
 A normal good is one for which consumption increases as income increases.
 Typically, poor families have more children that rich families. Are children
inferior goods?
 Or, is the cost of children relatively higher in rich countries? Two possible
explanations:
 The opportunity cost of having children, especially for women, increases with
economic development.
 There is a quantity-quality trade off. At higher levels of development, parents
choose to invest more time and money per child.
Figure 3.7.
Fertility Choices
and Increases in
Income.
49
 A shift in preferences is another possible explanation for

fertility choices (Figure 3.8).
 The change in the relative benefits of having children
translates into a change in the shape of indifference curves
between consumption and children.
 Preference can change for noneconomic reasons,
especially social norms.
 A reduction in benefits will result in fewer children and
more consumption of goods and services.
 The benefits of having children are at least partly related
to the differences in institutions across countries.
Figure 3.8.
Reduction in the
Benefits of
Having
Children.
51
Fertility Choices and Institutions

 In societies with no social security, children may represent an
investment for retirement and a source of financial security.
Therefore, poor parents tend to have more offspring where:
 Infant and childhood mortality are high.
 Children tend to be less educated, because they may potentially be
less productive.
 Because of extreme poverty children move away from their parents.
 In advanced economies, parents can use the promise of inheritance as an
incentive to make sure their children take good care of them.
 In the absence of economic incentives, however, you often find that poor
countries enforce parental support through cultural and social values
52
Gender bias also plays a key role in determining fertility

decisions.
 In many societies boys are preferred to girls.
 This bias may go as far as inducing female infanticide.
 In general, from a biological perspective, there should be
more females than males.
 However, gender bias for boys in some cases
dramatically changes population sex ratios.
Figure 3.9. Population Sex Ratios in 2010.
Fertility Choices and the Demographic Transition
 Changes in fertility usually happen gradually as people take time to adapt to
the evolving economic and social trends.
 In the different stages of the demographic transition the costs and benefits of
having children change.
 Improved income, medical technologies, education, urbanization, pension
systems, and opportunities for women outside the home all have pushed fertility
rates down.
 Social norms evolve or disappear.
55
Fertility Choices and the Demographic Transition

 Changes in fertility usually happen gradually as people take
time to adapt to the evolving economic and social trends.
 In the different stages of the demographic transition the costs
and benefits of having children change.
 Improved income, medical technologies, education,
urbanization, pension systems, and opportunities for women
outside the home all have pushed fertility rates down.
 Social norms evolve or disappear.
Figure 3.10. China’s Population Pyramid in 2000, with Projections for 2050.
Family planning policies
57
To speed up the demographic transition, governments often

intervene with policies directed at lowering fertility.
 The “one child” policy implemented in China in the late 1970s
is an extreme example of population growth control.
 Recently, the gender imbalance has made it difficult for men to find
wives.
 Similar extreme policies were enacted in India during the
1960s-1970s. These included, for example, forced sterilization.
 More recent policies aimed at reducing fertility include
improving education and increased access to contraceptives.
 Does there exist an imperfect market for contraceptives?
Family planning policies
58
Externalities and the Economics of Family Planning

The economic justifications of family planning are often based on the
existence of externalities. In this case, while it is rational for a family to
have many children, this decision may impose negative impacts on others
and society as a whole, such as:
 Strain on natural resources
 Congestion (increased demand) in the provision of public services, especially
education.
 Population pressure in low-growing developing countries can induce people
to emigrate to developed nations.
 However, an influx of immigrants may be beneficial for developed nations,
especially in the case of an aging population (i.e., immigrants contribute to the
workforce).
59
 Fertility may be lowered with:

 Improved women’s education, role, and status
 Female nonagricultural wage employment
 Rise in family income levels through shared growth
 Reduction in infant mortality, better health care
 Development of old-age and social security plans
 Expanded schooling opportunities, lowered real costs
 Lowered prices and better information on contraceptives
 Direct incentives such as subsidy benefits
 Policies that have the effect of reducing boy preference
 The above list provides a framework for policy.
The Consequences of High Fertility: Some
Conflicting Perspectives
 Population growth: “It’s Not a Real Problem”:
 The real problem is not population growth but the following,
 Underdevelopment
 World resource depletion and environmental destruction
 Population Distribution
 Subordination of women
 “Overpopulation is a Deliberately Contrived False Issue”
 “Population Growth is a Desirable Phenomenon”
The Consequences of High Fertility: Some
Conflicting Perspectives
 “Population Growth Is a Real Problem”
 Empirical arguments
 Lower economic growth
 Poverty
 Adverse impact on education
 Adverse impact on health
 Food constraints
 Impact on the environment
 Frictions over international migration
Goals and Objectives: Toward a
Consensus
 Despite the conflicting opinions, there is some common
ground on the following:
 Population is not the primary cause of lower living levels, but may be
one factor
 Population growth is more a consequence than a cause of
underdevelopment
 It’s not numbers but quality of life
 Market failures: potential negative social externalities
 Voluntary decreases in fertility is generally desirable for most
developing countries with still-expanding populations
Goals and Objectives: Toward a
63
Consensus
 Some Policy Approaches
 Attend to underlying socioeconomic conditions that
impact development
 Family planning programs should provide education
and technological means to regulate fertility
 Developed countries have responsibilities too
 Address gender bias, causes of boy preference

Some Policy Approaches
 What Developing Countries Can Do
 Persuasion through education
 Family planning programs
 Address incentives and disincentives for having children through the
principal variables influencing the demand for children
 Coercion is not a good option
 Raise the socioeconomic status of women
 Increase employment opportunities for women (increases opportunity
cost of having more children, as in microeconomic household theory)
 Help facilitate genuine and faster development of developing countries
that still have high fertility rates
Some Policy Approaches
 What the Developed Countries Can Do Generally
 Address resources use inequities
 More open migration policies
 How Developed Countries Can Help Developing

Countries with Their Population Programs
 Research into technology of fertility control
 Financial assistance for family planning programs
Elena Meschi
a.a. 2017/2018
Lecture 6
Education and fertility
Paper Osili and Long, 2008
Education and fertility
2
 References:
 Osili,Una Okonkwo and B. T. Long. (2008) “Does
Female Schooling Reduce Fertility? Evidence from
Nigeria.” Journal of Development Economics 87, 1 :
57‐75
 Jungho Kim (2016) “Female education and its impact
on fertility”, IZA World of Labor: 228
Negative association between
3
education and fertility
Osili & Long (2008)
4
 Over the past two decades, many developing countries

have adopted policies designed to reduce rapid
population growth.
 Among policy alternatives, educating girls and young
women is considered a highly effective means of
lowering fertility and accomplishing this goal (United
Nations, 1995)
 Moreover, growing consensus that investments in the
education of young girls and women yield additional
private and social returns, including improved child
health and nutrition outcomes
Osili & Long (2008)
5
 Purpose of the paper: establishing a causal effect of

education on fertility by exploiting an exogenous variation
in the supply of primary education in Nigeria.
 Economic theory provides several explanations for female
education influencing fertility. Education:
 increases the opportunity cost of childbearing among educated
women.
 improves child health and reduces rates of child mortality: women
need to have fewer births to reach a given desired family size.
 improves knowledge and more effective use of contraceptive
methods.
 increases female autonomy and bargaining power in fertility
decisions.
Osili & Long (2008): Empirical issues
6
 Simple negative associations between education and fertility,

often found in literature, does not necessarily imply a causal
effect
 A negative association may depend on omitted variables, such as
individual ability or household and community resources, which
affect both schooling and fertility decisions.
 Schooling opportunities are not randomly placed in communities
and, for instance, could be correlated with the presence of family
planning services.
 If fertility choices lead to interruptions in schooling, then fertility
may be an endogenous variable within the context of schooling
decisions.
Osili & Long (2008): Empirical issues
7
 To identify causal impacts we need an exogenous

source of variation that hits schooling without
having a direct impact on fertility besides the effect
mediated by schooling itself
 Such an exogenous source of variation is

represented by the UPE - Universal Public
Education – program in Nigeria
Osili & Long (2008): UPE
8
 Large-scale, nationwide program designed to increase

educational attainment
 Introduced in September 1976
 Funded by the federal government, it provided tuition-free
primary education and increased the number of primary school
classrooms and teacher-training institutions
 During the UPE program, the N of primary school children in
Nigeria increased from 4.4 million stud in 1974 to 13.8 million
by 1981
 in September 1981, the UPE program ended
 reduced funding for primary schools
 reintroduction of school fees
 Gross primary enrollment fell in many states
Osili & Long (2008): Nigeria
9
Osili & Long (2008): UPE
10
 To capture the intensity of the program by state, the

analysis focuses on the amount of federal capital funds
disbursed for classroom construction in 1976.
 As shown in Table 1, the states that received the highest
levels of federal capital funds for classroom construction
per capita were located outside the former Western region
of Nigeria with the exception of Lagos.
 These were the states that had relatively low primary
school enrollment rates and levels of educational inputs
prior to the UPE program.
 We use the term high-intensity to refer to these states, which
experienced a significant expansion in educational inputs.
Osili & Long (2008): UPE - data
11
12
Number of primary school students 1970-1981

13
Number of primary schools 1970-1981
The introduction of the

nationwide UPE had a
major impact on student
participation in primary
school education and the
number of primary schools
in Nigeria, especially in the
high-intensity, non- Western
states.
14
 The introduction of the nationwide UPE had a major

impact on student participation in primary school
education and the number of primary schools in
Nigeria, especially in the high-intensity, non-
Western states.
Osili & Long (2008): Data
15
 1999 Nigerian Demographic Health Survey

(NDHS).
 This dataset is a nationally representative survey of
women, with detailed information on fertility, family
planning and education (less detailed as regards
income).
 it contains rich information on socioeconomic and
demographic variables for nearly ten thousand
Nigerian women.
Osili & Long (2008): Empirical strategy
16
 To estimate the impact of UPE on educational and fertility

decisions, the paper utilizes two empirical strategies that
exploit the variation in which cohorts and in which
geographical areas students were affected by UPE.
 First, the effects of the reform are assessed:
 DiD: difference in differences
 treated cohort vs untreated - earlier - cohort
 high-intensity regions vs low-intensity regions
 Next, the causal effect of education on fertility is estimated:
 IV: the amount of federal capital funds for classroom construction
is used as instrument
17
 In general a DiD design exploits the discontinuity introduced by

a given policy on an underlying trend of the treated regions,
which would have been otherwise common between the treated
and non treated (= control) regions in the absence of treatment.
 The general DiD specification is
 where
 T is for time: a dummy which takes 0 before the treatment/policy and
1 after the treatment
 P is the policy variable which takes 1 in treated regions and 0 in non-
treated regions
18
 The (continuous) variable funds per capita (P) is made

dichotomous: 1 high intensity 0 low intensity (see Table 1
- Lagos excluded)
 The before/after variable is the birth cohort (T):
 primary school lasts for six years in Nigeria. Pupils
should start school at age 6. Given the timing of the
program, UPE should have primarily affected
individuals born between 1970 and 1975 due to high
prevalence of underage and overage enrollments in first
grade in Nigeria, the control group is composed of
women who were aged 15 to 20 when the UPE was
initiated (born between 1956 and 1961)
Osili & Long (2008): Results
19
Notes: t-statistics are shown in parentheses. The t-statistics

reported are based on standard errors that are clustered at the
year*state level.
20
 Problems with a direct regression of fertility on

years of education (OLS)
 Unmeasured individual, household, and community-level
resources may affect both education and fertility
decisions. For example, an increase in the level of
economic development may lead to higher educational
attainment and lower fertility.
 Education may serve as a proxy for unobservable
factors, such as ability, cognitive skills, motivation, and
parental background, and these factors may be
important determinants of a woman's fertility choices
21
 The instrumental variables approach allows to

identify the causal effect of education
 Valid instruments are variables that affect the level of
educational attainment but have no direct impact on
fertility.
 Federal disbursed funds per capita for primary school
construction in the state where an individual was
educated are assumed to have no direct effect on
fertility, other than through their effect on educational
attainment (Is it obvious?)
Osili & Long (2008): Results
22
Osili & Long (2008): Conclusions
23
 At the mean, for each additional 100 naira per capita spent on primary
school classroom construction in 1976, we estimate a 2-year increase in
educational attainment
 Estimates suggest that an additional year of schooling reduces the
number of children born before age 25 by 0.26.
 UPE may also have positively affected the outcomes of the children of
the affected women. For example, health and schooling indicators for
children are known to improve with the level of female education
(Schultz, 1998).
 not to speak of the direct effect on human capital accumulation and of
the reduced population growth  economic growth
 Returns to universal primary education programs may be substantial
over the long term
Elena Meschi
a.a. 2017/2018
Lectures 7, 8, 9
Economic Growth
Outline of the lecture
 Determinants of economic growth.

 Neoclassical Solow growth model and the
endogenous growth model.
 Empirical evidence on the determinants of growth
Introduction
3
 Economic growth measures only growth in economic production.

 Economic development measures overall improvements in living
standards and in the quality of life.
 Economic development cannot take place without economic growth.
 What are the determinants of economic growth?
 Why are some countries rich and others poor?
 Capital accumulation only explains part of the differences in economic
growth across countries.
 Recently, economist have been paying special attention to the role of
determinants like geography and institutions.
Growth and factors of production
4
 The Solow growth model is the starting point to

determine why growth differs across similar countries
 it builds on the Cobb-Douglas production model by
adding a theory of capital accumulation
 developed in the mid-1950s by Robert Solow of MIT, it
is the basis for the Nobel Prize he received in 1987
 It highlights the limitations of capital accumulation, and
how it leaves a significant part of economic growth
unexplained.
5
Factors of Production
 A firm generates output by combining factors of production (e.g. labor,
capital, land) to produce output.
 To what extent do more primary factors such as labor and capital allow
for more production?
 GDP (the most common measure of aggregate production in an
economy) equals the sum of value added across the economy.
 To a large extent, the productivity of each factor depends on the
quantity of the other factor.
 Factor abundance is the relative availability of the different factors of
production. Which factor of production is more abundant in developed
economies? In developing economies?
6
 The production function is an equation of the joint effect of the

factors of production on output:
Yt = F(Kt,Lt)
 Assume Cobb-Douglas production function:
 Kt is the stock of capital at year t, Lt the amount of labor and Yt

output (e.g., GDP). At is a scale factor that links units of labor and
capital to dollars of output.
 This production function exhibits constant returns to scale. (e.g., if we
multiply each factor of production by a number, output is also
multiplied by that number).
7
Factor productivity
 Factor productivity measures the contribution of each factor of
production to output.
 Average labor productivity is obtained by dividing national
output by total employment in the economy:
 Kt / Lt = capital intensity (i.e., the amount of capital per

worker)
8
 Marginal productivity is the increase in output caused by an

additional unit of labor or capital.
 The marginal productivity of labor is:
 The marginal productivity of capital is:
 Both labor and capital exhibit diminishing marginal productivity

 Also, the MPL with respect to capital is positive. Adding an extra worker
in a highly mechanized firm adds more output than adding the same
worker in a less mechanized firm. Adding more mechanization makes
existing workers more productive.
9
Factor shares are the proportions of national income used to pay for
the capital and labor in production.
 In theory, in a competitive economy a factor is paid its marginal
productivity. Labor will be added as long as the MPL exceeds the
wage rate. Capital will be added as long as the MPK is higher than
the opportunity cost of capital, the interest rate.
 Thus, MPL = W and MPK = R.
 A useful result from the Cobb-Douglas production function is that, in
equilibrium, the factor shares are:
MPK .Kt MPL.Lt
  (1   )
Yt Yt
 In advanced economies the share of capital is about 1/3 of total income
while the share of labor is approximately 2/3.
 Growth accounting estimates the percentage of an economy’s growth rate
that can be explained using the growth rate of the labor force, the growth
rate of the capital stock, and residual factors.
 In the Cobb Douglas function, replace At with eat to assume that At growth at
constant rate a over time:
Take logs of both sides:
Differentiate wrt time to obtain growth rates:

gy = a + αgK + (1-α) gL
 Growth in national output is decomposed among:

1. the growth rate of capital multiplied by weight (factor share) α
2. the growth rate of labor multiplied by weight (1-α)
3. Parameter a, the growth rate of total factor productivity (TFP)
 Example: China between 2000 and 2005 where average gy = 9.5,

α =.5, gK = 12.6, and gL = 1.0. Therefore, TFP: 9.5 – (.5 * 12.6) – (.5 * 1) = 2.7
 TFP is defined as the part of an economy’s growth rate that cannot be
explained with the growth of labor and capital stock.
 TFP is often associated with technological progress, but may also be due to
other things, such as institutional reforms.
 TFP usually explains a big share of an economy’s growth rate, and is the
biggest source of variation for cross-country growth differences.
The Neo-classical Solow Growth Model
 “Neoclassical” growth model assumes competitive markets and a diminishing
marginal product of capital.
 Links production function with savings, investment, population growth, and
technical change
 Divide both sides of Cobb Douglas function by Lt :
Yt/Lt = A Ktα Lt (1-α-1) = A (Kt / Lt )α
or, if yt =Yt/Lt and kt = Kt / Lt the above is expressed:
yt = A kt α
 Labor productivity increases with capital intensity, or, in other words, at a
given labor force income per capita increases with capital accumulation.
Figure 4.1. The Solow Model.
The slope of this function is the

marginal product of capital per
change in y
worker. y
MPK 
change in k
MPK = f(k+1)–f(k) y=f(k)
Change in y
It tells us the change in output

per worker that results when we Change in k
increase the capital per worker
by one.
k
 National Income is allocated between consumption and saving:
Yt = Ct + St
 Define savings rate (s) where: St = s Yt
 Capital stock increases through capital investment (assuming δ is
depreciation rate):
It = Kt+1 – Kt + δ Kt
 Assume closed economy (no trade) and no government, then:
Yt = Ct + It so It = St
 When an economy is in macroeconomic equilibrium, then
St = sYt = It = Kt+1 – Kt +δKt

 Rearranging terms to find capital accumulation:
Kt+1 = (1-δ) Kt + sYt

 Labor increases with the population growth rate of n:
Lt = Lt-1 (1 + n)
 Divide capital stock equation by Lt:
Kt+1/ Lt = (1-δ) Kt / Lt + sYt / Lt
 Replacing notation:
(1 + n) kt+1 = (1-δ) kt + sA ktα
 Capital intensity in period t+1 is a function of capital intensity in the
previous period, the savings rate (+), the depreciation rate (-), and the
population growth rate (-).
17
 What is the dynamic of capital intensity and income per capita

over time?
 Figure 4.2 includes two functions representing both the left- and the
right-hand side of the capital intensity equation
 Left-hand side (1+n)kt+1 is a straight line with slope (1+n)
 Right-hand side 1 − 𝛿 𝑘𝑡 + 𝑠𝑦𝑡 is the sum of a linear function
1 − 𝛿 𝑘𝑡 and of a concave function 𝑠𝑦𝑡 = 𝑠𝑘𝑡𝛼
 Both functions instersect at a point where 𝑘𝑡+1 = 𝑘𝑡 = 𝑘 ∗ , that is
capital intensity and thus output per capita remain stable over time.
This intersection point is called “steady state”
 In the steady state the population grows at rate n, the saving rate is
s and the depreciation rate is δ and capital accumulation is such that
the level of capital remains stable over time
The Neo-classical Solow Growth
18
Model – steady state
 Steady state equation
 Steady state capital intensity increases with the saving rate s

and total factor productivity A and decreases with the growth
rate of population and n and the depreciation rate of capital
𝛿
Figure 4.2. Steady State in the Solow Model.
20
 Simpler version (ignore population growth):

 Kt+1/ Lt = (1-δ) Kt / Lt + sYt / Lt
 kt+1 = (1-δ) kt + syt
 Capital accumulation over time
 kt+1 – kt = syt – δkt  kt+1 – kt = s𝐴𝑘𝑡𝛼 – δkt
 If investment > depreciation  capital intensity increases
 If investment < depreciation  capital intensity decreases
 In the steady state kt+1 – kt =0  s𝐴𝑘𝑡𝛼 = δkt

When investment is greater than depreciation, the capital stock increases
The capital stock rises until investment equals depreciation: At this steady state point,
ΔK = 0
Investment, depreciation
Depreciation: d K
Investment: s Y
Net investment
K0 K* Capital, K
The Solow Diagram graphs the production function and the capital
accumulation relation together, with Kt on the x-axis:
Investment,
Depreciation
At this point,
dKt = sYt, so
Capital, Kt
Suppose the economy starts at K0:
Investment, •The yellow line is above the

green at K0:
Depreciation
•Saving = investment is greater
than depreciation at K0
•So ∆Kt > 0 because
•Since ∆Kt > 0, Kt increases

from K0 to K1 > K0
Capital, Kt
K0 K1
Now imagine if we start at a K0 here:
Investment,
•At K0, the green line is above the
Depreciation yellow line
•Saving = investment is now less
than depreciation
•So ∆Kt < 0 because
•Then since ∆Kt < 0,

Kt decreases from K0 to K1 < K0
Capital, Kt
K1 K0
We call this the process of transition dynamics: Transitioning from any
Kt toward the economy’s steady-state K*, where ∆Kt = 0 and growth
ceases
Investment,
Depreciation
No matter where
we start, we’ll
transition to K*!
At this value of K,
dKt = sYt, so
Capital, Kt
K*
We can see what happens to output, Y, and thus to growth if we rescale
the vertical axis:
Investment, • Saving = investment and

Depreciation, Income depreciation now appear
here
Y* • Now output can be

graphed in the space
above in the graph
• We still have transition
dynamics toward K*
• So we also have
dynamics toward a
steady-state level of
income, Y*
Capital, Kt
K*
The Solow Diagram with Output
At any point, Consumption is the difference between Output and
Investment: C = Y – I
Investment, depreciation,
and output
Output: Y
Y*
Consumption
Depreciation: d K
Y0
Investment: s Y
K0 K* Capital, K
 The Steady State in the Solow Growth Model
 At given rates of savings, population growth, and depreciation, the steady state is
reached where the growth of capital intensity and productivity (income per capita) is
zero.
 Transition growth paths:
 If (1 + n) kt+1 > (1-δ) kt + sA ktα capital intensity falls, capital is “diluted”
because total investment (savings) cannot overcome population growth and
depreciation.
 If (1 + n) kt+1 < (1-δ) kt + sA ktα capital intensity rises because investment
is high enough to overcome population growth and depreciation.
 The production function assumes diminishing marginal product of capital. Assuming
similar functions, therefore, poor countries that start out with lower capital intensity
have the potential to grow faster than rich countries.
29
– steady state
 the economy will settle in a steady state because the investment
curve, sY, has diminishing returns
 The rate at which production and investment rise is smaller as the
capital stock is larger
 A constant fraction of the capital stock depreciates every period,
which implies depreciation is not diminishing as capital increases
 In summary, as capital increases, diminishing returns implies that
production and investment increase by less and less, but
depreciation increases by the same amount .
 Eventually, net investment is zero and the economy rests in steady
state.
 Technological Progress and the Steady State
 What drives per capita growth in the steady state?
 Figure 4.3: An increase in TFP from A1 to A2 (represented by an upward
shift in the production function) acts to increase capital intensity, and,
therefore, output per person.
 Absent new TFP shocks, the economy will settle into a new steady state at k*2
 In the Solow Model, these positive shocks to TFP are usually thought of as
exogenous.
 The possibility of new technology diffusion increases the growth potential of
developing economies.
Figure 4.3. Technological Progress in the Solow Model.
The Effect of Different Savings Rates

 An increase in the savings rate leads to an increase in per capita
output in the steady state.
 Historically, many high-savings Asian economies also grew rapidly.
High investment rates also encourage adoption of new technology.
 On the other hand, China, the Soviet Union, and other centrally-
planned economies invested at a high rate, but growth sputtered
as government directed investment was often wasted.
 In general, economists agree that while high rate of savings and
investment is important, the efficiency of investment is just as
important.
 Differences in Human Capital
 Human capital is the skills, talents, and knowledge embodied in the people (i.e.
education).
 Modify the Solow Model to incorporate h, the level of human capital (average years of
schooling for instance):
(23)
 And the steady state income per capita increases:
(24)
 Differences in human capital can explain much cross-country variation in growth and
development, and human capital is a better predictor than investment rates.
 However, education does not account for a large part of growth variation, and some
highly educated nations lag in development.
 Moreover, like most factors, the direction of causality between human capital and
economic growth is unclear.
Human capital and growth
34
Plot of a regression of the average annual rate of growth (in percent) of real GDP per
capita in 1960-2000 on the initial level of real GDP per capita in 1960, average test
scores on international student achievement tests, and average years of schooling in
1960 (Source: Hanusheck and Woesmann, 2007)
 Income Convergence
 The most important prediction of the Solow model is income convergence: poor
countries should catch-up in terms of income per capita to rich countries.
 Assumes all else being equal (savings, depreciation, but especially similar production
functions i.e., technology)
 A greater marginal product of capital (return) means that poor countries have the
potential to grow faster than rich countries.
 There is not much evidence of this convergence in the data (Figure 4.4). Results, at
best, are mixed. Some countries have successfully converged, others have drifted
further behind.
 Most variation appears to be embodied in TFP (technology development and
adoption) which is exogenous to the Solow Model. How can we endogenously
explain differences in technological change across countries?
Figure 4.4. Growth of

Income Per Worker
(1960–2008) and 1960
Initial GDP Per Worker.
37
 Weaknesses of the Solow model:

1. It focuses on investment and capital, while the
much more important factor of TFP is still
unexplained.
2. It does not explain why different countries have
different investment and productivity rates.
3. The model does not provide a theory of
sustained long-run economic growth.
The Solow Growth Model: Summary
1. The starting point for the Solow model is the
production model described at the beginning of
these slides. To that framework, the Solow model
adds a theory of capital accumulation. That is, it
makes the capital stock an endogenous variable.
2. The capital stock is the sum of past investments. The
capital stock today consists of machines and
buildings that were bought over the last several
decades.
3. The goal of the Solow model is to deepen our
understanding of economic growth, but in this it’s only
partially successful. The fact that capital runs into
diminishing returns means that the model does not lead
to sustained economic growth. As the economy
accumulates more capital, depreciation rises one-for-
one, but output and therefore investment rise less than
one-for- one because of the diminishing marginal
product of capital. Eventually, the new investment is only
just sufficient to offset depreciation, and the capital
stock ceases to grow. Output stops growing as well, and
the economy settles down to a steady state.
4. There are two major accomplishments of the Solow
model. First, it provides a successful theory of the
determination of capital, by predicting that the
capital-output ratio is equal to the investment-
depreciation ratio. Countries with high investment
rates should thus have high capital-output ratios,
and this prediction holds up well in the data.
5. The second major accomplishment of the Solow model
is the principle of transition dynamics, which states that
the farther below its steady state an economy is, the
faster it will grow. While the model cannot explain
long-run growth, the principle of transition dynamics
provides a nice theory of differences in growth rates
across countries. Increases in the investment rate or
total factor productivity can increase a country’s
steady-state position and therefore increase growth,
at least for a number of years. These changes can be
analyzed with the help of the Solow diagram.
6. In general, most poor countries have low TFP levels
and low investment rates, the two key
determinants of steady-state incomes. If a country
maintained good fundamentals but was poor
because it had received a bad shock, we would
see it grow rapidly, according to the principle of
transition dynamics.
Endogenous Growth Theory
Endogenous growth theory suggests that growth is generated by
endogenous technical change resulting from innovation.
Boundless Knowledge-Based Growth
 Assuming constant technological change, potential for growth can be boundless
surmounting even limits to resources.
 Technology depends on knowledge, and the more knowledge there is in an
economy, the higher the growth rate.
 Unlike human capital, human knowledge is not embodied in people but can be
transmitted across time and space.
 Technology and knowledge develop through economic incentives
communicated through the market and, sometimes, through government policy.
 Knowledge can spillover, grow and accumulate indefinitely. It is only
constrained by the number of people producing it.
 Knowledge as a Non-Rival Good
 Knowledge is a non-rival good and can still be consumed by the seller even
after it is sold.
 Knowledge is also an excludable good. A seller can exclude others from its
consumption.
 Enforced property rights (such as patents and copyrights), for example,
prevent access to knowledge.
 Such exclusions can be important because people and firms need to have an
incentives (monopoly rents) to invest in research and development.
 Imperfect competition is necessary to encourage growth. (Recall that the Solow
model assumes perfect competition.)
 A balance must be struck between granting temporary monopoly rights and
promoting competition over the long run.
 Basic Equations of the Romer Model
 Labor is divided between the productive sector (Y) and the research sector (A):
where human capital is the stock of knowledge at time t (At) times productive
labor Lyt at time t.
 The production function:
 Since knowledge (At) depends on the allocation of labor to research, it is

endogenous to the model and can vary over time.
 The increment of knowledge is given by the stock of knowledge multiplied by the
labor force in the research sector and a parameter that can represent, say, the
efficiency of research:
 The Romer Model versus the Solow Model
 In the Romer Model, imperfect competition is necessary for innovation and
growth, while the Solow model assumes perfect competition.
 Innovation is endogenous in the Romer model, depending on stock of
knowledge and the R&D needed to build that knowledge. Innovation is
exogenous in the Solow model.
 Romer predicts that rich countries should grow faster than poor countries
because of a higher stock of knowledge (i.e., divergence). Solow predicts
that countries with low capital intensity should grow faster (convergence).
 Note the basic Romer model ignores possibilities of trade and technology
transfer between rich and poor countries.
 Are the protection of intellectual property rights (IPR) beneficial for poor
countries?
Empirical analysis of economic growth
What drives the main differences in economic growth between rich and poor
countries?
 Output per worker (labor productivity) is highly correlated with long-term growth.
 Table 4.1: In 1988 Canada’s productivity was about 94% of the U.S. level. The
capital-output ratio (K/Y) and TFP (A) level were about the same between the
countries. Most of the difference is explained by lower human capital ratio (H/L) in
Canada.
 The same is generally true with West European economies.
 Between the developed and developing economies differences in TFP seems to be
the main factor.
 What drives this difference in TFP?
 Geography (Sachs) or Institutions (North)?
 The growing consensus is that countries with stronger institutions develop faster.
Table 4.1. Decomposition of Productivity Differences (Ratios to the United States).
 Two geographic factors influencing growth: distance from the equator and
geographical isolation
 Rich countries tend to be located in temperate zones, poor countries in the
tropics:
• tropical disease (malaria, sleeping sickness),
• intense heat makes hard labor (farming, construction) difficult,
• rainfall volatility affects agricultural productivity and flooding.
 Landlocked areas tend to be underdeveloped:
• lower population density means less challenges and cooperation in using resources,
• lack of access to trade and outside influences (e.g., high transportation cost).
 Societies in tropics have not always been poor (e.g., Egypt, Aztecs, Inca, etc.)
Acemoglu, Johnson, and Robinson found negative correlation between
urbanization in 1500 (proxy for development) and income per capita now.
Figure 4.5. Log GDP Per
Capita in 1995 among
Former European
Colonies and
Urbanization Rate in
1500.
Empirical analysis of economic
51
growth
 The reversal of fortune
 When did the reversal take place?
 Not when the Europeans plundered the previously
rich societies or killed of their populations.
 In the 19th century, and intimately related to
industrialization.
52
 Douglass North wrote (modern) seminal analysis arguing that institutions are
key to understanding why some countries developed earlier than others.
 Since then, much empirical analysis has shown a strong causal link from
institutional quality to long term growth.
 Another paper by Acemoglu, et. al. shows that protection from
expropriation risk enhances growth (Figure 4.6).
 Used instrumental variable technique to establish causation. Extractive or
predatory institutions established where settler mortality was high (e.g.,
Peru, Mexico, Congo, etc.) while more efficient institutions established where
settler mortality was low (e.g., North America, New Zealand, etc.).
 Most of the current research in development economies tries to understand
the role of institutions and their effects.
Figure 4.6. Average
Protection against Risk of
Expropriation and Log
GDP Per Capita.
Figure 4.7. A
Negative
Correlation
between Settler
Mortality and
Protection
Against
Expropriation.
Elena Meschi
a.a. 2017/2018
Lecture 10
Institutions
Economic development and institutions
2
 Why growth rates differ across countries? Economic growth related

to amount of human capital, physical capital and technology:
Yt=f(At, kt, ht)
 yt : income per worker(capita)
 kt : physical capital per worker
 ht : human capital per worker
 At : productivity (residual, proxy for technology)
 Standard economic answers:

 Physical capital differences (poor countries don't save enough)
 Human capital differences (poor countries don't invest enough in education
and skills)
 Technology differences (poor countries don't invest enough in R&D and
technology adoption, and don't organize their production efficiently)
3
 But why countries differ in human capital, physical capital,

and technology? We need to understand why poor
countries don't save enough, don't invest enough, don't
develop and use technologies.
 Imagine that different laws and regulations, different
level of corruption, different political systems have a
major effect on investment, education and allocation
decisions and thus on economic development, then...
 Institutions matter
 But why do societies choose different institutions?
 And what are institutions anyway?
4
What are institutions?
5
 A very broad concept.

 D. North (1990, p.3) defines institutions as: "Institutions include
any form of constraint that human beings devise to shape human
interaction"
 Therefore institutions:
 are " humanly devised" (under human control).
 are "the rules of the game" setting "constraints" on human

behavior.
 constraints (rules) placed by law and social norms on human
behavior
 include all constraints imposed on individuals and groups by a
society 's system of beliefs and values. They shape incentives.
6
Institutions underlie well-functioning markets:

 Without quality institutions transaction costs will be
higher and thus less economic development is likely to
take place.
 The more complex products and markets, the greater
the importance of institutions. Since complexity
increases with development, institutions become more
important over time.
7
 There are two types of institutions:

 Formal institutions: codified in writing
 Political institutions set rules for major political figures, bodies, parties, etc.
Political rules of the game (democracy versus dictatorship, electoral laws,
constraints)
 Legal institutions: Non-political institutions codified by laws. Ex. common law
and civil law systems, religious law.
 Economic institutions: economic rules of the game (property rights,
contracting institutions)
 Informal institutions: not legally codified, based on social norms,
conventions, social values, and culture. Determine how a given set of
formal rules and formal institutions function in practice
 Enforced informally (e.g., peer pressure, moral obligations). Religion also
plays a role in shaping social norms and values.
8
Interactions Between Institutions

 Formal institutions (laws) evolve with informal institutions
(culture and social norms - e.g., capital punishment).
 Difficult to establish formal laws which conflict with
social norms (e.g., border versus ethnic division in
Africa)
 Formal institutions can strengthen informal institutions
and vice versa. They are complementary.
9
 Good institutions imply:

 Political
stability
 Absence of corruption
 Enforcement of law
 Absence of conflicts
Measuring institutional quality
10
 Different indicators:
 Survey based indicators: governance effectiveness
(e.g., World Bank)
 Objectives measures: Constraint of the executive,
constitutional rules (e.g., Polity IV)
Measuring institutional quality:
11
governance effectiveness
 Kaufmann, Kraay and Mastruzzi (2010), "The worldwide
Governance Indicators: A summary of the methodology, Data and
Analytical Issues", World Bank Policy Research Working Paper
5430: 'Governance' indicators created by World Bank.
 Governance defined as "the traditions and institutions by which
authority in a country is exercised"
 Based on subjective or perceptions based data on governance
reflecting the views of a diverse range of informed stakeholders,
including ten of thousands of household and rm survey
respondents, as well as thousands of experts working for the
private sector, NGOs, and public sector agencies.
 Cover 212 countries and territories for 1996, 1998, 2000, and
annually for 2002-2016.
12
 Six dimensions (-2.5 + 2.5)
 1. Voice and Accountability
 2. Political Stability and Absence of
Violence/Terrorism
 3. Government Eectiveness
 4. Regulatory Quality
 5. Rule of Law
 6. Control of Corruption
13
1. Voice and Accountability: "captures perceptions of the
extent to which a country's citizens are able to participate in
selecting their government, as well as freedom of expression,
freedom of association, and a free media"
2. Political Stability and Absence of Violence (PV): captures
perceptions of the likelihood that the government will be
destabilized or overthrown by unconstitutional or violent
means, including politically, motivated violence and terrorism.
3. Government Effectiveness(GE): captures perceptions of the
quality of public services, the quality of the civil service and
the degree of its independence from political pressures, the
quality of policy formulation and implementation, and the
credibility of the government's commitment to such policies.
14
4. Regulatory Quality (RQ): captures perceptions of the ability
of the government to formulate and implement sound policies
and regulations that permit and promote private sector
development.
5. Rule of Law (RL): captures perceptions of the extent to which
agents have confidence in and respect the rules of society,
and in particular the quality of contract enforcement,
property rights, the police, and the courts
6. Control of Corruption (CC): captures perceptions of the
extent to which public power is exercised for private gain,
including both petty and grand forms of corruption, as well as
"capture“ of the state by elites and private interests.
15
 Institutions are persistent
16
17
18
19
20
21
Indicators of political institutions
22
There are two main political regime types:

 Autocracy: all power resides in a single person or narrow
group
 Democracy: power held by citizens through elections of
representative government
 In most democracies, citizens have fundamental rights of freedom of
expression, freedom of association, and freedom of movement.
 Indicators:
 Freedom House Democracy standard (Figure 9.1)
 The Polity index ranks democracies from -10 (full autocracy) to
10 strong democracy) (Figure 9.2).
Indicators of political institutions
23
 Indicators of democracy:
 https://freedomhouse.org/report/freedom-
world/freedom-world-2017
 http://www.systemicpeace.org/polity/polity4.htm
Political regimes
Figure 9.1. Freedom House Electoral Democracies in 2008.
Political regimes
Figure 9.2. The 2010 Polity IV Index of World Democracy.
Elena Meschi
a.a. 2017/2018
Lectures 11, 12
International Trade
Outline
 Recent trends in world trade
 Gravity model explains who trade with whom
CHAPTER CHECKLIST
 Theory of comparative advantage explains patterns of

trade
 Economics of trade barriers
 What are the benefits of free trade versus the costs of free
trade?
 Trade Policy in Developing Countries
 Import-substituting industrialization
 Trade liberalization since 1985
World trade evolution and developing
countries
Increased Openness of Developed and Developing Economies
 Trade openness is measured as the share of exports to GDP.
 Trade share is a region’s exports as a percentage of world exports.
 The share of exports in GDP has trended up for all countries but especially
for developing countries (Figure 6.1).
A Diverse Trade Performance across Regions

 Large heterogeneity among developing countries: exports of East Asian
countries have gone up spectacularly, exports from Africa and the Middle
East have grown more slowly (Figure 6.2).
 The pattern of export specialization by developing countries is changing
over time. Traditionally, developing economics specialized in primary
commodities (food, minerals including fuels), but most recent growth is in
manufactured goods, especially in Asia (Figure 6.3).
countries
Figure 6.1. Export Shares of GDP (Total Exports over GDP, Both in Constant
2000 dollars).
countries
Figure 6.2. World Trade Shares of Developing Regions (as a Percentage of
World Exports).
countries
Figure 6.3. The Composition of Developing Countries’ Exports in 1980 and 2000.
countries
 Low- and middle-income countries have changed the
composition of their trade.
 In 2001, about 65% of exports from low- and middle-income countries
were manufactured products, and only 10% of exports were
agricultural products.
 In 1960, about 58% of exports from low- and middle-income countries
were agricultural products and only 12% of exports were manufactured
products.
 More than 90 percent of the exports of China, the largest
developing country and a rapidly growing force in world
trade, consist of manufactured goods.
The Changing Composition of Developing-Country
Exports
The Composition of World Trade, 2011
Gains from Trade
 That there are gains from trade is probably the
most important insight in international economics.
 Countries selling goods and services to each other
almost always generates mutual benefits.
1. When a buyer and a seller engage in a voluntary
transaction, both can be made better off.
 Norwegian consumers import oranges that they would
have a hard time producing.
Gains from Trade
How could a country that is the most (least) efficient producer of
everything gain from trade?
 Countries use finite resources to produce what they are most
productive at (compared to their other production choices),
then trade those products for goods and services that they
want to consume
 Countries can specialize in production, while consuming many
goods and services through trade.
Gains from Trade
 Trade benefits countries by allowing them to export
goods made with relatively abundant resources and
imports goods made with relatively scarce resources.
 When countries specialize, they may be more efficient
due to larger-scale production.
 Countries may also gain by trading current resources for
future resources (international borrowing and lending)
and due to international migration.
Gains from Trade
 Trade is predicted to benefit countries as a whole in
several ways, but trade may harm particular groups
within a country.
 Internationaltrade can harm the owners of resources
that are used relatively intensively in industries that
compete with imports.
 Trade may therefore affect the distribution of income
within a country.
Who trades with whom?
14
 Gravity model:
 Trade is affected by:
 Economy’s size
 Distance, barriers, borders and other trade
impediments
Who Trades with Whom?
 More than 30% of world output is sold across national

borders.
 The 5 largest trading partners with the U.S. in 2012 were
Canada, China, Mexico, Japan, and Germany.
 The largest 15 trading partners with the
U.S. accounted for 69% of the value of U.S. trade in 2012.
Total U.S. Trade with Major Partners, 2012
Size Matters: The Gravity Model
 3 of the top 10 trading partners with the U.S.
in 2012 were also the 3 largest European economies:
Germany, the United Kingdom, and France.
 Why does the United States trade more with these European
countries than with others?
 These countries have the largest gross domestic product (GDP),
the value of goods and services produced in an economy, in
Europe.
 Each European country’s share of U.S. trade with Europe is
roughly equal to its share of European GDP.
Size Matters: The Gravity Model
 The size of an economy is directly related to the

volume of imports and exports.
 Larger economies produce more goods and services, so
they have more to sell in the export market.
 Larger economies generate more income from
the goods and services sold, so they are able to buy
more imports.
 Trade between any two countries is larger, the
larger is either country.
Using the Gravity Model: Looking for
Anomalies
 A gravity model fits the data on U.S. trade with
European countries well but not perfectly.
 The Netherlands, Belgium and Ireland trade much
more with the United States than predicted by a
gravity model.
 Irelandhas strong cultural affinity due to common
language and history of migration.
 The Netherlands and Belgium have transport cost
advantages due to their location.
Impediments to Trade: Distance,
Barriers, and Borders
Other things besides size matter for trade:
1. Distance between markets influences transportation costs and
therefore the cost of imports and exports.
2. Cultural affinity: close cultural ties, such as a common language,
usually lead to strong economic ties.
3. Geography: ocean harbors and a lack of mountain barriers make
transportation and trade easier.
4. Multinational corporations: corporations spread across different
nations import and export many goods between their divisions.
5. Borders: crossing borders involves formalities that take time, often
different currencies need to be exchanged, and perhaps monetary
costs like tariffs reduce trade.
Barriers, and Borders
 Estimates of the effect of distance from the gravity model
predict that a 1% increase in the distance between countries is
associated with a decrease in the volume of trade of 0.7% to
1%.
 Besides distance, borders increase the cost and time needed to
trade.
 Trade agreements between countries are intended to reduce
the formalities and tariffs needed to cross borders, and
therefore to increase trade.
Barriers, and Borders (cont.)
 Yet even with a free trade agreement between the U.S. and
Canada, which use a common language, the border between
these countries still seems to be associated with a reduction in
trade.
 Data shows that there is much more trade between pairs of
Canadian provinces than between Canadian provinces and
U.S. states, even when holding distance constant.
 Estimates indicate that the U.S.-Canadian border deters trade
as much as if the countries were 1,500-2,500 miles apart.
The Changing Pattern of World Trade:
Has the World Gotten Smaller?
 The negative effect of distance on trade according to the
gravity models is significant, but has grown smaller over time
due to modern transportation and communication.
 Technologies that have increased trade:
 Wheels, sails, compasses, railroads, telegraph, steam
power, automobiles, telephones, airplanes, computers, fax machines,
Internet, fiber optics, personal digital assistants, GPS satellites…
24
The Changing Pattern of World Trade:
 Political factors, such as wars, can change trade patterns
much more than innovations in transportation and
communication.
 World trade grew rapidly from 1870 to 1913.
 Then it suffered a sharp decline due to the two world wars and the
Great Depression.
 It started to recover around 1945 but did not recover fully until around
1970.
 Since 1970, world trade as a fraction of world GDP has

achieved unprecedented heights.
 Vertical disintegration of production has contributed to the rise in the
value of world trade through extensive cross-shipping of components.
The Fall and Rise of World Trade
Patterns of Trade
 The pattern of trade describes who sells what to whom.
 Differences in climate and resources explain why Brazil
exports coffee and Saudi Arabia exports oil.
 But why does Japan export automobiles, while the U.S.
exports aircraft?
 Why some countries export certain products can stem from
differences in:
 Labor productivity
 Relative supplies of capital, labor and land and their use in the
production of different goods and services
The theory of comparative advantage
 The theory of comparative advantage (C.A.) explains how economies
benefit from trade through specialization and, therefore, greater
efficiency.
 C.A. also provides the justification for free trade that is almost
unanimously accepted among economists.
The Benefits of Exchange and Specialization

 Exchange benefits all parties to the exchange. Why?
 Countries (or firms) specialize in areas where their relative
opportunity costs of production are lower.
 Relative costs are found by comparing production price ratios.
In Table 6.1, the price of a DVD player in terms of shirts is 6

(30/5), and in China the same ratio is 10 (60/6). Japan has a
comparative advantage in DVD players and China in shirts.
Table 6.1. Production of DVD Players and Shirts in Japan and China (in Hours).
30
 A country has a comparative advantage in producing a

good if the opportunity cost of producing that good is
lower in the country than in other countries.
 The opportunity cost of producing something measures the
cost of not being able to produce something else with the
resources used.
 For example, a limited number of workers could produce
either DVD players or shirts.
 The opportunity cost of producing DVD is the amount of shirts not
produced.
 The opportunity cost of producing shirts is the amount of DVD not
produced.
Factor Endowments and Comparative Advantage

 Relative factor endowments, the capacities of a country in terms of
capital and labor, are a key determinant of an economy’s
comparative advantage.
 In general, capital is relatively abundant in rich countries, while labor is
more abundant in developing economies.
 Hence, rich countries have comparative advantage in capital intensive
processes while poor countries have comparative advantage in labor
intensive processes.
 In the neoclassical trade model, the production possibilities frontier
(PPF), which is the maximum production of one good, for a given
quantity produced of another good, given the economies endowment
of labor and capital.
 Figure 6.4: Because of abundant labor, the opportunity costs of

producing a DVD player in terms of shirts is higher in China than in
Japan.
 Without international trade (autarchy), production in each
country is determined at the tangency point (point A) between the
PPF and an indifference curve (determined by domestic demand
preferences).
 With trade at the world relative price ratio, domestic production
is at point B and consumption is at point C, a higher level of utility
in each country.
 Trade is indicated by trade triangles: for each country exports
are given by line BD and imports by BC.
Figure 6.4. Factor Endowments and Comparative Advantage.
Comparative Advantage and Patterns of Trade

 C.A. explains some patterns of trade, especially specialization
based on factor endowments.
 C.A. can also help describe how endowments and trade
patterns change through time.
 Successful exporters start with labor and/or resource intensive
products and manufacturers, but as capital accumulates they
diversify into more sophisticated manufactured products.
 Asian Exporters: Japan, South Korea, Taiwan, Thailand, China
Trade Specialization and Export Price Risk

 Trade specialization is sometimes risky. Columbia’s dominance in coffee
leaves it vulnerable to price fluctuations. Steep price falls are especially
dangerous.
 Developing economies are more prone to such risk: Primary commodity
prices are especially volatile, but export diversification in small
developing economies is usually low.
 Export diversification often requires a sustained effort.
 A country’s terms of trade is the ratio of export prices to import prices.
Volatility in the terms of trade is a good measure of export price risk.
 Figure 6.5 shows that volatility in terms of trade has been decreasing
over time (until more lately).
Figure 6.5. Terms
of Trade.
The politics of trade
Winners and Losers in International Trade
 The political economy of trade liberalization analyzes conflicts between
those who benefit and those who don’t from liberalization.
 Trade liberalization helps exporting sectors and firms and harms import-
competing firms and sectors. The gains and losses for such firms often
fall narrowly and can be quite large for individuals.
 Consumers of intermediate and final goods benefit. These benefits,
while substantial, are spread over a large constituency and seem very
small to individual consumers.
The politics of trade
 Policy makers affect the amount of trade through
 tariffs: a tax on imports or exports,
 quotas: a quantity restriction on imports or exports,
 export subsidies: a payment to producers that export,
 or through other regulations (ex., product specifications)
that exclude foreign products from the market, but still
allow domestic products.
 What are the costs and benefits of these policies?
The cost of trade barriers
Protectionism and Tariffs
 Protectionism occurs when a country erects trade barriers to protect
domestic industry.
 Import tariffs (taxes) reduce the volume of imports because consumers
pay a higher price.
 Figure 6.6 shows the economics of tariffs:

Figure 6.6. Economic Losses Due to a Tariff Barrier.
41
 Before tariffs, domestic price is $100 per DVD player (same a

import price), domestic demand is Qd, domestic supply is Qs, and
import is Qd - Qs.
 After $20 tariff, domestic prices rises to $120, domestic demand is
Qd’, domestic supply is Qs’, import is Qd’–Qs’.
 Consumers pay more for reduced demand, rectangle HDGF.
 Government collects CDEF as tariff revenue.
 Chinese firms earn rents of HCGA.
 Welfare loss 1: triangle ACE is loss to rest of economy of allowing
resources to be used inefficiently
 Welfare loss 2: triangle DBF is loss to consumer surplus
Nontariff Barriers
 Nontariff barriers are often more insidious than tariffs because they
are less visible but just as harmful.
 Import quotas limit the total volume of imports to a country over a
period of time and are often regulated via “licenses.”
 Voluntary export restraints limit exports to individual countries to
preempt protectionist backlash (Japanese cars in 1980s).
 Regulatory barriers take the form of government regulations that limit
imports (e.g., product standards, arbitrary branding). These are
often difficult to detect.
 Subsidies used to resist imports or encourage exports. The U.S.
subsidizes cotton industry exports.
Trade institutions: bilateral vs.
Multilateral
 Bilateral trade negotiations are held between two countries.
 Multilateral trade negotiations are held among multiple countries,
usually under the auspices of a multilateral framework or
organization such as the WTO (formerly GATT).
 Bilateral liberalization often results in trade diversion, inefficient
trade flows that would not exist under full multilateral trade
liberalization.
 Regional trade agreements often promote wasteful trade diversion
(e.g., MERCOSUR).
 Multilateralism enlarges the set of possible exchanges between
different countries. It works better and is often easier to achieve.
Multilateral
The World Trade Organization (WTO)
 Succeeded GATT in 1995, 153+ member countries
 Main role is for multilateral negotiations, but also helps implement
agreements (rules), settle trade disputes, and enforce compliance.
 Nondiscrimination:
• Cannot have different trade rules for different exporting countries (MFN:
most favored nation clause)
• Same treatment of all products in the domestic market (national treatment
clause)
 Reciprocity: concession from one country should be matched by trade partners
 Figure 6.7: Nondiscrimination and reciprocity drive down trade barriers
 Some argue that the WTO promotes the agenda of rich countries (agriculture,
TRIPS).
Multilateral
Figure 6.7. Effects of Reciprocity and Nondiscrimination.
Exchange rates policies in developing
countries
 International transactions involve currency exchange. Exchange rate valuations
affect trade.
 In general, a “lower” or “weaker” exchange rate tends to encourage exports
and discourage imports, and a “higher” or “stronger” currency hurts exports and
promotes imports.
 Convertibility:
• Trade convertibility (current account): importers and exporters face no restrictions in exchanging
proceeds of trade
• Financial convertibility (capital account): flows of capital between countries are unrestricted
 Current account convertibility is vital for trade, but there is no consensus on the net
benefits of capital account convertibility.
 Absence of financial convertibility creates financial repression (e.g., households in China
have only narrow choices for savings).
 Full financial convertibility, however, poses substantial risk of financial crises and bank runs
(East Asian Crisis of 1997-1998).
Exchange rates policies in developing
countries
 Fixed exchange rates: the price of a country’s currency remains fixed in terms of other
currency(/ies)
• When the supply of a currency is not equal to demand, the central bank intervenes by selling or buying
reserves of foreign currencies as needed to keep exchange rates fixed. Lose control over money supply
 Floating exchange rates: Exchange rate set by supply and demand

 Volatility (large, sudden depreciation or appreciation) in exchange rates is harmful to
exporters and it complicates price and output management.
 Seeking stability especially for price expectations, developing economies sometimes
peg their currency to a major currency.
 Fixed exchanges rates are used as an “anchor” for a credible anti-inflationary
commitment. A currency board is a fixed exchange regime (e.g., Argentina).
 Fixed exchange rates are not as efficient for dealing with external and internal
economic shocks, especially under full capital account convertibility (a major lesson of
the Asian crisis).
48
Trade Policy in Developing Countries
 Import-substituting industrialization
 Trade liberalization since 1985
 Trade and growth: takeoff in Asia
Import-Substituting Industrialization
 Import-substituting industrialization was a trade policy
adopted by many low- and middle-income countries before
the 1980s.
 The policy aimed to encourage domestic industries by limiting
competing imports.
Effective Protection of Manufacturing in Some Developing
Countries (percent)
Import-Substituting Industrialization (cont.)
 The principal justification of this policy was/is the infant

industry argument:
 Countries may have a potential comparative advantage in some
industries, but these industries cannot initially compete with well-
established industries in other countries.
 To allow these industries to establish themselves, governments should
temporarily support them until they have grown strong enough to
compete internationally.
Problems with the
Infant Industry Argument
1. It may be wasteful to support industries now that will have a

comparative advantage in the future.
2. With protection, infant industries may never “grow up” or
become competitive.
3. There is no justification for government intervention unless
there is a market failure that prevents the private sector
from investing in the infant industry.
Infant Industries and Market
Failures
 Two arguments for how market failures prevent infant
industries from becoming competitive:
1. Imperfect financial asset markets
 Because of poorly working financial laws and markets (and more
generally, a lack of property rights), firms cannot or do not save
and borrow to invest sufficiently in their production processes.
 If creating better functioning markets and enforcing laws is not
feasible, then high tariffs would be a second-best policy to
increase profits in new industries, leading to more rapid growth.
Infant Industries
and Market Failures (cont.)
2. The problem of appropriability

 Firms may not be able to privately appropriate the benefits of their
investment in new industries because those benefits are public goods.
 The knowledge created when starting an industry may not be
appropriable (may be a public good) because of a lack of property
rights.
 If establishing a system of property rights is not feasible, then high
tariffs would be a second-best policy to encourage growth in new
industries.
Import-Substituting Industrialization
(cont.)
 Import-substituting industrialization in Latin American countries
worked to encourage manufacturing industries in the 1950s
and 1960s.
 But economic development, not encouraging manufacturing,
was the ultimate goal of the policy.
 Did import-substituting industrialization promote economic
development?
 No, countries adopting these policies grew more slowly than others.
Import-Substituting Industrialization (cont.)
 It appeared that the infant industry argument was not as valid

as some had initially believed.
 New industries did not become competitive despite or because
of trade restrictions.
 Import-substitution industrialization involved costs and
promoted wasteful use of resources:
 It involved complex, time-consuming regulations.
 It set high tariff rates for consumers, including firms that needed to buy
imported inputs for their products.
 It promoted inefficiently small industries.
Trade Liberalization
 Some low- and middle-income countries that had relatively
free trade had higher average economic growth than those
that followed import substitution.
 By the mid-1980s, many governments had lost faith in import
substitution and began to liberalize trade.
 Dramatic fall in tariff rates in India and Brazil, and less drastic reductions in
many other developing countries.
Trade Liberalization (cont.)
 Trade liberalization in developing countries occurred along
with a dramatic increase in the volume of trade.
 The share of trade in GDP has tripled over 1970–1998, with most of
the growth happening after 1985.
 The share of manufactured goods in developing-country exports surged,
coming to dominate the exports of the biggest developing economies.
 A number of developing countries have achieved
extraordinary growth while becoming more, not less, open to
trade.
Fig. 11-1: Tariff Rates in Developing Countries
Fig. 11-2: The Growth of Developing-Country Trade
 Has trade liberalization promoted development? The evidence
is mixed.
 Growth rates in Brazil and other Latin American countries have been
slower since trade liberalization than they were during import-
substituting industrialization.
 But unstable macroeconomic policies and financial crises

contributed to slower growth since the 1980s.
 Other countries like India have grown rapidly since liberalizing trade in
the 1980s, but it is unclear to what degree liberalized trade contributed
to growth.
 Some economists also argue that trade liberalization has contributed to
income inequality, as the Heckscher-Ohlin model predicts.
Trade and Growth: Takeoff in Asia
 Instead of import substitution, several countries in East Asia
adopted trade policies that promoted exports in targeted
industries.
 Japan, Hong Kong, Taiwan, South Korea, Singapore, Malaysia, Thailand,
Indonesia, and China have experienced rapid growth in various export
sectors and rapid economic growth in general.
Trade and Growth: Takeoff in Asia (cont.)
 These high-performance Asian economies generated a high volume of

exports and imports relative to total production.
 Their policy reforms were followed by a large increase in openness, as

measured by their share of exports in GDP.
 So it is possible to develop through export-oriented growth.
 However, Latin American nations such as Mexico and Brazil, which also
sharply liberalized trade and shifted toward exports, did not see
comparable economic takeoffs.
 These Latin American results suggest that other factors must have played a
crucial role in the Asian miracle.
Fig. 11-3: The Asian Takeoff
Fig. 11-4: Asia’s Surging Trade
Trade and Growth: Takeoff in Asia (cont.)
 It’s unclear if the high volume of exports and imports caused

rapid economic growth or was merely correlated with rapid
economic growth.
 High saving and investment rates could have led to both rapid economic
growth in general and rapid economic growth in export sectors.
 Rapid growth in education led to high literacy and numeracy rates
important for a productive labor force.
 These nations also undertook other economic reforms.
Summary
1. Import-substituting industrialization aimed to promote economic
growth by restricting imports that competed with domestic
products in low- and middle-income countries.
2. The infant industry argument says that new industries need
temporary trade protection due to market failures:
 imperfect asset markets that restrict saving, borrowing, and investment
in production processes
 problems of appropriating gains from private investment in production
processes
Summary (cont.)
3. Import-substituting industrialization was tried in the 1950s
and 1960s but by the mid-1980s it was abandoned for
trade liberalization.
4. The effect of liberalized trade on national welfare is still
being debated.
 Trade helped growth in some sectors, but saying that trade caused
higher overall economic growth has attracted some skepticism.
 Some argue that trade has caused increased income inequality.
Summary (cont.)
5. Several East Asian economies adopted export- oriented
instead of import-substituting industrialization.
 High export and import volumes and relatively low trade restrictions
were characteristics of this policy.
 It’s unclear to what degree this policy contributed to overall economic
growth, especially since other countries have not had similar successes.
1
Policy evaluation methods

 Based on Markus Goldstein, Poverty Reduction
Group, The World Bank
Outline - methods
 Monitoring and impact evaluation
 Why do impact evaluation
 Why we need a comparison group
 Methods for constructing the comparison group
 When to do an impact evaluation
Impact evaluation
3
 Impact is the difference between outcomes with the

program and without it
 The goal of impact evaluation is to measure this
difference in a way that can attribute the
difference to the program, and only the program
Why it matters
 We want to know if the program had an impact
and the average size of that impact
 Understand if policies work
 Justification
for program (big $$)
 Scale up or not – did it work?
 Meta-analyses – learning from others
 (withcost data) understand the net benefits of the
program
 Understand the distribution of gains and losses
What we need
 The difference in outcomes with the program versus
without the program – for the same unit of analysis
(e.g. individual)
 Problem: individuals only have one existence
 Hence, we have a problem of a missing counter-
factual, a problem of missing data

Thinking about the counterfactual
 Why not compare individuals before and after (the

reflexive)?
 The rest of the world moves on and you are not sure
what was caused by the program and what by the rest
of the world
 We need a control/comparison group that will
allow us to attribute any change in the “treatment”
group to the program (causality)
comparison group issues
 Two central problems:
 Programs are targeted
 Program areas will differ in observable and unobservable ways
precisely because the program intended this
 Individual participation is (usually) voluntary
 Participants
will differ from non-participants in observable and
unobservable ways
 Hence, a comparison of participants and an arbitrary
group of non-participants can lead to heavily biased
results
Example: providing fertilizer to
farmers
 The intervention: provide fertilizer to farmers in a
poor region of a country (call it region A)
 Program targets poor areas
 Farmers have to enroll at the local extension office to
receive the fertilizer
 Starts in 2002, ends in 2004, we have data on yields for
farmers in the poor region and another region (region B) for
both years
 We observe that the farmers we provide fertilizer to
have a decrease in yields from 2002 to 2004
Did the program not work?
 Further study reveals there was a national drought,

and everyone’s yields went down (failure of the
reflexive comparison)
 We compare the farmers in the program region to
those in another region. We find that our “treatment”
farmers have a larger decline than those in region B.
Did the program have a negative impact?
 Not necessarily (program placement)
 Farmers in region B have better quality soil (unobservable)
 Farmers in the other region have more irrigation, which is key in this
drought year (observable)
OK, so let’s compare the farmers in
region A
 We compare “treatment” farmers with their neighbors. We
think the soil is roughly the same.
 Let’s say we observe that treatment farmers’ yields decline by
less than comparison farmers. Did the program work?
 Not necessarily. Farmers who went to register with the program may
have more ability, and thus could manage the drought better than their
neighbors, but the fertilizer was irrelevant. (individual unobservables)
 Let’s say we observe no difference between the two groups.
Did the program not work?
 Not necessarily. What little rain there was caused the fertilizer to run
off onto the neighbors’ fields. (spillover/contamination)
The comparison group
In the end, with these naïve comparisons, we cannot
tell if the program had an impact
 We need a comparison group that is as identical in
observable and unobservable dimensions as
possible, to those receiving the program, and a
comparison group that will not receive spillover
benefits.
How to construct a comparison group –
building the counterfactual
1. Randomization
2. Matching
3. Difference-in-Difference
4. Instrumental variables
5. Regression discontinuity
1. Randomization
 Individuals/communities/firms are randomly assigned into
participation
 Counterfactual: randomized-out group
 Advantages:
 Often addressed to as the “gold standard”: by design: selection bias is zero on
average and mean impact is revealed
 Perceived as a fair process of allocation with limited resources
 Disadvantages:
 Ethical issues, political constraints
 Internal validity (exogeneity): people might not comply with the assignment (selective
non-compliance)
 Unable to estimate entry effect
 External validity (generalizability): usually run controlled experiment on a pilot, small
scale. Difficult to extrapolate the results to a larger population.
Randomization in our example…
 Simple answer: randomize farmers within a

community to receive fertilizer...
 Potential problems?
 Run-off(contamination) so control for this
 Take-up (what question are we answering)
3. Difference-in-difference
 Observations over time: compare observed changes in the
outcomes for a sample of participants and non-participants
 Identification assumption: the selection bias is time-invariant

(‘parallel trends’ in the absence of the program)
 Counter-factual: changes over time for the non-participants
 Constraint: Requires at least two cross-sections of data, pre-

program and post-program on participants and non-participants
 Need to think about the evaluation ex-ante, before the program
As long as the bias is additive and time-
invariant, diff-in-diff will work ….
Y1
Impact
*
Y1
Y0
t=0 t=1 time

Journal of Development Economics 87 (2008) 57 – 75
www.elsevier.com/locate/econbase
Does female schooling reduce fertility? Evidence from Nigeria

Una Okonkwo Osili a,⁎, Bridget Terry Long b,c
a
Department of Economics, Indiana University–Purdue University at Indianapolis, United States
b
Harvard Graduate School of Education, United States
c
NBER, United States
Received 27 July 2004; received in revised form 22 June 2007; accepted 4 October 2007
Abstract
The literature generally points to a negative relationship between female education and fertility. Citing this pattern,
policymakers have advocated educating girls and young women as a means to reduce population growth and foster sustained
economic and social welfare in developing countries. This paper tests whether the relationship between fertility and education is
indeed causal by investigating the introduction of universal primary education in Nigeria. Exploiting differences in program
exposure by region and age, the paper presents reduced form and instrumental variables estimates of the impact of female education
on fertility. The analysis suggests that increasing female education by one year reduces early fertility by 0.26 births.
© 2007 Elsevier B.V. All rights reserved.
JEL classification: J13; D10; O10

Keywords: Fertility; Female education; Instrumental variables; Universal education; Nigeria
1. Introduction association between female education and fertility

(Schultz, 1998). In fact, the negative relationship
Over the past two decades, many developing between female education and fertility has been
countries have adopted policies designed to reduce described as “one of the most clear-cut correlations” in
rapid population growth. Among policy alternatives, the social science literature (Cochrane, 1979).
educating girls and young women is considered a highly Economic theory provides several explanations for
effective means of lowering fertility and accomplishing why female education influences fertility. First, female
this goal (United Nations, 1995). Moreover, there is a schooling may increase the opportunity cost of child-
growing consensus that investments in the education of bearing and rearing among educated women (Becker,
young girls and women yield additional private and 1981; Schultz, 1981). Second, education may lower
social returns, including improved child health and fertility through improvements in child health and
nutrition outcomes (Schultz, 2002; Thomas et al., 1991). reduced rates of child mortality as women need to
Several empirical studies point to a robust, negative have fewer births to yield the same desired family size
(Lam and Duryea, 1999; Schultz, 1994a,b). Finally,
female schooling may affect fertility through knowledge
⁎ Corresponding author. Department of Economics, IUPUI, 425
and more effective use of contraceptive methods
University Boulevard, Indianapolis, IN 46202, United States.
Tel.: +1 317 278 7219; fax: +1 317 274 0097.
(Rosenzweig and Schultz, 1985, 1989) or by increasing
E-mail addresses: uosili@iupui.edu (U.O. Osili), female autonomy and bargaining power in fertility
longbr@gse.harvard.edu (B.T. Long). decisions (Mason, 1986). However, a survey of the
0304-3878/$ - see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.jdeveco.2007.10.003
58 U.O. Osili, B.T. Long / Journal of Development Economics 87 (2008) 57–75
existing literature suggests a need for caution in affected female educational attainment by exploiting
interpreting the observed relationship between female regional and age differences in the extent to which the
education and fertility as causal (Bledsoe et al., 1999). A policy affected investments in schooling. Second, the
negative association may arise due to omitted variables, paper links these changes in education to fertility
such as individual ability or household and community outcomes thereby testing whether exogenous changes
resources, which affect both schooling and fertility in education affect births in a causal manner. We employ
decisions. In addition, schooling opportunities often are instrumental variable methods to estimate the causal
not randomly placed in communities (Duflo, 2001; Pitt relationship between schooling and fertility. This paper
et al., 1993). Furthermore, if fertility choices lead to joins a growing number of studies that identify the
interruptions in schooling, then fertility may be an impact of education policies and programs in developing
endogenous variable within the context of schooling countries.2 In addition, the paper addresses the question
decisions (Angrist and Evans, 1999). of whether investments in education, specifically in
The ideal method to examine the issue of causality is to primary education, impact fertility as there is little
use an exogenous source of variation in schooling that is consensus about what level of education should be
not related to fertility outcomes. This paper does so by expanded to affect fertility outcomes.
focusing on the Universal Primary Education (UPE) Our results suggest that changes in schooling costs
program in Nigeria. Introduced in September 1976, the and the expansion of primary classrooms associated
UPE was a large-scale, nationwide program designed to with the Nigerian UPE had a substantial impact on
increase educational attainment. Funded by the federal female education and fertility before the age of 25. We
government, it provided tuition-free primary education provide evidence that female education has a strong,
and increased the number of primary school classrooms negative association with early fertility even after
and teacher-training institutions throughout the country, accounting for the possible endogeneity of the schooling
thereby marking a significant change in the educational decision. The paper also considers the rapid advances in
opportunities available to young Nigerian children. female schooling and demographic outcomes such as
During the UPE program, the number of primary school the state-level expansion in civil service employment
children in Nigeria increased from 4.4 million students in during this period.3 However, the results remain robust
1974 to 13.8 million by 1981 (Federal Office of Statistics, even with these additional controls.
Annual Abstract of Statistics, various years). 1 In This paper is organized as follows: Section 2 provides
September 1981, the UPE program ended when the background on the Nigerian UPE program and a
federal government handed over the financing of primary descriptive analysis of changes in educational outcomes
schools to states and regional governments. With reduced after the introduction of the policy. Section 3 presents
funding for primary schools, the reintroduction of school data sources and the empirical framework. Section 4
fees, and declining oil revenues in the 1980s, the gross examines the impact of UPE on educational outcomes
primary enrollment rate stagnated or fell in many states and fertility, and Section 5 presents conclusions.
after this date (Francis, 1998; World Bank, 2002).
In a seminal paper, Duflo (2001) examines the effect 2. Educational policy in Nigeria and the UPE
of a large-scale school construction program in Indone-
sia on educational attainment and wages by exploiting 2.1. Background on Nigeria and education before UPE
regional differences in program intensity and exposure
across birth cohorts induced by the timing of the Nigeria is an intriguing environment in which to
program. Following this strategy, using the UPE study the impact of female education on fertility. Some
program in Nigeria as an exogenous change in primary of the earliest research on the economic determinants of
schooling investments, this paper examines whether
increases in female schooling cause reductions in
fertility. First, the paper examines how the UPE program 2
For example, Case and Deaton (1999) use data from South Africa
to study the impact of increased resources on schooling outcomes.
Breierova and Duflo (2004) rely on an individual's date of birth and
1
The gross primary enrollment rate in Nigeria increased from 50.3 region of schooling to identify the impact of male and female
in 1975 to 120.7 in 1981. In comparison, in the five years before the schooling on child mortality and fertility. Angrist et al. (2002) study
UPE program commenced (1970–75), gross enrollment rates the impact of school vouchers in Columbia.
3
increased only by 17% from 43.7 in 1970 to 50.3 in 1975. By Unlike other related work, however, changes in contraceptive use
1985, several years after the program ended, the gross enrollment rate and legal statues, such as the legalization of abortion and anti-
had fallen to 91.8 (World Bank, 2005). discrimination laws, are less likely to be relevant in our environment.
U.O. Osili, B.T. Long / Journal of Development Economics 87 (2008) 57–75 59
Fig. 1. The 19 states of Nigeria in 1976. Source: Bray (1981).
fertility is based on survey evidence and field research carved into three states, was a forerunner in education. In
from Nigeria and other parts of sub-Saharan Africa (Van January 1955, the Western Region was also the first area
de Walle, 1965; Caldwell, 1968, 1977). In addition, total in Nigeria to experiment with universal education.4
fertility rates in sub-Saharan are among the highest in Tuition fees were abolished for all levels of primary
the developing world at 5.9 lifetime births per women schooling, and according to Awokoya (1952), the
(World Bank, 2005). A noteworthy feature of the education minister for the Western Region, advances in
Nigerian environment is the absence of a sustained education were “imperative” and “urgent.”5 Education
national family-planning program and government imbalances across regions were substantial in the decade
efforts to promote modern contraceptive methods that followed independence. The Northern, and to a lesser
(Caldwell et al., 1992). However, Nigeria's average extent the Eastern regions did not implement sustained
total fertility rate, at 5.6 lifetime births per woman, is universal education programs in the pre-independence
comparable to the overall mean for sub-Saharan Africa period. Specifically, in the Northern region, the legacy of
(UNICEF, 2005). the colonial policy to not interfere in Islamic religious
Prior to independence in 1960 from the United practices had limited the expansion of formal primary
Kingdom, Nigeria was divided into three administrative
units governed as semi-autonomous regions. Nigeria's 4
In 1843, the Methodist mission established the first primary school
regions were composed of diverse ethno-linguistic groups in the Western region of Nigeria, partially due to its proximity to the
often with distinct religious traditions. The Northern Atlantic coast (Fafunwa, 1974). Lagos, also formerly in the Western
region was a predominantly Muslim area while the Region and the capital of Nigeria after independence, also introduced
Eastern region was predominantly Christian (with both its own universal primary education program in January 1957 and had
religions present in the Western region). After indepen- achieved near-universal enrollments by the late 1960s.
5
The priority afforded to education was reflected in the Western
dence in 1960, four administrative regions were formed. Region's budget, with primary education consuming nearly 40% of
Each region had developed its own education policies the government's recurrent budget in comparison to 10-20% in the
during this period. The Western Region, which was later Northern Region of the country (Nwachukwu, 1985).
schooling. It was against this background that the national To capture the intensity of the program by state, the
UPE program was introduced. analysis focuses on the amount of federal capital funds
disbursed for classroom construction in 1976.9 Fig. 1
2.2. An overview of UPE provides a map of Nigeria in 1976.
Table 1 displays these amounts standardized by state
Nigeria's 1976 UPE nationwide initiative was charac- population. Throughout the paper, we map the variables
terized as “one of the most ambitious education projects in of interest to match the boundaries of the 19 states that
African history” due to the magnitude of the program in existed when the UPE program commenced in 1976.10
terms of government resources and the number of children As shown in Table 1, the states that received the
who were expected to benefit (Bray, 1981, p.1). Fueled by highest levels of federal capital funds for classroom
revenues from the oil boom of the time, the nationwide construction per capita were located outside the former
UPE Program was announced formally on October 1, Western region of Nigeria with the exception of
1974 and commenced in September 1976. As part of the Lagos.11 These were the states that had relatively low
UPE program, the Nigerian central government provided primary school enrollment rates and levels of educa-
tuition-free primary education nationwide. To enable the tional inputs prior to the UPE program. Throughout the
expansion of primary education, the government also rest of this paper, we use the term “high-intensity” to
recognized the need to construct a large number of new describe these states, which experienced a significant
primary school classrooms. According to the Nigerian expansion in educational inputs.
Third National Development Plan 1975–80, the country Our calculations suggest that changes in gross
planned for the provision of 150,995 new classrooms at enrollment ratios were also not distributed evenly across
the primary-school level (Federal Ministry of Economic the country.12 Using the projected state population for
Development and Reconstruction and Central Planning each Nigerian state, our calculations suggest that gross
Office, 1975). This represented a 1.4-fold increase in the enrollment rates increased much faster in the high-
number of primary school classrooms in the country in intensity states that experienced large increases in
1965. To finance this expansion, about 700 million naira federal capital expenditures and regional gaps appear
was disbursed to states by the Federal government for
primary classroom construction between 1974 and 1979.6
The UPE appeared to have a substantial impact on 9
Alternative measures of the intensity of the program include the
both male and female enrollments. The gross primary
number of planned classrooms in primary schools and teacher training
enrollment rate for boys increased from 60.3 in 1974 to institutions. However, information on these amounts is only available
136.8 in 1981. Similarly, although the gross female for the 12 states that existed in Nigeria in 1974 and not the 19 states
primary enrollment rate in Nigeria was only 40.3 in that existed by 1976 when the UPE was introduced.
10
1974, by the end of the program in 1981, it had risen to In 1965, Nigeria had 4 administrative areas (i.e. the Western,
Eastern, Northern, and Midwestern regions). At the time of the UPE's
104.7 (World Bank, 2002).7 These achievements are
announcement, these regions were split into 12 states. However, by
particularly impressive given that sub-Saharan and 1976, when the UPE program commenced, additional states were
northern Africa have some of the lowest levels of female created for a total of 19 states (the focus of this study). Currently,
educational attainment, and many young girls and Nigeria has 36 states and 1 Federal Capital Territory. The number of
women have little or no exposure to formal schooling.8 states has changed over time to improve equity in the revenue-sharing
system at the federal level. Federal government revenue allocations to
states and local governments are governed by a formula based on
6
Federal Office of Statistics, Social Statistics in Nigeria, 1979, population, need, and, to a lesser extent, derivation.
11
page 31. In 1976, the nominal exchange rate was N0.788 (naira) to $1 The population figures for Lagos are likely to be underestimated due
U.S. To meet the increased demand for teaching resources, the to significant population growth resulting from rural-to-urban migration.
government announced plans to expand existing teacher-training During the first decade after independence in 1960, metropolitan Lagos
institutions. According to Nwachukwu (1985), the UPE program was estimated to have experienced a growth rate of 14 per cent per annum
required about 80,000 new teachers, and the government had also (Lagos Executive Development Board, 1971). This would greatly reduce
planned for 6699 new classrooms in teacher-training institutions. the actual funds per capita amount. However, when Lagos is eliminated
7
The gross enrollment rate is calculated as the percentage of school- from the sample, the results of the paper do not change.
12
aged children who are enrolled in primary school in a given year. The The state population estimates used to standardize the expenditure
rate can be over 100.0% due to the fact that older or younger students numbers are based on the 1953 Population Census of Nigeria, the last
may be in the enrollment count and also due to grade repetitions. census conducted by the British Colonial Administration. Due to
8
During the first year of UPE, the number of students enrolled in significant controversies with the 1963 and 1973 censuses, we are less
Grade 1 increased by 82% to 3 million students. This growth confident in the population figures from these sources. However, as
exceeded predictions that only 2.3 million would enroll (Federal shown in the final columns, the ranking and relative amounts by state
Government of Nigeria, 1978-1979). are similar when using estimates based on the 1963 census.
Table 1
Federal capital funds allocated for primary school construction in 1976 (in naira)
State Region Funds allocated Using 1953 census population Using 1976 state population
estimates for towns projections based on 1963
census
Population Funds/capita Population Funds/capita
Low-intensity areas
Oyo Western 1,744,305 1,243,090 1.40 7,330,400 0.24
Ogun Western 321,524 166,274 1.93 2,182,600 0.15
Ondo Western 717,838 219,741 3.27 3,841,400 0.19
Lagos (Capital Region) Western 13,890,626 267,407 51.95 2,244,500 6.19
High-intensity areas
Anambra Eastern 8,342,532 213,561 39.06 5,061,500 1.64
Borno Northern 2,601,302 77,730 33.47 3,415,500 0.76
Kaduna Northern 11,116,441 145,440 76.43 5,168,500 2.15
Rivers Eastern 5,821,876 71,634 81.27 2,420,400 2.41
Imo Eastern 8,271,194 93,633 88.34 3,666,300 2.26
Kano Northern 12,131,038 130,173 93.19 8,126,800 1.49
Sokoto Northern 8,369,744 87,845 95.28 6,387,300 1.31
Kwara Northern 9,538,412 94,264 101.19 2,412,800 3.95
Bauchi Northern 2,973,215 29,075 102.26 3,421,500 0.87
Gongola Northern 5,005,510 47,643 105.06 4,894,700 1.02
Bendel Midwestern 10,062,666 76,092 132.24 3,462,300 2.91
Niger Northern 2,025,000 12,810 158.08 1,681,000 1.20
Plateau Northern 6,287,450 38,527 163.20 2,852,100 2.20
Benue Northern 3,175,804 16,713 190.02 3,463,300 0.92
Cross-river Eastern 10,256,206 46,705 219.60 4,218,300 2.43
Source: Federal Office of Statistics, Social Statistics of Nigeria (1979) p. 30. Information is not available on actual expenditures on classroom
construction. Population figures are from national censuses. Notes: The region refers to one of the 4 administrative areas defined in 1965 shortly after
independence. At the time of the program's announcement, Nigeria's four administrative regions had been split into 12 states. However, by 1976, when the
UPE program commenced, additional states were created for a total of the 19 states shown in this table. Throughout this paper, we map each variable of interest
to match the 19 states that existed when the UPE program commenced in 1976. The population estimates derive from the 1953 Population Census, the last
census conducted by the British Colonial Administration, which is considered the most accurate due to controversies surrounding the 1963 and 1973
Censuses. When mapping the 1953 town population numbers to the state boundaries of 1976, only towns with more than 15,000 people were included. This
suggests that the population estimates are more accurate for the more urbanized Western Region and that the funds per capita estimates may be slightly inflated
for the non-Western Regions. However, as shown in the final columns, the ranking and relative amounts by region are similar when using 1976 projected state
estimates from the 1963 census. The population estimates for Lagos are likely to be underestimated due to significant population growth resulting from rural-
to-urban migration. During the first decade after independence in 1960, metropolitan Lagos was estimated to have experienced a growth rate of 14 per cent per
annum (Lagos Executive Development Board, 1971). However, when Lagos is eliminated from the sample, the results of the paper do not change.
to have been substantially reduced during the years of As shown in the figures, the introduction of the nationwide
the UPE program.13 UPE had a major impact on student participation in
Figs. 2 and 3 summarize the state trends in the number primary school education and the number of primary
of students and primary schools before and during UPE schools in Nigeria, especially in the high-intensity, non-
with information from various years of the Annual Western states. The combined effects of the UPE program
Abstract of Statistics and Social Statistics in Nigeria.14 through the increase in capital funds for classroom
construction and the elimination of primary school tuition
13 fees appeared to been greatest in these states.
We calculate the gross enrollment rates in the Western region to be
0.65 in 1975–76 and 0.93 in 1980–81. For states outside the Western
region, our estimates suggest that the gross primary enrollment rate 3. Data and empirical framework
increased from 0.45 in 1975–76 to 0.96 in 1980–81. In summary, by
the time the program ended in 1981, many high-intensity states had 3.1. The data: The Nigerian Demographic Health Survey
experienced large changes in schooling inputs and enrollments.
14
Although the ideal trend to display is enrollment rates, we do not
have accurate population estimates for this time period because the
To study the impact of the UPE program, we rely
1973 population census in Nigeria was widely disputed and not mainly on the 1999 Nigerian Demographic Health Survey
publicly released. (NDHS). This dataset is a nationally representative survey
containing rich information on socioeconomic and

demographic variables for nearly ten thousand Nigerian
women. The NDHS survey is ideal for this study because
by design it provides reliable information on levels and
trends in education, fertility, and family-planning prac-
tices for a large number of women. Tables 2A and
Tables 2B provides summary statistics for the data. To
study the impact of the UPE on education and early
fertility, we use information from the NDHS on a
respondent's region of birth to link individuals to
educational policies and institutions. Tables 2A and
Tables 2B also include summary statistics on the 12
states that existed when the UPE program was announced
in 1974 and the 19 states that existed when the program
began in 1976.15 An important advantage of the NDHS is Fig. 2. Number of primary school students, 1970–1981. Source:
that it contains detailed information on religion and Federal Office of Statistics (various years), Annual Abstract of
ethnicity, and we can use these as control variables. Statistics. Notes: High-intensity states refer to states that experienced
the largest changes in schooling inputs during the program. These states
Although the NDHS survey has many important
are located largely outside of the Western region, while low-intensity
advantages, the data also have some shortcomings. First, refers to states of the Western Region (defined to include Lagos).
we would ideally like to know precisely when and where
an individual started and completed primary school. This possible measurement error, and the results are robust to
is because our analysis exploits geographic variation to excluding Lagos.
study the impact of the UPE. The NDHS survey has only Another shortcoming of the 1999 NDHS is that it
limited information on migration.16 Fortunately, two- provides only limited information on wages or income
thirds of the sample in the 1999 NDHS has never moved. at the household or individual level. Finally, because the
When comparing the characteristics of individuals who cohort born after the initiation of the UPE program
did and did not move, we do not find significant (1976–1981) was just reaching their mid-20s by 1999,
differences.17 In addition, recent studies on migration we cannot fully use them to study the impact of
patterns in Nigeria in the 1990s suggest that most education on fertility.
Nigerians do not move or only move within state. As
noted by National Population Commission (1998), “If 3.2. Empirical strategy #1: Differences-in-differences
migration is defined as moves across state boundaries —
most Nigerians can be classified as non-migrants. The To estimate the impact of UPE on educational and
only state with a sizeable share of migrants is Lagos fertility decisions, the paper utilizes two empirical
State, with 87% of its population migrated from other strategies that exploit the variation in which cohorts
states.” (p. 285). Therefore, the subsequent analysis will and in which geographical areas students were affected
display the results with and without Lagos to account for by UPE. The first strategy uses a differences-in-
differences (DD) technique.
By noting that the UPE occurred between 1976 and
15
As with Table 1, state population estimates are based on the 1953 1981 and had a larger impact on certain states (namely, the
Population Census in Nigeria, and we use only population of towns high-intensity states), we examine whether the introduc-
greater than 15,000 to maintain consistency in administrative boundaries. tion of universal primary education caused discontinuities
16
We can identify women who migrated, however, we are unable to in educational attainment and early fertility for the treat-
determine if a woman moved across or within a state, and if the move ment group relative to a control group. This approach has
occurred across states, we do not know in which state she lived during
her school-age years. If an individual moved to their current state of been used in many recent studies, including Duflo (2001).
residence before age six, then we can assume with some degree of The first difference we use is exposure to the UPE as
certainty that education was completed in that state. However, if the measured by age. In general, primary school lasts for six
individual moved after age six, then the educational opportunities she years in Nigeria, and we assume that most individuals
faced may have been in a different state. start school at age 6. Given the timing of the program,
17
We do not find significant differences in the mean years of schooling
completed or likelihood of completely primary-school between movers
UPE should have primarily affected individuals born
and non-movers. However, non-migrants tend to be younger and have between 1970 and 1975. However, there is a high
fewer children than movers. prevalence of underage and overage enrollments in first
enrollments in high-intensity states with educational

attainment increasing faster in comparison to smaller
effects in the low-intensity states. Therefore, our control
group comprises of individuals in the low-intensity
states that did not experience large growth in enroll-
ments or schooling inputs. These states serve to account
for any general trends that affected all states.
There are two ways to measure exposure to the UPE by
state. The first is just to use a dummy variable that is equal
to one if the woman grew up in a high intensity state. In
this case, the differences-in-differences estimation of the
impact of UPE on schooling can be described as:

Sijk ¼ a0 þ a1 Xijk þ a2 High Intensityk 4UPE Cohortj
ð1Þ
Fig. 3. Number of Primary Schools in Nigeria, 1970–1981. Sources: þ a3 High Intensityk þ a4 UPE Cohortj þ eijk
Nigeria Federal Office of Statistics (various years), Annual Abstract of
Statistics. Nigeria Federal Office of Statistics (various years), Social
where i indexes individuals, j indexes cohorts, and k
Statistics in Nigeria. Notes: High-intensity states refer to states outside
of the Western region, while low-intensity refers to states of the indexes states. Sijk is the years of schooling completed by
Western Region (defined to include Lagos, the former capital
territory).
Table 2A
Summary statistics of the 1999 NDHS data
grade in Nigeria.18 For this reason, the program may have Mean Std dev
affected individuals outside of the main age range. Number of children born (at the time of the survey) 2.16 (2.78)
Therefore, rather than using a cohort just a little older Number of births before age 25 2.35 (1.81)
than the target group for the control group, we use Number of births before age 25 2.67 (1.51)
women who were age 15 to 20 when the UPE was (conditional on births N 0)
First birth before age 16 0.14
initiated (born between 1956 and 1961); this group is less
First birth before age 18 0.25
susceptible to overage enrollment. The main tables Years of education: female 5.00 (4.71)
compare the outcomes of the group born 1970–75 (i.e. Completed seven or more years of education: female 0.34 (0.47)
the main UPE group) to the group born 1956–1961 (i.e.
out of the primary school age range before UPE). We also Year of Birth
Born 1982–1989 0.26
investigate the impact of UPE on additional cohorts
Born 1976–1981 0.22
(those born 1964–69 and 1976–81) as they may have Born 1970–1975 0.18
benefited from UPE due to overage enrollments and the Born 1964–1969 0.15
expansion in resources (the building of primary school Born 1958–1963 0.12
classrooms and an increased number of trained teachers). Born 1952–1957 0.07
Born 1946–1951 0.02
The second difference we use is state of residence. As
shown above, the introduction of the UPE to Nigeria had Religion and ethnicity
varying effects on the states within the country. Because Muslim 0.44
many non-Western, high-intensity states did not have Christian 0.54
tuition-free primary schooling prior to the nationwide Other religion 0.02
Hausa 0.24
UPE program, the program represented both an
Yoruba 0.22
expansion in schooling inputs and a reduction in the Ibo 0.14
cost of schooling. As a result, the national UPE program Other ethnicity 0.40
should have had a greater impact on primary- school
Residency characteristics
Low-intensity state (western region — 0.18
18 1965 definition)
According to a Multiple Indicator Cluster Survey (MICS)
High-intensity state (non-western region) 0.82
conducted by UNICEF in 1999, underage enrollments, overage
Number of observations 9810
enrollments, and grade repetitions are common in Nigeria. Of the
sample of 977 female students enrolled in first grade (primary one), Source: 1999 Nigerian Demographic and Health Survey. Notes: The
about 4% of the sample was age four, 19% were age five, 31% were sample includes all respondents in the 1999 NDHS. The average number
age six, 21% were age seven, and about 15% were above age seven. of births before age 25 is calculated for women age 25 and older.
Table 2B with data on the projected primary school classroom

1976 state characteristics (N = 19) construction detailed in Nigeria's Third National
Mean Std dev Development Plan. It is also highly correlated with the
Federal capital funds disbursed in 1976 to 6,455,404 (4,145,556) projected expansion of classrooms at teacher training
state for primary school classroom institutions.19
construction) in 1976 (in naira) Using the continuous program intensity measure, the
Average population (1953 census estimates) 160,465 (272,468)
differences-in-differences estimation of the impact of
Population (1963 census estimates) 2,843,827 (1,269,354)
Projected population in 1976 4,013,221 (1,773,927) UPE on schooling can be described as:
(based on 1963 census)
Projected enrollments in grade 1–1976 120.94 (53.06) Sijk ¼ b0 þ b1 Xijk þ b2 UPE Inputsk ⁎UPE Cohortjk
(in '000) þ b3 UPE Inputsk þ b4 UPE Cohortjk þ eijk
Actual enrollments in Grade 1–1976 157.48 (79.53)
(in '000) ð2Þ
Area in square miles 18,035 (14,477)
Share of females in civil service employment 0.02 (0.01) The variable “UPE Inputs” is measured as the per
Sources: Nigeria Federal Office of Statistics (various years), Annual capita federal funds disbursed for the primary school
Abstract of Statistics. Nigeria Federal Office of Statistics (various construction in the state where an individual was
years), Social Statistics in Nigeria. Notes: The 19 states in the sample educated. The parameter β2 is the reduced-form estimate
represent the 19 states that existed when UPE was introduced in 1976. of the effect of UPE. It measures whether individuals who
grew up in states that received more resources during UPE
individual i of cohort j in state k. Xijk includes controls for experienced increases in educational attainment larger
year of birth, religion, and ethnicity. The variable “High than women in states that did not receive as much.
Intensity” is a dummy variable equal to one if the woman The DD methodology can also be used to examine the
was educated in a high-intensity state (i.e. a non-Western impact of UPE on early fertility. The same regressions are
state). “UPE Cohort” is a dummy variable that captures run with the outcome being a measure of fertility rather
whether an individual was born between 1970 and 1975 than years of education.
and therefore would have been the age for primary school If women in the “UPE Cohort” who were educated in
while UPE program was implemented. The parameter α2 high intensity states had fewer births than women in the
is the reduced-form estimate of the effect of UPE. We control groups, then we would expect that differences in
expect that the change in schooling outcomes should be fertility are due to the increased educational attainment
larger in the high-intensity states for the “UPE Cohort,” associated with the UPE.
and therefore, α2 is expected to be positive. In particular, it
measures whether individuals in high-intensity states, 3.3. Empirical strategy #2: Instrumental variable
who experienced large changes due to the UPE program, approach
also experienced more rapid growth in schooling in
comparison to individuals in the low-program intensity To determine the causal impact of education on
states, who did not experience much change due to UPE. fertility, we use the instrumental variables approach. As
A more detailed way to measure program intensity is noted earlier, an important issue within the existing
to use the 1976 per capita amount of federal funds literature concerns the causal interpretation of the effect
disbursed to each state for classroom construction of education on fertility. If education has a causal effect
(shown in Table 1). This measure of UPE program on fertility, then an expansion in schooling should
intensity has several significant advantages in our induce lower fertility rates, holding other variables
context. First, the measure is available for all 19 states constant. However, unmeasured individual, household,
that existed when the UPE program commenced in and community-level resources may affect both educa-
1976. In contrast, data on planned classroom construction and fertility decisions. For example, an increase in
tion, which previous studies have used, is only available the level of economic development may lead to higher
for the 12 states that existed when the program was first educational attainment and lower fertility. Furthermore,
announced in 1974. Second, the federal capital funds
measure represents actual amounts disbursed to the 19
states rather than the planned allocation of resources. We have also found that data on projected classroom construction
detailed in Nigeria's Third National Development Plan (available for
This allows us to track more closely changes in only 12 states) are highly correlated with actual changes reported in
schooling inputs that took place during the course of the number of classrooms in each state based on the Federal Office of
the program. Finally, our measure is strongly correlated Statistics, Annual Abstract of Statistics.
education may serve as a proxy for unobservable on investments in female schooling. In the initial
factors, such as ability, cognitive skills, motivation, analysis, we limit our sample to the women of two
and parental background, and these factors may be cohorts: those born between 1956 and 1961 (ages 15–20
important determinants of a woman's fertility choices when the program started) and those born between 1970
(Thomas, 1999). Ignoring these factors would lead to and 1975 (ages 1–6 when the program started).
biased estimates of the impact of education on fertility Individuals born between 1970 and 1975 (i.e. the
within the context of simple Ordinary Least Square “UPE cohort”) are likely to have been the primary
(OLS) estimation. The DD methodology described beneficiaries of the program. Because primary school
above attempts to isolate the causal impact by including lasts for six years in Nigeria, an individual who was born
a control group, but an instrumental variables approach in 1970 and enrolled in primary school in 1976 (and
provides an additional option. attended school continuously) was most likely exposed
Valid instruments are variables that affect the level of to a maximum of five years of the UPE program, which
educational attainment but have no direct impact on ended in 1981. In contrast, individuals who were born
fertility. Again, our key measure of program intensity is between 1956 and 1961 are likely to have been too old
federal disbursed funds per capita for primary school to benefit from the educational policy. Other cohorts
construction in the state where an individual was also may have benefited from the UPE due to underage
educated. If we assume that the UPE program had no and overage enrollments and so are not used as control
direct effect on fertility, other than through its effect on groups; they are examined in a later section.
educational attainment, then we can use exposure to the The left side Table 3 reports the effect of UPE on the
UPE program as an instrument for schooling. The age number of years of schooling completed using the
and state of schooling determines an individual's differences-in-differences methodology. Panel A (the top
exposure to the UPE program. The key dependent panel) uses a dummy variable for growing up in a state
variable is the number of children born before age 25. that had a high level of UPE intensity. Panel B (the
The instrumental variable approach is described as bottom panel) instead uses the 1976 per capita federal
follows: funds disbursed for primary school construction as the
measure of the UPE program intensity.20 We present
several specifications of the empirical model. The
Nijk ¼ ai þ BXijk þ Sijk þ eijk where
ð3Þ baseline model (specification 1) includes controls for
Sijk ¼ f ðUPE Inputsk Þ þ mijk background such as religion and ethnicity. Because one
might expect more growth in educational attainment in
In Eq. (3), Nijk represents the number of children areas where fewer students were in school, we also
born before age 25 to an individual i of cohort j living in control for the female share of total primary school
state k and Sijk is instrumented for using f (UPE Inputsk). enrollment in the state in 1970 to deal with the possibility
More formally, OLS estimates may lead to biased of mean reversion. We also recognize that there may
estimates if ɛijk is correlated with schooling due to have been other programs in place to encourage women
unmeasured ability and other factors, such as family to become educated or enter the labor force. During this
background, social norms, or community wealth. We period in Nigeria, public-sector employment as well as
assume that exposure to the program (which is based on the wages of civil servants significantly expanded. This
age and region of residence) is correlated with schooling expansion in the federal civil-service labor force may be
outcomes after other explanatory variables are included. correlated with the timing of the UPE program, and so to
Eqs. (1) and (2) above form the basis for the first-stage account for this factor, the baseline model also controls
regression represented by f(UPE Inputsk). the time-varying share of female civil service employ-
ment in the state of residence.
4. The impact of the UPE
4.1. Differences-in-Differences analysis: The impact on 20

To obtain per capita amounts, we use state population estimates
schooling and fertility from the 1953 census. We use this earlier census because the 1963 and
1973 censuses suffered from considerable controversy, and so we
In our empirical analysis, we first examine the impact have much less confidence in those numbers. In fact, the 1973 census
was not publicly released. However, when we re-calculate the per
of the UPE program on female educational levels. capita amount using 1976 projected population numbers (based on
Before addressing the larger question of how schooling the 1963 census) and re-estimate our regression models, the results do
affects fertility, these results study the effect of the UPE not differ from our main measure using the 1953 population numbers.
Table 3
The impact of the UPE program — DD analysis
A. Program intensity measured with a dummy variable for high-intensity (non-Western) states
Dependent var Years of schooling Number of kids before age 25
Baseline Add 1970 state Add year Baseline Add 1970 state Add year
FE FE FE FE
(1) (2) (3) (4) (5) (6)
Born 1970–75 ⁎ 1.632⁎ 1.573⁎ 1.537⁎ − 1.110⁎⁎⁎ − 1.142⁎⁎⁎ −1.086⁎⁎⁎
High-intensity state (1.77) (1.76) (2.22) (4.35) (4.49) (5.03)
Born 1970–75 − 0.297 −1.188 3.676⁎⁎⁎ 3.774⁎⁎⁎
Dummy variable (014) (0.61) (4.35) (4.38)
High-intensity − 0.605 0.766⁎⁎⁎
State dummy variable (0.64) (3.28)
R-squared 0.383 0.393 0.404 0.102 0.110 0.130
Observations 2646 2646 2642 2646 2646 2646
B. Program intensity measured by 1976 per capita federal funds disbursed for primary school construction
Dependent var. Years of schooling Number of kids before age 25
FE FE FE FE
(1) (2) (3) (4) (5) (6)
Born 1970–75 ⁎ Classroom 0.007 0.007⁎ 0.008⁎⁎ − 0.003⁎⁎ − 0.003⁎⁎ −0.003⁎⁎⁎
Construction funds per capita (1.61) (1.72) (2.43) (2.00) (2.24) (2.91)
Born 1970–75 1.903 1.144 1.874⁎⁎⁎ 1.922⁎⁎⁎
Dummy variable (1.41) (0.90) (2.80) (2.78)
Classroom − 0.001 0.012⁎⁎ 0.012⁎⁎ 0.002 − 0.001 −0.001
R-squared 0.383 0.395 0.407 0.096 0.104 0.125
Observations 2646 2646 2646 2646 2646 2646
Treatment: born 1970–1975 (age 1–6 in 1976); control: born 1956–1961 (age 15–20 in 1976).
⁎ Significant at 0.1 level. ⁎⁎ Significant at 0.05 level. ⁎⁎⁎Significant at 0.01 level. Notes: t-statistics are shown in parentheses. The t-statistics
reported are based on standard errors that are clustered at the year ⁎ state level. The baseline models include dummy variables for religion (Muslim,
Catholic, Protestant, Other Christian, Traditional Religion with “Other” being the left out group), ethnicity (Hausa, Yoruba, and Igbo with “Other”
being the left out group), the female share of total primary school enrollment in 1970 in the state, this variable is also interacted with a dummy variable
for being born 1970–75, and the proportion of civil servants in the state who were female the year the individual was age 6. State fixed effects refer to
the 12 states that existed in Nigeria in 1970 when the program was announced while classroom construction funds are measured at the 1976 state
level. The program intensity measure is available for all the 1976 states (19 states) that existed at the onset of the UPE program.
Even with all the controls discussed above, there still unobserved state characteristics, such as pre-UPE school
may be unobserved state-level heterogeneity. This availability and teacher quality, and initial differences in
concern is important, because the high-intensity UPE the level of economic development in the state, which
states that experienced large changes in enrollments and may affect educational attainment. Finally, specification 3
schooling inputs during the program also had significantly includes year-of-birth fixed effects to deal with govern-
lower educational and social indicators prior to the ment programs, policies, and other trends that took place
introduction of the UPE. For example, school availability during this period. We report robust t-statistics that are
and teacher quality tended to be lower in the high- clustered at the state level and year of birth.
intensity states, particularly in the northern states of As shown in the first three columns of Panel A of
Nigeria. Therefore, the second specification includes state Table 3, the UPE is estimated to have increased
fixed-effects to deal with time-invariant, unobserved educational attainment. The results show a sizeable and
heterogeneity at the state-level.21 The fixed-effects statistically significant effect of the program whether we
regressions allow us to “sweep out” time-invariant, measure exposure to UPE by the high-intensity dummy
variable or, as in Panel B, classroom construction funds
21
The state fixed effects that we include refer to the state of residence per capita, a continuous measure of investments due to
as a child and are based on the 1970 boundaries of Nigerian states. UPE. Our preferred estimates are shown in specification 3
of each panel. These results include all of the controls and 4.2. DD Analysis: Evidence in favor of the identification
both state and year of birth fixed-effects.22 From Panel A, assumption
exposure to the UPE program is estimated to have
increased the years of completed schooling by 1.54 for One interesting question to consider is whether other
women age 1–6 in 1976 in comparison to the control cohorts, beyond the main cohort that was primary
group (age 15–20 in 1976). This represents a 0.32 school age during the UPE, also benefited from the
standard deviation gain in schooling years completed and program. Table 4 presents results that examine this issue
suggests that the UPE had a sizeable effect on schooling and provide a robustness check. If the UPE really caused
outcomes given that the average years of schooling in the the effects on educational attainment and fertility, we
NDHS sample is very low (5.0 years). Based on the results should find smaller positive effects for the additional
in Panel B using classroom construction funds to measure cohorts (the farther away from 1970, the smaller
the treatment, we find a $100 naira increase in disbursed effects). Similar to the earlier results, the control group
funds for primary school construction per capita increased is women born between 1956 and 1961.
the number of years of schooling completed by 2 years.23 Table 4, Panel A shows results for women born
This represents a 0.42 standard deviation gain in between 1962 and 1965 who would have been between
schooling years completed. the ages of 11 and 14 when the program started in 1976.
Table 3 also examines the impact of the UPE program The impact of the UPE is likely to be limited for this
on fertility. Ideally, we would like to observe completed group based on their age. However, to the extent that
fertility. However, since the UPE cohort was a relatively overage enrollments occurred, this group may have
young group at the time of the 1999 NDHS survey, we enjoyed some benefits from the program. In Panel A,
limit our analysis to early fertility indicators. Our key although the coefficients are positive for educational
dependent variable here is the number of children born attainment and negative for fertility, none of them are
before age 25. Otherwise, the specifications follow the statistically significant suggesting that the UPE did not
same pattern as the educational attainment estimates. have a strong effect on this cohort.
The baseline estimates in specification 4 of each panel Table 4, Panel B focuses on the cohort of individuals
show a robust, negative impact of the UPE on the number who were born between 1966 and 1969 and were age
of births. Again, our preferred estimates include year of seven to ten when the program was launched. This
birth and state fixed- effects and are shown in specifica- group is particularly interesting because it is very likely
tion 6. In Panel A, we find that the UPE program reduced they also benefited from the UPE program but mainly in
the number of births by 1.09. This represents a 0.6 the latter years of primary schooling or as a result of
standard deviation reduction in the number of births. overage enrollments. From Panel B, Column 3, we find
From Panel B, we estimate that a 100 naira increase in some evidence that these women who were educated in
classroom construction funds per capita associated with high-intensity UPE states also enjoyed higher educa-
the UPE program reduced the number of births by 0.4 tional attainment compared to those of unaffected
(Column 6). Given that average number of children born cohorts. This point justifies our decision not to use
before age 25 is 2.35, this represents a 16% reduction in this cohort as the control group for the main results in
early births. We interpret these results as initial evidence Table 3 (instead opting to use women age 15 to 20 in
that the UPE program had an impact on fertility through 1976). We also examine fertility outcomes for this
its effect on female schooling.24 cohort, but the UPE is not found to have impacted the
number of children before age 25.25
22
The dummy variables “Born 1970-75” and “High-intensity state” Because the program had a limited duration (1976–81),
are not shown in specifications 2, 3, 5, and 6 because their it is also possible to examine the impact of exposure to the
interpretation changes with the inclusion of state fixed effects and UPE program for cohorts educated after the program
the year of birth fixed effects.
23
In 1976, the nominal exchange rate was N 0.788 (naira) to $1 U.S. ended. This analysis is shown in Panel C. Individuals who
This suggests that a 100 naira increase corresponded to about $130 would have been age six after the program ended (born
increase (World Bank, 2002).
24
While it is also possible that the number of children born before
age 25 may also be lower among the UPE cohort due to other omitted
25
variables that are correlated with the program such as birth control, One caution in interpreting these results is that children who were
changes in the diffusion of modern contraceptive methods are not born between 1965 and 1969 and were between ages 7 and 12 when
likely to be important here as in other developing countries because the program commenced may have been influenced by the Nigerian
the prevalence of such methods remains relatively low in Nigeria Civil War (1969–1970). This factor may have led to some schooling
(Caldwell et al, 1992). interruptions in certain states for this cohort.
Table 4
Evidence in favor of the identification assumption: UPE on unaffected cohorts
A. Born before UPE was announced — born 1962–1965 (age 11–14 in 1976); control group: born 1956–1961 (age 15–20 in 1976)
FE FE FE FE
(1) (2) (3) (4) (5) (6)
Born 1962–65 ⁎ Classroom 0.005 0.005 0.006 −0.002 − 0.001 − 0.001
R-squared 0.350 0.357 0.374 0.031 0.050 0.060
Observations 1723 1723 1723 1723 1723 1723
B. Born shortly before UPE was announced — born 1966–1969 (age 7–10 in 1976); control group: born 1956–1961 (age 15–20 in 1976)
FE FE FE FE
(1) (2) (3) (4) (5) (6)
Born 1966–69 ⁎ Classroom 0.007 0.006 0.006⁎ −0.001 − 0.001 − 0.001
R-squared 0.361 0.372 0.385 0.037 0.052 0.062
Observations 1956 1956 1956 1956 1956 1956
C. Born after UPE program ended — born 1976–1981; control group: born 1956–1961 (age 15–20 in 1976)
Dependent var. Years of schooling
Baseline Add 1970 state Add year
FE FE
(1) (2) (3)
Born 1976–81 ⁎ Classroom 0.003 0.004 0.004
Construction funds per capita (0.67) (0.86) (1.26)
R-squared 0.425 0.447 0.456
Observations 2936 2936 2936
⁎ Significant at 0.1 level. ⁎⁎ Significant at 0.05 level. Notes: t-statistics are shown in parentheses. The t-statistics reported are based on standard errors
that are clustered at the year ⁎ state level. The baseline models include dummy variables for religion (there are six religious categories, including
Muslim, Catholic, Protestant, Other Christian, Traditional Religion with “Other” being the left out group), ethnicity (Hausa, Yoruba, and Igbo with
“Other” being the left out group), the female share of total primary school enrollment in 1970 in the state, this variable is also interacted with a dummy
variable for being born 1970–75, and the proportion of civil servants in the state who were female the year the individual was age 6. State fixed effects
refer to the 12 states that existed in Nigeria in 1970 while classroom construction funds are measured at the 1976 state level. The program intensity
measure is available for all the 1976 states (19 states) that existed at the onset of the UPE program.
after 1976) may have also enjoyed some benefits if the results in Panel C are positive in terms of years of
program had a long-lasting impact on schooling due to schooling, the results are not statistically significant. We
classroom construction and teacher-training efforts. do not present fertility outcomes for this group because
However, the end of the UPE program also marked most were too young at the time of the 1999 NDHS
significant reduction in funding resources available to survey to construct a complete measure of births before
primary schools. In most states, primary school tuition age 25.
fees were reintroduced after 1981 as the federal
government handed over control of the schools to states. 4.3. DD analysis: Further evidence in favor of the
Furthermore, the federal government no longer provided identification assumption
grants to states for teacher salaries and training. With
these caveats in mind, we do not find evidence that there One concern with the analysis is that schooling
were long-term effects of the UPE program. Although the outcomes may be higher for the treated cohort because
of reasons other than the UPE program. Instead, they may funds per capita to measure UPE intensity, and likewise,
be due to a pre-existing differential trend in educational no statistically significant results are found for the variable
attainment across states. To address this concern, in of interest.
Table 5, we compare the 1956–61 cohort to the 1950–55 Table 6 presents results for all women born between
cohort. Neither group should have been affected by UPE. 1949 and 1975. These results allow us to further compare
However, if schooling levels were increasing faster in the several cohorts who should not have been exposed to the
high-intensity states prior to the UPE program, then we program to cohorts that may have been exposed to the
should find a spurious significant coefficient for the program either at the beginning or towards the end of
“younger” unaffected cohorts in high-intensity states. As primary school. In particular, women who were born in
shown in Panel A, however, when we compare successive the 1950s should have had no exposure to the program
cohorts who were unaffected by the UPE program, we (i.e. the control group, born 1949–1961), while women
find no evidence that schooling was increasing faster in born in the 1960s may have had some exposure to the
the high-intensity areas prior to the program. Similar program in the later years of primary school or due to
analysis is shown in Panel B using classroom construction overage enrollments. From the results, we find that
Table 5
Additional evidence in favor of the identification assumption — robustness check
A. Program intensity measured with a dummy variable for high-intensity (non-Western) states
FE FE FE FE
(1) (2) (3) (4) (5) (6)
Born 1956–61 ⁎ High-intensity − 0.447 − 0.400 − 0.765 0.230 0.210 0.221
State (0.42) (0.38) (0.93) (0.72) (0.64) (0.76)
Born 1956–61 − 0.789 − 0.555 0.772 0.807
Dummy variable (0.36) (0.27) (0.84) (0.89)
High-intensity −0.104 0.681⁎⁎
State dummy variable (0.16) (2.55)
R-squared 0.253 0.273 0.291 0.026 0.034 0.046
Observations 1552 1552 1552 1552 1552 1552
B. Program intensity measured by 1976 per capita federal funds disbursed for primary school construction
FE FE FE FE
(1) (2) (3) (4) (5) (6)
Born 1956–61 ⁎ Classroom 0.0003 − 0.0003 − 0.002 0.0001 0.0007 0.0007
Born 1956–61 − 1.599 − 1.224 0.976 1.043
Dummy variable (1.01) (0.93) (1.23) (1.35)
Classroom 0.0001 0.003 0.005 0.0001 − 0.005 − 0.005⁎
R-squared 0.253 0.273 0.291 0.016 0.036 0.047
Observations 1552 1552 1552 1552 1552 1552
False treatment: born 1956–1961 (age 15–20 in 1976). Control: born 1950–1955 (Age 21–26 in 1976).
⁎ Significant at 0.1 level. ⁎⁎ Significant at 0.05 level. Notes: t-statistics are shown in parentheses. The t-statistics reported are based on standard errors
that are clustered at the year ⁎ state level. The baseline models include dummy variables for religion (Muslim, Catholic, Protestant, Other Christian,
Traditional Religion with “Other” being the left out group), the female share of total primary school enrollment in 1970 in the state, this variable is
also interacted with a dummy variable for being born 1970–75, and the proportion of civil servants in the state who were female the year the
individual was age 6. State fixed effects refer to the 12 states that existed in Nigeria in 1970 while classroom construction funds are measured at the
1976 state level. The program intensity measure is available for all the 1976 states (19 states) that existed at the onset of the UPE program.
Table 6 adults. While most of the households in our sample have

The impact of UPE by year of birth control: born 1949–1961 (age 15–26 never moved, among movers we do not know precisely
in 1976)
where the woman was educated. While most movers
Dependent var Years of Number of kids appear to stay within their state birth, evidence from
schooling before age 25
Nigeria suggests that Lagos is the only state with a
(1) (2) significant share of migrants from other Nigerian states
Born 1962 or 1963 ⁎ classroom 0.0029 −0.0003 (National Population Commission, 1998). Therefore,
Construction funds per capita (0.73) (0.22) Table 7 displays the results found in previous tables with
Born 1964 or 1965 ⁎ Classroom 0.0085 −0.0027⁎
and without Lagos to account for measurement error that
Construction funds per capita (1.33) (1.94)
Born 1966 or 1967 ⁎ Classroom 0.0039 −0.0021 could result from migration patterns. As shown in
Construction funds per capita (1.11) (1.16) Panel A, the main results for the cohort age 1 to 6 in
Born 1968 or 1969 ⁎ Classroom 0.0055⁎ −0.0024 1976 are robust to the exclusion of Lagos from the
Construction funds per capita (1.87) (1.39) sample. In fact, as theory would predict, the results
Born 1970 or 1971 ⁎ Classroom 0.0096⁎⁎ −0.0022⁎
become stronger for both schooling outcomes and
Construction funds per capita (2.47) (1.75)
Born 1972 or 1973 ⁎ Classroom 0.0062⁎ −0.0018 fertility when excluding this group that may suffer
Construction funds per capita (1.76) (1.29) with measurement error. The same is true in Panel B for
Born 1974 or 1975 ⁎ Classroom 0.0030 −0.0024⁎ the additional cohort shown to have increased their
Construction funds per capita (0.89) (1.73) educational attainment due to UPE (those age 7 to 10 in
R-squared 0.3883 0.0820
1976). Once we exclude women from Lagos in Column
Observations 5135 5135
2, the estimate increases in size and statistical
⁎ Significant at 0.1 level. ⁎⁎ Significant at 0.05 level. Notes: t-statistics
significance.
are shown in parentheses. The t-statistics reported are based on standard
errors that are clustered at the year ⁎ state level. All models include
dummy variables for the two-year groups of the interactions (1962–63, 4.4. Analysis: The impact on fertility
1964–65, 1966–67, 1968–69, 1970–71, 1972–73, and 1974–75),
religion (Muslim, Catholic, Protestant, Other Christian, Traditional We now focus on the impact of the UPE on fertility
Religion with qOtherq being the left out group), ethnicity (Hausa, using the instrumental variables strategy. The ideal
Yoruba, and Igbo with qOtherq being the left out group), the female
share of total primary school enrollment in 1970 in the state, this
candidate for an instrumental variable would affect
variable is also interacted with dummy variables for being born 1962– schooling outcomes but have no direct impact on
65, 1966–69, and 1970–75, the proportion of civil servants in the state fertility. As shown earlier, year of birth interacted with
who were female the year the individual was age 6, and state and year the state classroom construction funds per capita had a
fixed effects. State fixed effects refer to the 12 states that existed in
positive and significant impact on years of completed
Nigeria in 1970 while classroom construction funds are measured at the
1976 state level. The program intensity measure is available for all the schooling. Therefore, we use this measure as an
1976 states (19 states) that existed at the onset of the UPE program. instrument for schooling in our analysis. However,
before displaying the IV result, we begin by estimating
OLS regressions, which ignore the concern about the
endogeneity of the schooling decision. Table 8, Panel A
cohorts born in the 1950s did not experience more reports OLS estimates of the effect of education on
education in the high-intensity areas relative to the low- fertility, measured as the number of children born before
intensity areas suggesting that they were unaffected by the age 25. Similar to the order of the specifications in the
UPE program. In addition, except for the cohort born above tables, all models include controls for religion and
1964–65, we do not find that fertility was decreasing ethnicity, a control for the female share of total primary
faster in the high-intensity areas for cohorts born prior school enrollment in 1970 in the state, this variable
to the UPE program. Consistent with the earlier findings, interacted with a dummy variable for being in the
the UPE program had the largest impact on women Treatment cohort, and the proportion of civil servants in
born between 1970 and 1975. As expected, the largest the state who were female the year the individual was
impact on education was for the cohort born 1970–1971, age 6. In Column 2, we add 1970 state fixed effects.
the group of women who had the longest exposure to Finally, in Column 3, we also add year of birth fixed
UPE. effects.
Table 7 displays an important set of additional As shown in Table 8A, Panel A, the OLS estimates
robustness checks. Our analysis is based on linking the suggest that increasing female schooling had a small,
resources and educational experiences of women while but precisely measured negative impact on early
they were children to schooling and fertility outcomes as fertility. From these results, a one-year increase in
Table 7
The impact of UPE by migration (including and excluding Lagos) — robustness check
A. Main treatment and control groups: treatment group: born 1970–1975 (Age 1–6 in 1976); control group: born 1956–1961 (age 15–20 in 1976)
Full sample Excl. Lagos Full sample Excl. Lagos
(1) (2) (3) (4)
Born 1970–75 ⁎ Classroom 0.008⁎⁎ 0.011⁎⁎⁎ − 0.003⁎⁎⁎ − 0.004⁎⁎⁎
Construction funds per capita (2.43) (3.22) (2.91) (3.39)
R-squared 0.407 0.409 0.125 0.123
Observations 2646 2503 2646 2503
B. Born shortly before UPE was announced: treatment group: born 1966–1969 (age 7–10 in 1976); control group: born 1956–1961 (age 15–20 in
1976)
Full sample Excl. Lagos Full sample Excl. Lagos
(1) (2) (3) (4)
Born 1966–69 ⁎ Classroom 0.006⁎ 0.007⁎⁎ −0.001 −0.002
Construction funds per capita (1.84) (2.23) (0.89) (0.97)
R-squared 0.385 0.380 0.062 0.057
Observations 1956 1840 1956 1840
⁎ Significant at 0.1 level. ⁎⁎ Significant at 0.05 level. ⁎⁎⁎Significant at 0.01 level. Notes: The t-statistics reported are based on standard errors that are
clustered at the year ⁎ state level. The t-statistics are clustered at the year ⁎ state level. All models include dummy variables for religion and ethnicity, a
control for the female share of total primary school enrollment in 1970 in the state, this variable interacted with a dummy variable for being in the
Treatment cohort, the proportion of civil servants in the state who were female the year the individual was age 6, state fixed effects, and year of birth
fixed effects. The results that exclude Lagos allow us to focus on women who did not migrate from the state in which they were educated. State fixed
effects refer to the 12 states that existed in Nigeria in 1970 while classroom construction funds are measured at the 1976 state level. The program
intensity measure is available for all the 1976 states (19 states) that existed at the onset of the UPE program.
female schooling is associated with a 0.11 reduction in We also compare our estimates to a number of recent
the number of births. Starting at the mean, this studies that have examined the relationship between
represents about a 5% reduction in fertility. fertility and female schooling in developing countries.
In Panel B, we present the IV results, which account for Ainsworth et al. (1996) use data from the Demographic
the endogeneity of the schooling decision by instrument- and Health Surveys for 14 sub-Saharan African
ing for years of schooling using UPE primary classroom countries and find that schooling has a negative and
construction funds per capita in the state of education significant association with the number of children ever
interacted with year of birth indicators. The IV regressions born in 13 out of the 14 countries included in their
suggest that a one-year increase in female schooling study. Although they do not attempt to instrument for
reduces fertility by 0.26 to 0.48 births — close to an 11 schooling to deal with possible endogeneity, their point
to 19% reduction in fertility depending on the speci- estimates on the impact of an additional year on the
fication of interest. These results suggest that the OLS number of children ever born range from − 0.06 to
estimates may underestimate the magnitude of the effect − 0.13. Using cross-country panel data, Schultz (1998)
of schooling on fertility. We compare the differences estimates that a one-year increase in adult female years
between our OLS and IV estimates and find that the of schooling reduces fertility by 13% or about 0.5
differences between the two estimates are statistically children per woman. Schultz (1994a,b) also finds that
significant.26 female education has a negative effect on fertility and
population, while family planning and other variables
do not appear to have a consistent, negative effect,
although the estimates do not address concerns about
26
Based on standard Wu-Hausman test comparisons, we find that
endogeneity. Relative to these earlier studies, our OLS
there are significant differences (at the 10% level) between the IV estimates are similar, but when we account for
coefficients in Panel B and OLS estimates in Panel A. endogeneity, we estimate a larger effect.
Table 8
The impact of education on fertility — instrumental variable estimates
A. OLS estimates
Baseline Add 1970 state Add year of birth
Fixed effects Fixed effects
(1) (2) (3)
Years of education −0.111⁎⁎⁎ − 0.113⁎⁎⁎ − 0.109⁎⁎⁎
(14.08) (14.35) (14.35)
R-squared 0.163 0.172 0.187
B. Instrumental variables estimates instruments used: year of birth dummies ⁎ state capital allocation for classroom construction
Baseline Add 1970 state Add year of birth
(1) (2) (3)
Years of education −0.444⁎⁎⁎ − 0.475⁎⁎⁎ − 0.263⁎⁎
(3.70) (3.99) (2.28)
Overidentification test (P-values) 0.086 0.032 0.281
R-squared 0.428 0.393 0.628
Treatment: born 1970–1975 (age 1–6 in 1976); control: born 1956–1961 (age 15–20 in 1976). Dependent variable: number of kids before age 25.
⁎ Significant at 0.1 level. ⁎⁎ Significant at 0.05 level. ⁎⁎⁎ Significant at 0.01 level. Notes: t-statistics are shown in parentheses. The t-statistics
reported are based on standard errors that are clustered at the year ⁎ state level. All models include dummy variables for religion (Muslim, Catholic,
Protestant, Other Christian, Traditional Religion with “Other” being the left out group), ethnicity (Hausa, Yoruba, and Igbo with “Other” being the left
out group), the female share of total primary school enrollment in 1970 in the state, this variable is also interacted with a dummy variable for being
born 1970–75, the proportion of civil servants in the state who were female the year the individual was age 6, and state and year fixed effects. State
fixed effects refer to the 12 states that existed in Nigeria in 1970 while classroom construction funds are measured at the 1976 state level. The program
intensity measure is available for all the 1976 states (19 states) that existed at the onset of the UPE program.
4.5. The impact of UPE on later fertility should note that for women age 45 and above, about 60%
of births occurred before age 30.
In the tables above, we have presented evidence on To examine the impact of the program on later births,
early fertility outcomes. One final concern is that early we use the 2003 NDHS, which allows us to examine
fertility may be less indicative of final fertility outcomes fertility outcomes four years later for the UPE cohort.
for women. Several studies (Schultz, 1994a,b, 1998) These results suggest that the UPE program had a
examine completed fertility, which is typically measured robust, negative impact on births before 30. The OLS
as the number of births for women age 45. The UPE estimates show that an additional year of schooling
cohort was relatively young at the time of the survey so it reduces the number of births before age 30 by 0.11 while
is not possible to investigate fertility to that age. the IV estimates suggest a reduction of 0.36 in the
However, among older cohorts, we observe that there number of births before age 30. These estimates are even
is a significant decline in births after age 30, with a very larger than our results using fertility up to age 25 as the
steep decline in births after age 35. Furthermore, using outcome. The complete results are shown in the
the retrospective data available in the NDHS data, Appendix. We do not include the 2003 NDHS in the
among women who are no longer in their reproductive main analysis because it has some disadvantages due to
years, we find that that early fertility is, in fact, strongly higher levels of migration observed in the data.
correlated with completed fertility. In particular, among
women age 45 and above, the correlation coefficient 5. Conclusions
between births before age 25 and completed fertility is
0.58. In addition, the correlation coefficient between In this paper, we investigate the causal link between
births before age 30 and lifetime births is 0.74. We education and fertility using a large- scale policy
experiment from Nigeria. Results from Nigeria suggest Appendix A. The impact of UPE on later fertility
that the change in education policy had a significant impact
on both female education and fertility decisions. At the Appendix Table 1
mean, for each additional 100 naira per capita spent on The Impact of the UPE program — DD Analysis
primary school classroom construction in 1976, we es- A. Program intensity measured with a dummy variable for high-intensity
timate a 2-year increase in educational attainment. These (non-western) states
results are robust to different specifications, and tests of Baseline Add 1970 state Add year
alternate explanations lend additional support to the idea FE FE
that the impact of schooling was due to the UPE program. (1) (2) (3)
Under the assumption that exposure to the UPE Born 1970–75⁎ −1.505⁎⁎⁎ −1.460⁎⁎⁎ −1.407⁎⁎⁎
program is a valid instrument for schooling, we construct High-intensity state (4.19) (4.32) (5.29)
IVestimates of the impact of female education on fertility. Born 1970–75 3.405⁎⁎⁎ 3.431⁎⁎⁎
Our IV estimates are generally higher than OLS estimates Dummy variable (3.32) (3.37)
High-intensity state 1.029⁎⁎⁎
and suggest that an additional year of schooling reduces Dummy variable (3.69)
the number of children born before age 25 by 0.26. R-squared 0.205 0.219 0.246
Therefore, the OLS estimates provide an underestimate of Observations 2688 2688 2688
the negative effect of schooling on early fertility.
In summary, our results provide robust evidence that B. Program Intensity measured by 1976 per capita federal funds
female education reduces the number of early births. An disbursed for primary school construction
important topic for future research would be to estimate Baseline Add 1970 state Add year
the social savings associated with higher female schooling. FE FE
Within the neoclassical framework, lower fertility impacts (4) (5) (6)
economic growth through several channels, including Born 1970–75⁎ classroom − 0.004⁎⁎ −0.004⁎⁎⁎ −0.004⁎⁎⁎
lower dependency burdens (share of workers to children in Construction funds (2.22) (2.64) (3.63)
the population), which then increases savings and invest- per capita
ment capital. We should note that calculating the social Born 1970–75 0.953 1.102
savings from a reduction in fertility would most likely Dummy variable (1.23) (1.43)
Classroom 0.003⁎⁎ −0.001 −0.001
provide an underestimate of the total benefits associated Construction funds (1.98) (0.29) (0.38)
with increased female schooling because higher female per capita
educational attainment likely would also affect wages, R-squared 0.197 0.213 0.240
child nutrition, child mortality, and other outcomes. Observations 2688 2688 2688
Based on macroeconomic evidence, it is likely the Treatment: born 1970–1975 (Age 1–6 in 1976).
UPE policy and other programs in the developing world Control: born 1956–1961 (Age 15–20 in 1976).
that introduce higher levels of female schooling would Dependent variable: number of kids before age 30.
⁎Significant at 0.10 level. ⁎⁎Significant at 0.05 level. ⁎⁎⁎ Significant
have a significant impact on economic growth. Using a at 0.01 level.
Solow growth framework, Knowles et al. (2002) Notes: This table is similar to Table 3 in the main text except that we
estimate that a 1% increase in female education would look at fertility outcomes up to age 30 rather than age 25. T-statistics
increase average GDP levels by 0.37%. Between 1974 are shown in parentheses. The standard errors are clustered at the
and 1979, the total cost of the UPE program in Nigeria year ⁎ state level. The baseline models include dummy variables for
religion (the six religion categories are Muslim, Catholic, Protestant,
(capital and recurrent expenditure) was about 2.8 billion Other Christian, Traditional Religion with qOtherq being the left out
naira (in 1995 naira) — on average, about 3% of annual group), ethnicity (Hausa, Yoruba, and Igbo with “Other” being the left
real GDP during this period. In comparison, we estimate out group), the female share of total primary school enrollment in 1970
that the average increase in female schooling associated in the state, this variable interacted with a dummy variable for being
with the program was about 1.54 years, or female born 1970–75, and the proportion of civil servants in the state who
were female the year the individual was age 6. State fixed effects refer
schooling attainment was about 30% higher than the to the 12 states that existed in Nigeria in 1970 while classroom
sample mean of 5.0 years. Moreover, UPE may also construction funds are measured at the 1976 state level. The program
have positively affected the outcomes of the children of intensity measure is available for all the 1976 states (19 states) that
the affected women. For example, health and schooling existed at the onset of the UPE program.
indicators for children are known to improve with the
level of female education (Schultz, 1998). Therefore, the
returns to universal primary education programs over
the long term may be substantial.
Appendix Table 2 Appendix

(continued)
Table 3 (continued )
The impact of education on fertility — instrumental variable estimates Dependent var. Years of Number of kids before
A. OLS estimates schooling age 30
Baseline Add 1970 state Add year of birth (1) (2)
Fixed effects Fixed effects Born 1966 or 0.0058⁎ 0.00163

1967 ⁎ Classroom
(1) (2) (3) Construction funds (1.73) (0.91)
Years of education − 0.109⁎⁎⁎ −0.112⁎⁎⁎ −0.108⁎⁎⁎ per capita
(7.92) (10.59) (10.28) Born 1968 or 0.0042 − 0.0035⁎
R-squared 0.243 0.260 0.283 1969 ⁎ Classroom
Observations 2688 2688 2688 Construction funds (1.14) (− 1.68)
per capita
Born 1970 or 0.0129⁎⁎⁎ − 0.0036⁎⁎
B. Instrumental variables estimates instruments used: year of birth
1971 ⁎ Classroom
dummies ⁎ state capital allocation for classroom construction
Construction funds (2.70) (− 2.16)
Baseline Add 1970 state Add year of birth per capita
Born 1972 or 0.0068⁎ − 0.0035⁎⁎
1973 ⁎ Classroom
(1) (2) (3) Construction funds (1.94) (− 2.34)
Years of education − 0.297⁎ −0.400⁎⁎⁎ −0.359⁎⁎⁎ per capita
(1.67) (2.75) (2.69) Born 1974 or 0.0038 − 0.004⁎⁎⁎
Overidentification 0.0014 0.0022 0.408 1975 ⁎ Classroom
test (P-values) Construction funds (1.17) (− 2.67)
R-squared 0.666 0.6081 0.646 per capita
Observations 2688 2688 2688 R-squared 0.3696 0.1713
Observations 5168 5168
Treatment: born 1970–1975 (age 1–6 in 1976).
Control: born 1956–1961 (age 15–20 in 1976). Control: born 1949–1961 (age 15–26 in 1976).
⁎ Significant at 0.1 level. ⁎⁎ Significant at 0.05 level. ⁎⁎⁎ Significant
Dependent variable: number of kids before age 30.
⁎ Significant at 0.10 level. ⁎⁎ Significant at 0.05 level. ⁎⁎⁎ Sig- at 0.01 level.
nificant at 0.01 level. Notes: T-statistics are shown in parentheses. Notes: t-statistics are shown in parentheses. The t-statistics reported are
The t-statistics reported are based on standard errors that are based on standard errors that are clustered at the year ⁎ state level. All
clustered at the year ⁎ state level. The baseline models include models include dummy variables for the two-year groups of the inter-
dummy variables for religion (the six religion categories are Muslim, actions (1962–63, 1964–65, 1966–67, 1968–69, 1970–71, 1972–73,
Catholic, Protestant, Other Christian, Traditional Religion with and 1974–75), religion (Muslim, Catholic, Protestant, Other Christian,
“Other” being the left out group), ethnicity (Hausa, Yoruba, and Igbo Traditional Religion with “Other” being the left out group), ethnicity
with “Other” being the left out group), the female share of total (Hausa, Yoruba, and Igbo with “Other” being the left out group), the
primary school enrollment in 1970 in the state, this variable female share of total primary school enrollment in 1970 in the state, this
interacted with a dummy variable for being born 1970–75, and the variable is also interacted with dummy variables for being born 1962–65,
proportion of civil servants in the state who were female the year the 1966–69, and 1970–75, the proportion of civil servants in the state who
individual was age 6. State fixed effects refer to the 12 states that were female the year the individual was age 6, and state and year fixed
existed in Nigeria in 1970 while classroom construction funds are effects. State fixed effects refer to the 12 states that existed in Nigeria in
measured at the 1976 state level. The program intensity measure is 1970. The program intensity measure is available for all the 1976 states
available at the 1976 state level (19 states) that existed at the onset of (19 states) that existed at the onset of the UPE program.
the UPE program.
References
Appendix Table 3
The impact of UPE by year of birth (2003 sample)
Ainsworth, M., Beegle, K., Nyamete, A., 1996. The impact of
Dependent var. Years of Number of kids before women's schooling on fertility and contraceptive use: a study of
schooling age 30 fourteen sub-Saharan African countries. World Bank Economic
(1) (2) Review 10 (1), 85–122.
Awokoya, S.O., 1952. Proposals for an educational policy. Western
Born 1962 or 0.0061⁎ − 0.001 Regional Gazette 13 (10).
1963 ⁎ Classroom Angrist, J.D., Evans, W.N., 1999. Schooling and labor market
Construction funds (1.82) (− 0.53) consequences of the 1970 state abortion reforms. Research in
per capita Labor Economics 18, 75–113.
Born 1964 or 0.0105⁎⁎ − 0.0053⁎⁎⁎ Angrist, J.D., Bettinger, E., Bloom, E., King, E., Kremer, M., 2002.
1965 ⁎ Classroom Vouchers for private schooling in Colombia: evidence from a
Construction funds (2.09) (− 3.99) randomized natural experiment. American Economic Review 92 (5),
per capita 1535–1558.
Becker, G.S., 1981. Treatise on the Family. Harvard University Press, National Population Commission, 1998. 1991 Population Census of
Cambridge. the Federal Republic of Nigeria Analytical Report at the National
Bledsoe, Caroline H., Casterline, John B., Johnson-Kuhn, Jennifer A., Level. National Population Commission, Abuja, Nigeria.
Haaga, John G., 1999. Critical Perspectives on Schooling and Fertility Nwachukwu, A.E., 1985. An historical analysis of the roots of
in the Developing World. National Academy Press, Washington, D.C. universal public primary education in Nigeria (1900–1980).
Bray, M., 1981. Universal Primary Education in Nigeria: A Study of University of Kansas, Ph. D. dissertation.
Kano State. Routledge and Kegan Paul Ltd., London. Pitt, M., Rosenzwig, M.R., Gibbons, D.M., 1993. The determinants
Breierova, Lucia, Duflo, Esther, 2004. The impact of education on fertility and consequences of the placement of government programs in
and child mortality: do fathers really matter less than mothers? Indonesia. World Bank Economic Review 7 (3), 319–348.
National Bureau of Economic Research Working Paper No. 10513. Rosenzweig, M.R., Schultz, T.P., 1985. The demand and supply of
Caldwell, John C., 1968. The control of family size in tropical Africa. births and its life-cycle consequences. American Economic
Demography 5 (2), 598–619. Review 75 (5), 992–1015.
Caldwell, John C., 1977. The economic rationality of high fertility: an Rosenzweig, M.R., Schultz, T.P., 1989. Schooling, information, and
investigation illustrated with Nigerian survey data. Population nonmarket productivity: contraceptive use and its effectiveness.
Studies 31 (1), 5–27. International Economic Review 30 (2), 457–477.
Caldwell, John C., Orubuloye, I.O., Caldwell, Pat, 1992. Fertility decline in Schultz, T.P., 1981. Economics of Population Reading. Addison-
Africa: a new type of transition? Population and Development Review Wesley, MA.
18 (2), 211–242. Schultz, T.P., 1994a. Human capital, family planning, and their effects
Case, A., Deaton, A., 1999. Schooling inputs and educational outcomes on population growth. American Economic Review 84 (2),
in South Africa. Quarterly Journal of Economics 114 (3),1047–1084. 255–260.
Cochrane, S.H., 1979. Fertility and Education: What Do We Really Schultz, T.P., 1994b. Studying the impact of household economic and
Know? Johns Hopkins Press, Baltimore. community variables on child mortality. Population and Develop-
Duflo, E., 2001. Schooling and labor market consequences of school ment Review 10 (0), 215–235.
construction in Indonesia: evidence from an unusual policy Schultz, T.P. (Ed.), 1998. Demand for children in low income
experiment. American Economic Review 91 (4), 795–813. countries. Handbook of population and family economics 1A.
Fafunwa, Babs A., 1974. History of Education in Nigeria. Allen and Mark Rosenzweig and Oded Stark, eds. The Netherlands Elsevier
Unwin, London. Press, Amsterdam.
Federal Government of Nigeria, 1978–1979. Implementation Committee Schultz, Paul T.P. (Ed.), 2002. Why governments should invest more
for the National Policy on Education blueprint. to educate girls. World Development 30, 207–225.
Federal Ministry of Economic Development and Reconstruction, Thomas, D., 1999. Fertility, education and resources in South
Central Planning Office, 1975. Third National Development Plan, Africa. In: Bledsoe, C.H., Casterline, J.B., Johnson- Kuhn, J.A.,
1975–1980. Lagos, Nigeria. Haaga, J.G. (Eds.), Critical Perspectives on Schooling and
Federal Office of Statistics, various years. Annual Abstract of Statistics. Fertility in the Developing World. National Academy Press,
Federal Office of Statistics, various years. Social Statistics in Nigeria. Washington D.C., pp. 138–176.
Federal Office of Statistics, 1979. Social Statistics in Nigeria. Federal Thomas, D., Strauss, J., Henriques, M., 1991. How does mother's
Office of Statistics, Lagos, Nigeria. education affect child height? Journal of Human Resources 26 (2),
Francis, Paul A., 1998. Hard lessons: primary schools, community and 183–211.
social capital in Nigeria. World Bank Technical Paper No. 420, UNICEF, 1999. Multiple Indicator Cluster Survey.
Africa Region Series. World Bank, Washington D.C. UNICEF, 2005. Nigeria: country report. http://www.unicef.org/
Knowles, Stephen, Lorgelly, P., Owen, P., 2002. Are educational gender infobycountry/nigeria_statistics.html retrieved February 19, 2007.
gaps a brake on economic development? Some cross-country United Nations, 1995. Recent Evidence from the Demographic and
empirical evidence. Oxford Economic Papers 54 (1), 118–149. Health Surveys. United Nations, New York.
Lagos Executive Development Board, 1971. Annual report and Van de Walle, Etienne, 1965. An approach to the study of fertility in
accounts. Lagos Executive Development Board, Lagos, Nigeria. Nigeria. Population Studies 19 (1), 5–16.
Lam, D., Duryea, S., 1999. Effects of schooling on fertility, labor World Bank, 2002. African Development Indicators, Database.
supply and investments in children, with evidence from Brazil. World Bank, 2005. African Development Indicators. World Bank,
Journal of Human Resources 34 (1), 160–192. Washington D.C.
Mason, K.O., 1986. The status of women: conceptual and method-
ological debates in demographic studies. Sociological Forum 1,
284–300.
Journal of African Economies Advance Access published December 9, 2016
Journal of African Economies, 2016, 1–18

doi: 10.1093/jae/ejw027
Article
An African Growth Miracle?†

Dani Rodrik*
John F. Kennedy School of Government, Harvard University, Cambridge, MA 02138, USA
*Corresponding author: Dani Rodrik. E-mail: Dani_Rodrik@hks.harvard.edu
Downloaded from http://jae.oxfordjournals.org/ at Harvard Library on December 21, 2016

†
This is the revised text of the Richard H. Sabot Lecture, delivered at the Center for Global
Development, Washington, DC, on 24 April 2014. I am grateful to Nancy Birdsall for her invitation
and to two reviewers of this journal for their comments. A slightly edited version of the paper
was previously published as ‘Why An African Growth Miracle is Unlikely’, The Milken Institute
Review, Fourth Quarter 2014, 42–54.
Abstract
SSA has grown rapidly over the last decade, but a curious feature of this growth
was that it was accompanied by little structural change towards non-traditional
tradables (such as manufactures). Now that China, the advanced economies, and
most emerging markets are all slowing down, the question whether Africa’s high
growth can be sustained looms larger. This article looks at this question from the
lens of modern growth theory, paying particular attention to structural issues that
are crucial for low-income countries. It comes down on the pessimistic side, due to
what appear to be poor prospects for industrialization. This article also considers
alternative models of growth, based on services instead of manufactures.
Key words: Africa, economic growth, economic reform, industrialization

JEL classification: O40, O14, O55
There is much to celebrate in Africa’s recent economic performance. Gone are the traditional
pessimism about the continent’s growth prospects and the references to basket-case econ-
omies. They have been replaced by rosy scenarios replete with stories of African entrepreneur-
ship, expanding Chinese investments and a growing middle class. The turnaround is easy to
see in the numbers. Having spent a long time in negative territory during the 1980s and
1990s, SSA’s growth rate jumped up to close to 3% per annum in per-capita terms after
2000. This was not as stellar as East and South Asia’s performance but decidedly better than
what Latin America, undergoing its own renaissance of sorts, was able to achieve (Figure 1).
And it is not just a revival in investment. The region has been experiencing positive total fac-
tor productivity growth for the first time since the early 1970s (Figure 2).
The slowing down of emerging market growth and China’s rebalancing troubles have led
many to take another look at Africa’s future economic prospects. Concerns about inadequate
© The Author 2016. Published by Oxford University Press on behalf of the Centre for the Study of African Economies.
All rights reserved. For permissions, please email: journals. permissions@oup.com
2 Dani Rodrik
10.0%
1980–1990
1990–2000
8.0%
2000–2012
6.0%
4.0%
2.0%

0.0%
World Low income Middle income East Asia & South Asia Sub-Saharan Latin America &
Pacific Africa (developing Caribbean
(developing only) only) (developing only)
–2.0%
annual average per-capita GDP growth
Figure 1: Growth performance of country groups since 1980.

Source: World Development Indicators, World Bank.
Figure 2: Growth rate of TFP by subregion, 1960–2010.

Source: Calculations based on PWT8.0 (database) and Feenstra, Inklaar and Timmer (2013); UNECA (2014).
structural change have been raised, among others, by the UN Economic Commission for
Africa (UNECA, 2014) and the African Center for Economic Transformation (ACET, 2014).
As welcome as recent growth has been, the depth of the economic decline prior to the last
African Growth Miracle? 3

Figure 3: Economic performance in Sub-Saharan Africa, 1960–2012 (GDP per capita, constant 2005 $).
Source: World Development Indicators, World Bank.
decade means that many African countries still have not caught up with post-independence
income levels. If the World Banks’s figures are to be believed, the Central African Republic,
the Democratic Republic of Congo, Niger, Liberia, Cote d’Ivoire, Liberia, Zambia,
Zimbabwe and Senegal are all now poorer than they were in 1960 (Figure 3).
It is clear that Africa has benefited from a particularly favourable external environment
during the last two decades. Global commodity prices have been high and interest rates low.
Private capital flows have supplemented increased official assistance. China’s rapid growth
has fuelled demand for the region’s natural resources and has stimulated direct investment in
African economies. The global financial crisis, meanwhile, had little direct impact, given
African countries’ weak financial links with the ROW and low levels of financialisation.
Now that China, the advanced economies, and most emerging markets are all slowing
down, there is a genuine question about whether Africa’s growth can be sustained and, if so,
at what level. I will look at this question from the lens of modern growth theory, paying par-
ticular attention to structural issues that are crucial for low-income countries. I come down on
the pessimistic side, due to what I think are poor prospects for industrialisation. Even if my dis-
cussion does not yield decisive answers, I hope it clarifies the issues.
1. The economics of convergence

Neoclassical growth theory establishes a presumption that poor countries should grow fas-
ter than rich countries. After all, they have the advantages of economic backwardness: they
have low capital–labour ratios, which should raise the return to investment, everything else
being the same. Furthermore, they can rely on global capital markets to supplement domes-
tic saving, so the latter should not act as a constraint. Finally, they have access to global
markets so that they can expand output quicker in those tradable goods in which they have
comparative advantage.
The reality is that convergence has been the exception rather than the norm since the
great divergence spawned by the Industrial Revolution and the division of the world into a
4 Dani Rodrik
.03 1870–2008 1965–2005, decadal
HKG
.1
SGP
TWN
orthogonal component of growth

KOR JPN
FIN NOR
.05
VEN GRC SWE
.02
ESP IRL
THAMYS PRT CAN
ITA AUT
CHN
YUGMEX CZE FRA
DEU CHE USA
DNK
BGR POL CHL
ALB TUN BRA SYR
IRN TUR
HUN ARG NLD
BEL
JOR
0
IDN AUS
GBR
JAM
MMR
VNM MAR EGY LKA NZL
IND ZAF
LBNROM
PHL DZA URY
.01
GHA
–.05
NPL WBG
PRK
IRQ
–.1
0
6 6.5 7 7.5 8 4 6 8 10 12
log GDP per worker, 1870 log GDP per worker, at start of decade

Figure 4: Convergence is historically the exception rather than the norm.
Notes: For RHS chart, variable on the vertical axis is growth of GDP per worker over four separate
decades (1965–1975, 1975–1985, 1985–1995,1995–2005), controlling for decadal fixed effects.
Source: Rodrik (2013a, b), using data from Maddison (2010) and Heston et al. (2011).
rich core and a poor periphery (Figure 4). Except for the European periphery and East
Asia, sustained rapid growth in the lagging regions has been rare.
Growth theory has accommodated this empirical reality by distinguishing between uncon-
ditional and conditional convergence. So, growth in developing nations is held back by a var-
iety of country-specific obstacles, ranging from weak institutions to poor geography and from
lousy policies to poverty traps. Accordingly, developing nations converge to rich country
income levels only conditional on these disadvantages being overcome. Conditional conver-
gence can be expressed formally as follows:
yˆj = β (ln y ⁎ (Θj ) − ln yj ) + εj ,
where ŷj is the growth rate of per-capita (or per-worker) GDP, yj , in country j, Θj is a vector
of country-specific circumstances determining the long-run income level, β is the rate of
(conditional) convergence and εj is a random shock term.
What goes into Θj is what we might call the ‘growth fundamentals’—the set of factors
that condition long-run income levels. While this set could be quite large in principle, many
of the plausible members of the set are also endogenous in the long run. Typical condition-
ing variables used in growth regressions such as levels of investment, human capital and the
quality of policies might be all viewed as being ultimately determined, for example, by a
country’s quality of institutions (as has been argued forcefully by Daron Acemoglu, James
Robinson and assorted co-authors). Or they may be determined by geography and ecology
(as has been argued by Jeff Sachs and co-authors). Institutions themselves may be endogen-
ous to initial levels of human capital brought in by colonisers (as has been argued by
Glaeser and Shleifer).
For the purposes of the present discussion, I do not need to take a strong stand among
these contending perspectives on what the true growth fundamentals are. As long as we leave
room for human capital and institutions, I am happy to accept that geography matters too.
African countries cannot do much about their geography, but there is little doubt
that their growth fundamentals on all other dimensions have improved significantly.
Agricultural markets have been liberalised, domestic markets have been opened up to inter-
national trade, parastatals have been rationalised or closed down, macroeconomic stability
has been restored and exchange rate management is infinitely better than it used to be
(Figure 5). Beyond economic governance, political institutions have improved significantly
as well, with democracy and electoral competition becoming the norm rather than the
exception throughout the continent (Figure 6). Finally, some of the worst military conflicts
have ended, reducing the number of civil war casualties in recent years to historic lows for
the region (Figure 7).

Figure 5: Trends in Africa’s foreign currency black market premiums and index policy reform, 1960–2010.
Source: Giuliano, Mishra and Spilimbergo (2013); UNECA (2014).
Figure 6: Trends towards democracy and electroral competition, 1960–2010.

Source: www.systemicpeace.org/polity/polity4.htm; UNECA (2014).
6 Dani Rodrik
Figure 7: Africa’s fundamentals: fewer civil wars.

Source: Straus (2012).

That is all good news for Africa’s economic prospects, but how much growth should we
expect out of them? The improvement in the policy and institutional environment can be
expected to generate greater economic stability and prevent deep crises arising out of mismanage-
ment as in the past. But it is not clear that it provides a significant boost for economic growth
and nor that it acts, on its own, as the engine for a growth miracle. Work by Bill Easterly, myself
and others has shown that the relationship between standard measures of good policy (such as
trade liberalisation and low inflation) and economic growth is not particularly strong, leaving
extreme cases aside. A huge black market premium for foreign currency and hyperinflation can
drive an economy to ruin, but there is no predictable or large growth difference between an infla-
tion rate of 5% and 15% or an average tariff rate of 10% versus 25%. Economists have a pretty
good idea of what can cause economic collapse, but not so much about what can produce a mir-
acle. The upside potential of these policy reforms remains uncertain as a result.
What about institutions that have received so much attention in the literature? Is not
it the case that high-quality institutions make a huge difference to long-run income
levels and hence convergence patterns? Acemoglu et al. (2014) claim that differences in
institutional quality account for as much as 75% of the variation in income levels
around the world. This is a very big number. And it may well be right for the very long
run. The trouble is that even if it is correct, this long-run relationship tells us rather less
about growth prospects over the next decade or two. The empirical relationship
between institutions (or the change in the quality thereof) and growth rates tend not to
be that strong, unlike what the long-run relationship in levels suggests. Few would deny
that Latin America’s political and economic institutions have improved significantly
over the late 1980s and 1990s. Yet, the growth payoff has been meagre at best.
Conversely, high-performing Asian economies such as South Korea (until the late
1990 s) and China (presently) have been rife with institutional shortcomings such as
cronyism and corruption and yet have done exceedingly well.
Consider democracy. Despite an extensive empirical literature, the growth effects of dem-
ocracy still remain in question. The strongest recent statement about the growth-promoting
effects of democracy comes from Acemoglu et al. (2014), who find that full democratisation
produces roughly a 20% increase in GDP per capita over 30 years. This translates to a
growth effect of about 0.6% per year. This is not an insignificant effect, but it is temporary
and phased out over time. And it cannot account for a substantial part of income differences
across the world—nothing like the 75% claimed for ‘institutions’ in general.
To get large effects out of institutions, even for the long run, we need to use measures
such as the ‘rule of law’ or ‘expropriation risk’. An important problem is that these are out-
comes: they tell us something about investors’ evaluation of the economic environment but
not so much about how to get there. It remains unclear which policy levers have to be
pulled to get those outcomes. Surely what is required is more than passing the relevant
laws or regulations. And perhaps those same outcomes can be obtained through institu-
tional forms that look very different than those we associate with the ‘rule of law’ in west-
ern contexts. As I have argued elsewhere, the function that good institutions fulfil (about
which we have a fairly good idea) do not map into unique forms (about which we know a
lot less) (Rodrik, 2008a, b). The mapping depends on local context and opportunities, and
figuring it out can be quite hard. One lesson for Africa is that we should not be overly con-
fident about the growth payoffs when countries adopt the formal trappings of ‘good

institutions’.
2. A structural transformation perspective

So, the standard growth equation yˆj = β (ln y⁎ (Θj ) − ln yj ) + εj does not do a very good job
of describing growth miracles, at least with the usual fundamentals, Θ. A complementary
perspective is provided by the tradition of dual-economy models that have long been the
staple of development economics. The birth of modern growth economics has oversha-
dowed this tradition aside, but it is clear that the heterogeneity in productive structures that
dual-economy models capture continues to have great relevance to low-income economies
such as those in SSA. A hallmark of developing countries is the wide dispersion in product-
ivity across economic activities—modern versus traditional, formal versus informal, traded
versus non-traded, cash crops versus subsistence crops, etc.—and even within individual
sectors, as recent studies have documented.
What was implicit in those old dual-economy models was the difference in the dynamic
properties of productivity across the modern–traditional divide. Traditional sectors were stag-
nant, while modern sectors had returns to scale, generated technological spillovers and experi-
enced rapid productivity growth. This picture has been refined over time, and we no longer
think of traditional sectors—such as agriculture—as necessarily stagnant. But in one import-
ant respect, recent findings reinforce the dual-economy perspective. As I have shown (Rodrik,
2013a, b), modern, organised manufacturing industries are different: they do exhibit uncondi-
tional convergence, unlike the rest of the economy (Figure 8). The estimated beta-coefficient
in these industries is close to 3%, suggesting a half-life of convergence of 40–50 years.
This is a rather remarkable result. It says that modern manufacturing industries con-
verge to the global productivity frontier regardless of geographical disadvantages, lousy
institutions or bad policies. Under better conditions, convergence could be faster of course.
But what is striking is the presence of convergence, in at least certain parts of the economy,
even in the absence of good fundamentals.
In Rodrik (2013a, b), I show that this result is fairly general, regardless of time period,
region or level of aggregation. In particular, the twenty or so African countries that are
represented in the UNIDO data set follow the same pattern as the ROW (Figure 9). In this
respect, Africa is no different. So can Africa generate a growth miracle based on the per-
formance of these manufacturing industries?
8 Dani Rodrik

Figure 8: There is unconditional convergence—in (formal) manufacturing industries.
Notes: Vertical axis represents growth in labor productivity over subsequent decade (controlling for
period fixed effects). Data are for the latest 10-year period available.
Source: Rodrik (2013a, b).
Full sample: 115 countries Sub-Saharan Africa: 20 countries
Figure 9: African manufacturing seems no different.

Notes: Each observation represents a 2-digit manufacturing industry, for the latest 10 year period for
which data are available. The horizontal axis is the log of VA per worker in base period, and the verti-
cal axis is its growth rate over the subsequent decade, Period, industry, and period x industry con-
trols are included.
Let us first integrate this sectoral convergence result with the conditional convergence
framework for the entire economy. Divide the economy into two parts, the modern (or manu-
facturing) part, with the subscript M, and the rest (or traditional part) with subscript T.
Suppose only the M-sector exhibits unconditional convergence and the T-sector is subject to
conditional convergence as before. Now the growth rate of the economy can be decomposed
into three channels:
y = β (ln y ⁎ (Θ ) − ln y )
+ αM πM βM ( ln y ⁎ − ln yM )
+ (πM − π T ) dαM
The first of these is the conditional convergence channel we have looked at before. It
depends on the cumulative accumulation of fundamental capabilities, vague as the contents
of these may be, as I discussed before. The second channel is convergence within modern
industries. Its magnitude depends on the distance from the productivity frontier, the conver-
gence coefficient ( βM ), the productivity premium in M relative to the economy (πM ) and the
employment share of M (αM ). The third channel is the structural change term and captures
the growth effect of the reallocation of labour from low-productivity sectors (T) to high-

productivity sectors (M).
The two new terms can boost growth significantly and indeed have played a key role in
Asian growth miracles. Their quantitative magnitudes depend crucially on the size of the
modern/manufacturing sector and its rate of expansion (αM, dαM )—that is, the pace of
industrialisation. Rapid industrialisation produces fast growth into middle-to-upper income
status. In the later stages of growth, as industrial convergence runs out of steam, economic
progress begins to rely disproportionately on the fundamentals and growth slows down.
This framework produces the following typology of growth patterns.
A typology of growth processes/outcomes
As the 2 × 2 box makes clear, long-term convergence requires both structural change
and fundamentals. Rapid industrialisation without the accumulation of fundamental cap-
abilities (institutions, human capital) produces spurts of growth that eventually run out of
steam. But investment in fundamentals on its own produces moderate growth at best in the
absence of rapid structural change.
3. Structural change and industrialisation in Africa

So where does Africa stand in structural change? Here, the picture is considerably less
bright. While farmers have moved out of rural areas and the share of agriculture in employ-
ment and value added has dropped significantly since the 1960 s, the primary beneficiary
10 Dani Rodrik

Figure 10: GDP, employment, and relative productivity levels across countries and sectors, 1960–2010.
Source: de Vries et al. (2013).
has been urban services rather than manufactures. In fact, industrialisation has lost ground
since the mid-1970s, and not much of a recovery seems to have taken place in recent dec-
ades. Manufacturing industries’ share of employment stands well below 8%, and their
share in GDP is around 10%, down from almost 15% in 1975 (Figure 10). Most countries
of Africa are too poor to be experiencing de-industrialisation, but that is precisely what
seems to be taking place. Note that the data I am relying on here, from the Groningen
Growth and Development Center, cover only eleven countries in the entire continent. But
data from other sources (such as the World Bank’s World Development Indicators) tell a
broadly similar and not very encouraging story.
Figure 11 provides a visual comparison with Asian countries. African countries are
shown in blue, while Asian countries are red. Not surprisingly, African observations are
mostly on the lower left-hand side of the chart at low levels of income and industrialisation
compared to Asia. But more importantly, and less evidently, the industrialisation-income
relationship looks decidedly different in the two regions: African countries are under-
industrialised at all levels of income, relative to Asia.
Figures 12 and 13 compare patterns of structural change for specific countries. Look first
at Vietnam that exhibits the classic, growth-promoting pattern of structural change. Labour
has moved from agriculture into more productive urban occupations. Manufacturing has
expanded by 8% of the labour force over 1990–2008, but so has many services that are
comparatively of high productivity. The work by McCaig and Pavcnik’s (2013) shows that
these patterns of structural change account for around half of Vietnam’s impressive growth
over this period. The pattern in Africa, exemplified by Ethiopia and Kenya in Figure 13, is
much more mixed. In both cases, there has been outmigration from agriculture, but the con-
sequences have been less salutary. In Ethiopia, where there has been some growth-
promoting structural change, its magnitude is much smaller than in Vietnam. Manufacturing
industry, in particular, has expanded much less. In Kenya, meanwhile, structural change has
contributed little to growth. That is because the large number of workers leaving agriculture
have been absorbed mainly into services where productivity is apparently not much higher
than in traditional agriculture.
Manufacturing employment and GDP per capita Manufacturing value added/GDP and GDP per capita
.5
KOR
.4
KOR
KOR
manufacturing employment share
THA
THA KOR
KOR
HKG
HKG
HKG THA
THA KOR
HKGHKG
HKG
HKG THA
THA TWN
TWN
HKG THA TWN
TWNTWN KOR
KOR
KOR
THA
.4
HKG
HKG TWN
KOR
THA
THA
TWN KOR
TWN
KOR
KOR
KOR
HKG PHL
PHL THA
THA TWN
KOR
TWN
THATWN KOR
KOR
MYS TWN
HKG MYS
MYS TWN
.3
TWNTWN
TWN HKG
HKG PHL
PHL PHL
PHL
PHL
PHL
IDN
IDNIDN THA
TWN
THA KOR
TWN
SGP SGP
MYS
MYS
MYS
MYSSGP
SGP TWNTWN
TWN TWNTWN PHL
IDN
PHL
PHL
IDNPHL THA
TWN TWN
KORSGP
KOR
SGPSGP
MYS
MYS
MYS
SGP SGP
SGP
TWN
TWN
SGP
TWN
TWN TWN SGP
SGP
SGP
TWN
MUSTWN
TWN
MUS
MUS
MUS
MUS TWN HKG PHL
PHL
PHL
PHL PHL THA
IDN
PHL
PHL
PHL
THA
PHL TWN
PHL SGPMYS
MYS SGP
SGP SGP SGP TWN
SGP
SGPTWN
TWN
TWNSGP
MVA/GDP
MUS
TWN MUS
SGP TWN PHL IDN
IDN
THA
THATWN KOR
KOR
SGP
KOR MYS
MYS SGPTWN
SGPJPN SGP
SGP
.3
TWN
TWN MUS
SGP MUS
SGP
SGP
MUS
SGP
MUS
MUS
MUSMUS TWN
SGP
SGP HKG
TWN THA THA
THATHA
THA
THA THA
IDN
THA
THAKOR
KOR
SGPMYS MUS
MUS
MUS
MUS
MUS SGP
JPN JPN
JPN
JPN
JPN JPN
SGP
JPN
JPN SGP
JPN
SGP
JPN
JPN
TWN
TWN SGP
KOR
KOR
MUS
KOR
SGPKOR
MYS
MYS
SGP
MYS SGP
KOR
MYS SGP
SGP TWN
TWN
SGP
TWN
TWN
SGP
TWNTWN
TWN
HKG
TWN
TWN
TWN THA IDN
IDN
IDN
THA TWN ZAF
MUS MUS
MUS
MUS
MUS
MUSMUS
JPN
JPN
MUS JPN
JPNJPN
JPN
JPN
JPN
SGP JPN
JPN
JPN
JPN
JPN
MUS
SGPMYS
JPN
SGP
JPN
JPN MUS
MYS
MYS JPN
JPN
JPN
KOR
JPN
SGP
SGP SGP HKG IDN
IDN
THA TWNSGP
KORMYS
ZAF
ZAF
ZAF JPN
JPNMUS JPN
TWNJPN
JPN
JPN
JPN MUSKOR
MYS JPN
KOR
JPNSGP
KOR IDN SGPSGP
KOR
MUS MYSJPN
ZAF
MUS
ZAF
ZAF MUS
.2
JPN
JPN KOR
KOR
TWN SGP
KOR
SGP
KORMYS
MYSMYS
MYS JPN
KOR
JPN
JPN
JPN
JPN
KOR JPN
JPN
SGPJPN
JPN
JPN
SGP HKG
JPN
SGP
JPN IND
IND
IND IDN
THA SGP
SGPTWN ZAF
ZAF
JPN
ZAF
MYS
ZAF
ZAF
JPN
JPNMYS
MUS
ZAF MUS
ZAF
MYS
ZAF
ZAF MUS
MUS
ZAF
JPN
JPN
JPN TWN MUS
SGP
KOR
MUS
KOR
MUS
KOR
MUS MYSMUS
MUS KOR
KOR SGP
JPN
SGPSGP
HKG
SGP
SGP
SGP
JPN
SGP IND
IND INDTHA
GHA
THA
THA GHA
IND
GHA
INDIND
GHA
IND
INDTHA
GHA
IND
IND
IDN
SGP
GHA INDTWN
SGP
TWN KOR
KOR ZAF
MYS
MYS
JPN
MUS
MUS
MUS
MYS
ZAFMYS
MYS
ZAF
ZAF ZAFMUS
KOR
MUS MUS KOR SGP IND THA
IND
SEN
SEN
SEN
THASEN
GHA
GHA SGPSGP MUSMUS ZAF
.2
NGA
NGA
NGA JPN
JPN TWN
TWN KOR
MUSMYS MYS MUS
MUS KOR KOR
KOR
KOR
KOR HKG
JPN
JPN
HKG THA IND
IND
IND IND
IND
THA
IND
THA
THA IND
IND SEN
SENSEN
SEN
SEN
SEN SEN
SEN
GHAGHA
SEN
IDN
IDN
IDN TWN
TWN KOR MUS
MYS
JPN
MUSZAF
MYSMUS
MYS
JPN
ZAF
MUS
MUS
NGA
NGA
NGA JPN
JPN TWN
JPN KOR MYS JPN THA
THATHA
THA THA
IND
IND
THA
THAIND
IND ZMB
IND SEN
SEN
IND GHA
SEN
ZMB
SEN GHA
SEN
SEN
ZMBGHA
SEN
SENTWN MYS
KOR MUS
JPNZAF
MYS
MYS
NGA
GHA TWN
TWN
TWN KOR MUS ZAF
ZAF
ZAF
MYS
ZAFTHA
MYS
THA
THA
MYS
MYS
THA
MYS
MYS
ZAF
ZAF
MYS
THA HKG ZMB
IND
INDIND
ZMB
ZMBINDSEN
SEN
SEN
ZMB
KEN
ZMB
KEN
KEN
ZMB
GHA
GHA
ZMB IDN
KEN
ZMB
TZA
KEN
KEN GHA
IDN
KEN
ZMB
ZMB
ZMB
ZMB TWN
TWN
GHA
GHA
GHA
ZMB
ZMBKORJPNJPN
JPN
JPN ZAF
GHAGHA
GHA
NGA NGA
GHA
NGA
GHATWN
KOR
KOR IDN
IDN
IDN
KOR IDN
IDN
IDNIDNMYS
MUS THA
ZAF
ZAF
ZAF
ZAF THA
ZAFTHA
ZAF
THA
ZAF
ZAF HKG ZMB
ZMB
MWI IND
IND
IND
IND
ZMB
ZMB
MWI
ZMB KEN
ZMB
ZMB KEN
KENGHA
KEN
ZMB GHA
GHA
GHA
TZA
ZMB KEN
TZA
KEN
KEN
TZA
TZA
KEN BWA
IDN
TZA
GHA
IDN
GHA ZMB
KORJPN
JPN
ZMB ZMB
JPN HKG
KEN
KEN
KEN
KEN GHA NGA
GHA
GHA KOR
IND
IND
IND KOR
IND
IND
IND
GHAIND
GHA IDN
IDNMUS ZAF
ZAF
ZAF THA
ZAF
THAZAF
THAZAF HKG MWI
MWIMWI
MWIIND
MWI
MWI TZATZA
GHA
TZA IDN
GHA
BWA
KEN
GHA
GHA
GHA
ZMB
KEN
BWA
IDN BWA
GHA
GHA
ZMB KOR ZMB HKGHKG HKG
HKG
HKG
HKG
HKG HKG
HKG
.1
NGA
NGANGA
NGA NGA INDKEN
GHA
GHA NGA
IND
KOR
GHA
IND
GHA
GHA
KEN
IND
IND GHA
GHA
GHANGA
GHA
GHA
GHA
GHA
GHA IND
PHL
GHA
GHA
IDN
GHA
PHL IDN
IDN
IDN
PHL
PHL
PHL
PHL IDN
PHL
PHL
PHLPHL
PHLMUS THA
MYS
ZAF
THA
ZAF
PHL ZAF HKG MWI
MWI
MWI
MWI MWI
MWI
MWI
IND
IND
IND
MWI
MWIMWI
MWI
MWI TZA
TZA
TZA
IDN
MWI TZA
IDN
TZA GHA
GHA
TZA
IDN
IDN
TZA KEN KOR
KORGHA
ZMBGHA
BWA ZMB
BWA
GHA
GHA HKGHKG
HKG
HKG HKG
HKGHKG
HKG
HKG
HKG
IND
NGA PHL MWI
MWI TZA TZA HKG
.1
NGA
NGA IND
IND
IND
INDNGA
NGA
IND
GHA
IND
IND
IND IND KOR
KEN
SEN
KOR
INDKEN
SEN
NGA
KEN GHA
IDN
IDN
SEN IDN
IDN
IDN
IDN PHL
PHL
PHLTHA
THA
PHL PHL
PHL
THA
PHL
PHL
THA
PHL
THA
THA PHL
PHL
THA
ZAF
THA
THA HKG
HKG MWI
MWI TZA
TZA
TZATZA
TZA
TZA
TZA IDN
IDN
TZAIDN
TZA
TZA IDN TZA
KEN
KEN
KEN
KOR
KEN
BWAKEN BWA
ZMB BWAHKG
BWA
BWA
BWA
BWA BWA
BWA HKG
HKG
HKG
NGA IDN
IDN KOR
IDN
KOR IDN
IDN
IDNKEN
SEN
SEN
SEN
NGA KEN THA
THA
THA
THA
THA THA THA BWA HKG TZA KOR
KEN
KEN
KOR
KOR KEN
KEN
KOR BWA BWABWA BWA
BWA BWA
BWA
BWABWA
BWA
BWA BWA
BWA
BWABWA HKG
HKG
NGA
NGA ETH
ETH
ETH THA
THA IDN
THA
THA
THA KEN
SEN
NGA
SEN
SEN
SEN
KEN
THA
THA
THA
NGA SEN
SEN
SEN
SEN
SEN
KEN THA
SEN BWA
BWA
BWA BWA
BWA
BWA
BWA
BWA
BWA BWA
BWA
BWA
BWA
BWA
BWA BWA ETH
ETH
ETH
ETHETH
ETHETH
ETHETHBWA
ETH
ETH
ETH
BWA
ETH KOR
BWAKOR
KOR BWA NGA BWA
BWA BWA BWA HKG
HKG
NGA NGA
ETH
NGA
NGA
ETH
ETHMWI THA
NGA
MWI
MWI
MWI
MWI
MWI
MWI
MWI NGA NGA
ZMB
ZMB
ZMB
ZMB KEN
KENKENKEN
NGA
NGA
ZMBKEN
ZMB
ZMB
KEN
KEN
NGA
KEN NGA
ZMB
ZMBNGA
ZMB BWA
BWA NGA NGA
NGA
NGA ETH
NGA BWA
ETH
NGA
ETH
ETH
NGA ETH
ETH
ETH
BWA
NGA
NGA KOR NGANGA
NGANGA
NGA
NGA NGA
NGA HKG
HKG
ETH
ETH
ETH
ETH
ETH MWI
MWI
MWI
ETH
ETH
ETHETH
ETH MWI
MWI
MWI
MWI ZMB
ZMB
MWI
MWI MWI
ZMB
MWI
ZMB
ZMB
ZMB
MWI
TZA ZMB
ZMB
TZA
TZA
TZA TZA
TZA TZAKEN
ZMB
TZAZMB
ZMB ZMB
ZMB BWA
ZMBZMB
BWABWA
BWA NGA NGAETH
NGA
ETH
NGA
ETH
ETH
ETH NGA KOR NGA
NGA
NGA NGA
NGA NGA
NGA
NGA
NGA
ETH
BWA
ETH
BWA
ETH
ETHETH
ETH
TZATZA
ETH
TZA
TZA
ETH
ETH
TZA
BWA
BWA
BWATZA
BWA
BWA BWA
TZABWA
TZA
TZA
TZA
TZA TZA
TZA
TZA TZA
TZA
TZATZA
BWA
BWA BWA
TZABWA
BWAZMB
BWA BWA
BWA
BWA NGANGA
NGA NGA
NGA
NGA
0
0
4 6 8 10 4 6 8 10
log GDP per capita log GDP per capita
Africa Asia Africa Asia

Figure 11: African industrialisation is lagging behind, even controlling for incomes.
Notes: Africa: Botswana, Ethiopia, Ghana, Kenya, Mauritius, Malawi, Nigeria, Senegal, Tanzania,
South Africa, and Zimbabwe. Asia: Hong Kong, Indonesia, India, Japan, Korea, Malaysia, the
Philippines, Singapore, Thailand, Taiwan, and Vietnam.
Source: Based on data from Groningen Growth and Development Center.
1990–2008
Figure 12: Structural change in Vietnam.
The even worse news for African manufacturing is the degree to which it is dominated by
small, informal firms that are not particularly productive. The share of formal employment in
overall manufacturing employment appears to run as small as 6% in Ethiopia and Senegal
(Figure 14). Remember that the finding on unconditional convergence applies to formal, orga-
nised firms. There is little reason or evidence to believe that informal firms are on the same escal-
ator as modern firms with access to technology, markets and finance. The evidence on
12 Dani Rodrik
Correlation Between Sectoral Productivity and Correlation Between Sectoral Productivity and
Change in Employment Shares in Ethiopia (1990-2005) Change in Employment Shares in Kenya (1990-2005)
Log of Sectoral Productivity/Total Productivity

Log of Sectoral Productivity/Total Productivity
β = 9.4098; t-stat = 0.91 β = 0.0902; t-stat = 0.02 pu
3
4
fire
fire min
pu
3
2
tsc tsc
2
1
con
con
cspsgs
1
cspsgs wrt
min man
0
man
0
agr
agr wrt
–1
–1
–.3 –.2 –.1 0 .1 .2

–.1 –.05 0 .05
Change in Employment Share Change in Employment Share
(ΔEmp. Share) (ΔEmp. Share)
Fitted values Fitted values
*Note: Size of circle represents employment share in 1990 *Note: Size of circle represents employment share in 1990

**Note:β denotes coeff. of independent variable in regression equation: **Note:β denotes coeff. of independent variable in regression equation:
ln(p/P) = α + βΔEmp. Share ln(p/P) = α + βΔEmp. Share
Source: Authors' calculations with data from National Bank of Ethiopia and Source: Authors' calculations with data from Kenya National Bureau of Statistics,
Ethiopia's Ministry of Finance Central Bureau of Statistics, UN National Accounts Statistics and ILO's KILM
Figure 13: Structural change in Ethiopia and Kenya.

Source: McMillan and Rodrik (2011).
year UNIDO GGDC ratio (%)
BWA 2008 3.6 6.4 56

ETH 2008 0.3 5.3 6
GHA 2003 1.0 11.21 9
KEN 2007 1.5 12.9 12
MUS 2008 16.3 21.5 76
MWI 2008 0.7 4.3 16
NGA 1996 1.4 6.6 21
SEN 2002 0.5 8.9 6
TZA 2007 0.5 2.3 22
ZAF 2008 7.0 13.1 53
ZMB 1994 1.5 2.9 52
Notes: Difference in coverage between two data sets: GGDC (which covers
informal employment) and UNIDO (which is mostly formal, registered firms)
Figure 14: Informality dominates in African manufacturing. Manufacturing employment shares,

GGDC and UNIDO datasets, 1990.
informality suggests few small, informal firms eventually grow out of informality. So informality
is a drag on overall productivity, and this plays a large part in explaining why not just services
but also manufacturing in Africa has been falling behind the productivity frontier, even in recent
years with high growth (Figure 15).
To sum up, the African pattern of structural change is very different from the classic
pattern that has produced high growth in Asia, and before that, the European industriali-
zers. Labour is moving out of agriculture and rural areas. But formal manufacturing

Figure 15: An international perspective on productivity (USA = 100): manufacturing.
Source: de Vries et al. (2013).
industries are not the main beneficiary. Urban migrants are being absorbed largely into ser-
vices that are not particularly productive and into informal activities. The pace of industri-
alisation is much too slow for the convergence dynamics to play out in full force.
4. High-growth scenarios for Africa

Countries that have grown rapidly recently in Africa have relied on engines that appear to
be running out of fuel, such as commodity booms or foreign transfers. There are also exam-
ples of public investment–driven growth, among which Ethiopia stands out. Ethiopia has
experienced GDP growth of more than 10% per annum over last decade, due in large part
to the increase in public investment, from 5% to 19% of GDP. The resources spent on
investment—in basic infrastructure such as roads and hydroelectricity—appear to have
been well spent; they have raised the overall productivity of the economy and reduced rural
poverty. So far, the Ethiopian government has managed to avoid major fiscal imbalances
by reducing public consumption and relying on private savings channelled through finan-
cial repression (World Bank, 2016). But public investment booms often do not end well,
and there are questions about the sustainability of Ethiopia’s strategy into the future.
To generate sustained, rapid growth into the future, Africa has essentially four options.
The first one is to revive manufacturing and put industrialisation back on track, so as to repli-
cate as much as possible the traditional route to convergence. The second is to generate
agriculture-led growth, based on diversification into non-traditional agricultural products.
The third is to generate rapid growth in productivity in services, where most of the people
will end up in any case. The fourth is growth based on natural resources, in which many
African countries are amply endowed. Let me say a few words about each of these scenarios.
What are the prospects for a renewed industrialisation drive in Africa? While the bulk
of Chinese investment has gone to natural resources, there have been some hopeful signs of
14 Dani Rodrik
greenfield investments in manufacturing as well in many countries of the region, most not-
ably Ethiopia, Nigeria, Ghana and Tanzania. Looking at some of these green shoots, one
can perhaps convince oneself that Africa is well poised to take advantage of rising costs in
Asia and turn itself into the world’s next manufacturing hub. A recent McKinsey Global
Institute report reflects such optimism (Bughin et al., 2016). Yet, as we have seen, the aggre-
gate data do not yet show something like this happening.
There is almost universal consensus on what holds manufacturing back in Africa. It is
called ‘poor business climate’, a term that is sufficiently broad and all-encompassing that
there is room for virtually anything under its rubric. The very useful paper by Gelb et al.
(2014), for example, cites costs of power, transport, corruption, regulations, security, con-
tract enforcement and policy uncertainty, among other impediments. There is little doubt
that all of these raise the costs of doing business in Africa for an investor interested in start-

ing or expanding a manufacturing operation.
But there is also a hopeful side to this account. If the problem is that such costs act as a
tax on tradable industries, there is a relatively easy remedy that could compensate for them.
It is the exchange rate. A real exchange rate depreciation of, say, 20%, is effectively a 20%
subsidy on all tradable industries. It is a way of undoing the costs imposed by the business
environment in a relatively quick and easy manner. Where the culprit for slow industrialisa-
tion are market failures, an undervalued exchange rate also substitutes for industrial policy.
At the right exchange rate, many African manufacturers can compete with Chinese and
Vietnamese exporters, both externally and in the home market. As I and others have noted,
an undervalued real exchange rate may be the most effective tool for spurring industrialisa-
tion and hence growth (Rodrik, 2008a, b; Johnson et al., 2010).
Of course, achieving and sustaining a competitive/undervalued real exchange rate
requires an appropriate monetary/fiscal policy framework. In particular, it requires man-
aging or discouraging capital and aid inflows and a tighter fiscal policy than otherwise. But
these macroeconomic policy adjustments may be considerably easier to implement than the
endless series of policy reforms needed to fix the individual problems associated with the
‘poor business climate’. Once the economy is on a higher growth path, it may become eas-
ier to deal with those problems over time, reducing the reliance on the real exchange rate.
Yet, I have the suspicion that the obstacles industrialisation faces in Africa are more
deep-seated and go beyond specific African circumstances. For various reasons that we do
not quite understand, industrialisation has become really hard for all countries of the
world. The advanced countries are of course de-industrialising, which is not a big surprise
and can be ascribed both to shift in demand in services and imports. But middle-income
countries in Latin America are too. And industrialisation in low-income countries is run-
ning out of steam considerably earlier than has been traditionally case. This is the phenom-
enon that I have called ‘premature industrialisation’.
As Figure 16 shows, late developers have begun to de-industrialise at lower and lower levels
of income. The first wave of industrializers such as Britain and Germany put more than 30% of
their labour force in manufacturing before they began to de-industrialise. Among Asian expor-
ters, the most successful such as Korea reached a peak well below 30%. Today, countries such
as India, along with many Latin American countries, are de-industrialising from peaks that do
not exceed the mid-teens. Even Vietnam, which is one of the most successful recent industriali-
zers, shows signs of having peaked at 14% of employment. Yet, Vietnam is still a poor country,
and in an earlier period would have had many more years of further industrialisation.

Figure 16: Peak manufacturing levels.
The reasons for this common pattern of premature de-industrialisation are probably a com-
bination of global demand shifts, global competition and technological changes. Whatever the
reason, Africa finds itself in an environment where it is facing much stronger head winds.
Countries with a head start in manufacturing, having developed a large manufacturing base
behind protective walls as in both Europe and Asia, make it difficult for Africa to carve a space
for itself, especially as global demand shifts from manufacturing to services. Having liberalised
trade, African countries have to compete today with Asian and other exporters not only on
world markets but also in their domestic markets. Earlier industrializers were the product of
not just export booms but also considerable amount of import substitution. Africa is likely to
find both processes very difficult, even under the best of circumstances.
What about the second scenario of agriculture-based growth? Since so much of Africa’s
workforce is still in agriculture, does it not make sense to prioritise agricultural develop-
ment? Without question, there are many unexploited opportunities in African agriculture,
whether in perishable non-traditional products such as fruits and vegetables or perishable
cash crops such as coffee.
Agricultural diversification seems to be hindered by many of the same obstacles as
manufacturing. The term ‘poor business climate’ applies equally well here too (Golub and
Hayat, 2014). In addition, agriculture has special problems that governments need to fix,
such as extension, land rights, standard setting and input provision. Once again, the
exchange rate can be an important compensatory tool.
The main argument against this scenario is that it is very difficult to identify historical exam-
ples of countries that have pulled such a strategy off. Agriculture-led growth implies that coun-
tries would sell their agricultural surplus on world markets, and that their export basket would
remain heavily biased towards farm products. Yet one of the strongest correlates of economic
development is export diversification away from agriculture. It is true that Asian countries such
as China and Vietnam have benefited greatly from an early spurt in agricultural productivity—
something that is particularly helpful for poverty reduction. But in all cases, the subsequent and
more durable boost came from the development of urban industries. Moreover, even if modern,
16 Dani Rodrik
non-traditional agriculture succeeds on a large scale in Africa, it is unlikely that this will reverse
the process of outmigration from the countryside. More capital and technology-intensive farm-
ing may even accelerate this process. So one way or other African countries will need to develop
an array of high productivity sectors outside of agriculture.
The third scenario of growth in service productivity is one that perhaps raises the largest
numbers of questions. When I lay out my pessimism on industrialisation to audiences familiar
with Africa, invariably I hear back a litany of success cases in services—mobile telephony and
mobile banking are the most common—which seem to lead to a more optimistic prognosis.
With few exceptions, services traditionally have not acted as an escalator sector like
manufacturing. The essential problem is that those services that have the capacity to act as
productivity escalators tend to require relatively high skills. The classic case is information
technology, which is a modern, tradable service. Long years of education and institution

building are required before farm workers can be transformed into programmers or even
call centre operators. Contrast this with manufacturing, where little more than manual dex-
terity is required to turn a farmer into a production worker in garments or shoes, raising
his/her productivity by a factor of two or three.
So, raising productivity in services has typically required steady and broad-based accu-
mulation of capabilities in human capital, institutions and governance. Unlike in manufac-
turing, technologies in most services seem less tradable and more context-specific (again
with some exceptions such as cell phones). And achieving significant productivity gains
seems to depend on complementarities across different policy domains. For example, prod-
uctivity gains in a narrow segment of retailing can be accomplished relatively easily by let-
ting foreign firms such as Walmart or Carrefour come in. But achieving productivity gains
along the entire retail sector is extremely difficult in view of the heterogeneity of organisa-
tional forms and the range of prerequisites across different segments.
None of this is to say that the past will necessarily look like the future. Perhaps Africa
will be the breeding ground of new technologies that will revolutionise services for broad
masses, and do so in a way that creates high-wage jobs for all. Perhaps; but it is too early
to be confident about the likelihood of this scenario.
Finally, what about natural resource-based growth? Once again, the argument against
this scenario has to be the paucity of relevant examples in history. Almost all of the coun-
tries that have grown rapidly (say at 4.5% per annum) over a period of three decades or
more have done so by industrialising (Rodrik, 2013a, b). In the post-World War II period,
there were two such waves of countries, one in the European periphery (Spain, Portugal,
Italy, etc.) and one in Asia (Korea, Taiwan, China, etc.). Very few countries could achieve
such a performance based on natural resources, and those that did were typically very small
countries with unusual circumstances. Three of these countries were in SSA: Bostwana,
Cape Verde and Equatorial Guinea. What these countries demonstrate is that it is indeed
possible to grow rapidly if you are exceptionally rich in minerals and fuels. But it would be
a stretch of the imagination to think that these countries set a relevant or useful example
for countries such as Nigeria and Zambia, let alone Ethiopia and Kenya.
The downsides of natural resource-based growth patterns are well known. Resource sec-
tors tend to be highly capital-intensive and absorb little labour, creating enclaves within
economies. This is one reason why small economies can generally do better with resource
windfalls. Resource booms crowd out other tradables, preventing industries with escalator
properties from getting off the ground. Resource-rich economies experience substantial
volatility in their terms of trade. And they have great difficulty in managing/sharing resource
rents. Institutional underdevelopment is often the price paid for resource riches. All these fac-
tors help account for why resource-based growth has not paid off for most countries.
5. Is an African growth miracle possible?

The balance of the evidence I have reviewed here suggests caution on the prospects for high
growth in Africa. Much of the recent performance seems to be due to temporary boosts: an
advantageous external context and making up of lost ground after a long period of eco-
nomic decline. While the region’s fundamentals have improved, the payoffs to macroeco-
nomic stability and improved governance are mainly to foster resilience and lay the

groundwork for growth, rather than to ignite and sustain rapid productivity growth. The
traditional engines behind rapid growth and convergence, structural change and industrial-
isation, are operating at less than full power.
So, my baseline would be that we should expect moderate and steady growth, perhaps
as much as 2% per capita, as long as the external environment does not deteriorate signifi-
cantly, and China manages its own substantial challenges well. I hasten to point out that a
growth rate of 2% on a sustained basis is not bad at all. In all likelihood, this will also pro-
duce some convergence with the more advanced economies, largely because the latter will
not do very well in the decades ahead. My story is not one of Afro-pessimism, but one of
curbing our enthusiasm.
I can make another prediction, perhaps one that I feel even more confident about. If
African countries do achieve growth rates substantially higher than what I have surmised,
they will do so pursuing a growth model that is different from earlier miracles based on
industrialisation. Perhaps, it will be agriculture-led growth. Perhaps, it will be services. But
it will look quite different than what we have seen before.
References
Acemoglu D., Naidu S., Restrepo P., Robinson J. A. (2014) ‘Democracy Does Cause Growth’,
unpublished paper.
Acemoglu D., Gallego F. A., Robinson J. A. (2014) ‘Institutions, Human Capital, and
Development’, unpublished paper.
African Center for Economic Transformation (ACET), ‘Growth with Depth—2014 African
Transformation Report’, Accra, 2014.
Bughin J., Chironga M., Desvaux G., Ermias T., Jacobson P., Kassiri O., Leke A., Lund S., Van
Wamelen A., Zouaoui Y. (2016) ‘Realizing the Potential of Africa’s Economies’, McKinsey
Global Institute.
de Vries K., Timmer M., de Vries G. J. (2013) ‘Structural Transformation in Africa: Static gains,
Dynamic losses’, GGDC research memorandum #136.
Feenstra R. C., Inklaar R., Timmer M. (2013) PWT 8.0–a user guide. Groningen Growth and
Development Centre, University of Groningen.
Gelb A., Meyer C. J., Ramachandran V. (2014) ‘Development as Diffusion: Manufacturing
Productivity and Africa’s Missing Middle’, Washington, D.C.: Center for Global
Development, January.
Giuliano P., Mishra P., Spilimbergo A. (2013) ‘Democracy and reforms: evidence from a new
dataset’, American Economic Journal: Macroeconomics, 5 (4): 179–204.
18 Dani Rodrik
Golub S., Hayat F. (2014) ‘Employment, Unemployment, and Underemployment in Africa’,

WIDER Working Paper 2014/014.
Heston A., Summers R., Aten B. (2011) Penn World Table (PWT) 7.0. Center for International
Comparisons of Production, Income and Prices at the University of Pennsylvania.
Johnson S., Ostry J. D., Subramanian A. (2010) ‘The Prospects for Sustained Growth in Africa:
Benchmarking the Constraints’, IMF Staff Papers, vol. 57.
Maddison A. (2010) ‘Statistics on world population, GDP and per capita GDP, 1-2008 AD’,
Historical Statistics.
McCaig B., Pavcnik N. (2013) ‘Moving out of Agriculture: Structural Change in Vietnam’,
NBER Working Paper No. 19616.
McMillan M. S., Rodrik D. (2011) Globalization, structural change and productivity growth
(No. w17143). National Bureau of Economic Research.
Rodrik D. (2008a) ‘Thinking About Governance’, in Governance, Growth, and Development

Decision-Making, reflections by D. North, D. Acemoglu, F. Fukuyama, and D. Rodrik, The
World Bank, pp. 17–24.
Rodrik D. (2008b) ‘The Real Exchange Rate and Economic Growth’, Brookings Papers on
Economic Activity, 2: 365–412.
Rodrik D. (2013a) ‘Unconditional Convergence in Manufacturing’, Quarterly Journal of
Economics, 128 (1): 165–204.
Rodrik D. (2013b) ‘The Past, Present, and Future of Economic Growth’, in Behrman J., et al.
(eds), Towards a Better Global Economy: Policy Implications for Citizens Worldwide in the
21st Century. Oxford, New York: Oxford University Press, pp. 70–137.
Straus S. (2012) ‘Wars Do End! Changing Patterns of Political Violence in Sub-Saharan Africa’,
African Affairs, 111 (443): 1–23.
U.N. Economic Commission on Africa (UNECA), Economic Report on Africa 2014, Addis
Ababa, 2014.
World Bank (2016) Ethiopia: Priorities of Ending Extreme Poverty and Promoting Shared
Prosperity. Washington, DC: Systematic Country Diagnostics.

Ilovepdf Merged

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ilovepdf Merged

Uploaded by

Copyright:

Available Formats

Elena Meschi

INTERNATIONAL ECONOMICS AND

 Preliminary information and organization of the

 Slides of the lectures

 Detailed course outline, slides and reading list available in

 The outline may be updated and additional readings may

 Final grades will be awarded by weighting

 The final written exam consists of theoretical questions,

For general information about the services offered by our University:

“Development Economics” tries to provide solutions to help poor

 Nature of development economics:

 The Development Gap

 References: Roland, chapter 1

 Development multifaceted phenomenon

 The first basic indicator of development is real income

 GDP per capita needs to be expressed in a common

GNP Per Capita (US $)

 Big Mac Index is published by The Economist as an informal way of measuring

 Defining the Developing World

 Upper Middle Income Countries (UMC)

 High Income Countries (HIC)

 GNI and GDP per capita data, and income-level

 The Poverty Gap

 Policy interventions can make a difference in lowering child mortality. For

The Education Gap

The Urbanization Gap

Figure 1.8. Average Annual Growth

 World Bank in its 1991 WDR asserted that the “challenge

 The improved QOL involves higher incomes, better

 “Development is a multi dimensional process involving

 A standard definition of human development (1990

 Emphasizes that outcomes for people and their

 Initiated in 1990 and undertaken by UNDP in its

 HDI= 1/3(Income index)+1/3(Life expectancy

Each component of the HDI is measured in the

 Common characteristics among developing in

 Poverty and Inequality

 References: Roland, chapter 2

Poverty is the most visible feature of developing countries

 Absolute poverty: is a condition where household

 Relative poverty – A condition where household

 Poverty headcount: number of people (Q) in the

 Poverty headcount ratio (incidence of poverty):

 These indexes do not reflect the degree of

 Average poverty gap: 𝐴𝑃𝐺 = 𝑇𝑃𝐺/𝑁

 Table 2.1. Example of Poverty Headcount Ratio and Average Poverty

 Country B’s poverty is more serious

 Progress on Extreme Poverty

 The Millennium Development Goal of halving 1990

 The world's extremely poor people are distributed very

Poverty is still predominantly a

Very high among people in rural

Indigenous Poverty in Latin America

 Data on Inequality over time and across countries

 Economic interpretation of Inequality:

 Ifa single person holds all of a given resource,

 Size distributions (quintiles, deciles): useful to measure the

 The quantile ratio is the average income in the highest quantile

 The Lorenz curve plots the cumulative share of

 Curvature of Lorenz curve summarises inequality

 Complete inequality, all income held by just one

The Lorenz Curve and the Gini Coefficient

In this example, the Range = $1,000,000-$24,000

The range is simply the difference between the highest and