1
CHAPTER I
INTRODUCTION
This paper assesses international trade in Sub-Saharan Africa (SSA) and whether
trade is affected by regional trade agreements and quality of institutions. SSA countries
have abundant natural resources and opportunities, yet development in these countries
continues to stagnate, corruption is endemic in most countries, and industrialization and
specialization have not taken off. In fact, the 2010 Governance Matters Control of
Corruption Indicator shows that only two percent of SSA countries rank in the top 25
percent of countries and 80 percent are in the lowest fifty percent (Kaufmann, Kraay, and
Mastruzzi 2011).
Theories examined in this paper include those based both on economic geography
and the effect of institutions on development. Krugman (1991, 1), defines economic
geography as “‘the location of production in space’ [;] that is, that branch of economics
that worries about where things happen in relation to one another.” North (1994, 3)
defines institutions as
The humanly devised constraints that structure human interaction. They are made
up of formal constraints (e.g., rules, laws, constitutions), informal constraints
(e.g., norms of behavior, conventions, self-imposed codes of conduct), and their
enforcement characteristics. Together they define the incentive structure of
societies and specifically economies.
Acemoglu, Johnson, and Robinson (2004) and North (1990) identify that
economic institutions, which align resources with their most efficient resources, are the
2
determining factor for whether a country is able to develop and grow, and they focus
particularly on the importance of such institutions as property rights and open markets.
Acemoglu says that economic institutions define both an economy’s potential for
economic growth and the distribution of wealth resulting from economic growth.
Dufrenot, Diop, and Sanon (2010) believe that failures of governance and institutions
explain the stagnation of growth in per capita income in the ECOWAS (Economic
Community of West African States) countries.
Combes, Mayer, and Thisse (2008) state that economic geography addresses
factors impacted by location, such as where producers are located in relation to
consumers, and the benefits and costs of location-related choices. Puga and Venables
(1999) additionally propose that with agglomeration, all countries do not develop at the
same pace. They suggest that poor countries transition to a developed state one at a time,
at a rapid pace.
African countries have an abundance of labor and availability of similar primary
resources with similar relative factor costs in each country. With similar resources, there
may not be reason for trade among these countries. In general, lack of variability, means
that these countries should not have comparative advantage over each other. This limits
trade potential both among African countries and between African countries and the rest
of the world (Hanink and Owusu 1998). Yet the dilemma remains of how these countries
will grow and develop without trade. This paper focuses on these SSA trade agreements:
South African Development Community (SADC), East African Community (EAC), the
Common Market for Eastern and Southern Africa (COMESA), the Economic
3
Community of West African States (ECOWAS) and the Economic Community of
Central Africa States (ECCAS).
Musila (2005) analyzes trade creation and trade diversion in COMESA, ECCAS
and ECOWAS. Musila observes that prior analyses do not fully address South-South
trade and whether it is creating trade (countries start importing a product previously
produced locally by inefficient producers) or diverting trade (with imports from lower-
cost non-member states to higher cost member states). He cites that the Economic
Commission for Africa argues that regional trade can actually promote market
diversification and reverse deindustrialization and marginalization because it provides
wider trading and investment environment that leads to economies of scale, induces
backward and forward linkages and promotes diversification within regional markets to
build experience before entering global markets. This contradicts the prediction of Yeats
(1990) that these blocs will have a negative impact on industrialization and will divert
imports from low to higher cost sources.
Gbetnkom (2006) summarizes that ECOWAS continues to be effective in creating
trade after the Structural Adjustment Programme (SAP) liberalization reforms were
enacted. His analysis contradicts Musila’s posit that trade diversion does not occur in
ECOWAS, and he says that trade creation and trade diversion occur simultaneously.
Recent studies indicate that trade liberalization efforts made in the 1980s may have
resulted in increased intra-ECOWAS trade in the 1990s, and additional efforts to reduce
trade barriers, specifically tariffs, in the 21st century, may have brought further
improvements among ECOWAS countries (Arbache, Go, and Page 2008).
4
In an analysis of changes in the distribution of poverty throughout the world,
Sala-i-Martin observes that world poverty rates declined significantly in the thirty years
from 1970 to 2000, but that the improvements occurred primarily in Asian countries,
while the situation in African countries worsened during that same time period. He says
that poverty today is primarily an African problem, with 68 percent of the world’s poor
living in Africa (Sala-i-Martin 2006).
Baliamoune-Lutz and Ndikumana (2007) state that when trade openness is low
the combined impact of institutions and trade liberalization is negative on growth. As
trade openness increases, quality institutions have a positive impact on growth. North
(1991) explains the importance of institutions to trade, primarily because of their ability
to reduce the uncertainty inherent in transactions. In order to have effective trade across
international boundaries, there has to be an ability to establish agency with remote trading
partners, contract enforcement and secure property rights. These factors are generally
established by effective institutions.
De Groot et al (2004) analyze international trade data with the incorporation of
institutional quality indicators and determine that countries with similar institutional
frameworks experience 13% more bilateral trade on average. They conclude institutional
quality impacts bilateral trade flows. In addition, De Groot et al (2004) state that if
institutions are not taken into account, then the gravity model experiences omitted
variable bias. In relation to this paper, De Groot et al do not specifically examine Africa
nor do they examine impact of trade agreements. None of the prior works reviewed that
examine trade in SSA countries incorporate institutional metrics as explanatory variables.
5
The World Bank (Kaufmann, Kraay, and Mastruzzi 2009, 5) has defined
governance as “the traditions and institutions by which authority in a country is
exercised. This includes the process by which governments are selected, monitored and
replaced; the capacity of the government to effectively formulate and implement sound
policies; and the respect of citizens and the state for the institutions that govern economic
and social interactions among them.” The World Bank (2011) has developed six
indicators of good governance:
1. Voice and Accountability – measuring political, civil and human rights
2. Political Instability and Violence – measuring the likelihood of violent threats
to, or changes in, government, including terrorism
3. Government Effectiveness – measuring the competence of the bureaucracy and
the quality of public service delivery
4. Regulatory Burden – measuring the incidence of market-unfriendly policies
5. Rule of Law – measuring the quality of contract enforcement, the police, and
the courts, as well as the likelihood of crime and violence
6. Control of Corruption – measuring the exercise of public power for private
gain, including both petty and grand corruption and state capture
These six indicators are developed with a reduction technique called unobserved-
components-model which is a variant of factor analysis that extracts components
common to the 300 different measures in the original data (Kaufmann, Kraay, and
Mastruzzi 2011). The concept is that a good government is one that complies with Adam
Smith’s guidance that institutions result in greater prosperity (Smith 1776).
6
This analysis is conducted using the gravity model. The gravity model examines
reasons for trade among countries. The gravity model is a bilateral trade flow model
where relationships are examined from each country to each of the other countries being
examined, so that for n countries in the model, there are (n*(n-1)) relationships. It is
inspired by Newton’s gravity equation in which two bodies are attracted in proportion to
their masses and distance. The gravity model analyzes trade between a pair of countries
as an increasing function of incomes (size) and a decreasing function of their distance
(Aitken 1973).
Credit is given to both Tinbergen and Pöyhönen in the early 1960s for
concurrently and independently developing this model of international trade (Sandberg
2006). Tinbergen (1962) identifies the gravity model to examine the flow of trade
between two countries utilizing three primary indicators: the gross national product
(GNP) of the exporting country, the GNP of the importing country, and distance.
Tinbergen further explains the justification for these three indicators, the export GNP
indicates the size of a country’s market (or the amount that can be sold), the import GNP
indicates the amount a country can buy (or the country’s market) and distance is a proxy
for transportation costs.
The dependent variable for gravity model analysis is the amount of trade between
two countries. The gravity model assesses whether total exports, or trade, between two
countries is a function of the selected independent variables. Export data is collected
from the International Monetary Fund’s Direction of Trade Statistics dataset (2014). The
gravity model for this paper states that the trade between two countries is a function of
7
the GDP of both the exporting and importing countries, the distance between the major
cities of the two countries, dummy variables for whether or not the countries share a
common border, whether or not the countries have a common official or ethnic language,
whether or not the exporting or both countries are ECOWAS, EAC, COMESA, SADC or
ECCAS members, and what is the quality of institutions of the exporting and importing
country, using an average of the six individual indicators.
Hypotheses
H1: Membership in an SSA trade bloc increases trade for member countries:
Countries that are members of SADC, COMESA, EAC, ECCAS or ECOWAS
experience higher trade than countries that are not members of one of those five
trade blocs.
H2: Membership in an SSA trade bloc increases trade between members of the trade
bloc: Trade between countries within the trading blocs (member-to-member) is
greater than trade with countries outside of the trading bloc.
H3: Institutional quality affects trade in SSA countries: countries with higher quality
of governmental institutions experience more trade
Countries that trade are countries that grow (Dollar and Kraay 2003). The fact that
some countries trade more than others drives differentiation and results in some countries
being richer than others. SSA countries not only experience more less growth and more
poverty, but they have lower quality of institutions.
This paper utilizes regression analysis to assess the impacts of regional trading
blocs and institutional quality on trade. The model will utilize OLS regression in 2010
8
with comparison to 1980, 1990, and 2000, and will compare to Tobit regression
conducted in the same years. Countries included in the trade bloc and institutional
quality analysis are SSA countries in the five identified trade blocs, additional SSA
countries not in the trade blocs, and additional countries included in similar studies for
which export trade data is available for the selected years. The government quality data is
only available after 1995.
This paper continues with a review of literature in Chapter II, a discussion of
methods in Chapter III, a presentation of results and analysis in Chapter IV, and a
conclusion in Chapter V.
9
CHAPTER II
LITERATURE REVIEW
This dissertation examines the occurrence of trade over geographic space, and
how it is impacted by institutional quality and regional trade agreements. The paper looks
at economic geography and institutional quality and the impacts of regional trade
agreements.
All economic activities take place in a location, or across locations. This is easily
observable and can be viewed at the micro level of an individual business choosing a
location for a factory and at the macro level with trade occurring across industries and
nations. Underlying this simple observation are theories that seek to explain such details
as why a business decides to open a factory in a given location, why concentrations of
factories and industries occur, or why trade in goods occurs across space – whether that
be in towns, cities, regions or nations. Spatial economics, or economic geography,
attempts to address questions such as these.
Krugman (1991, 1) defines economic geography as “the location of production in
space; that is, that branch of economics that worries about where things happen in
relation to one another.” According to Krugman, economics is intrinsically geographic,
but examination and utilization of the geographic aspect has been either consciously
removed from theory or sometimes just forgotten since the 1960s. Combes, Mayer, and
Thisse (2008) state that economic geography addresses factors impacted by location, such
as where producers are located in relation to consumers, and the benefits and costs of
location-related choices.
10
Economies of scale, where it costs less to produce a lot of things at one location
than to produce fewer things at many locations, is part of a more comprehensive set of
events known as agglomeration. Agglomeration is the clustering of economic activity,
created and sustained by a sort of circular logic. Agglomeration can occur in a small local
market or across large areas that serve a world market. The circular concept is self-
reinforcing. Firms want to locate where market potential is high, and markets will tend to
be large where lots of firms locate (Krugman 1995). The circular phenomenon of
agglomeration is also known as the snowball effect, which results in its continuous
reinforcement once it is set in motion (Combes, Mayer, and Thisse 2008).
Occurrences of agglomeration have two key points in their development. The
break point is when a uniform spatial economy spontaneously begins to develop
concentrations of population and/or industry, and it truly becomes an agglomeration
when it reaches the sustain point where the established agglomerations are able to survive
even under conditions that would not have caused them to form in the first place (Fujita,
Krugman, and Venables 1999).
The question remains of how the agglomeration actually forms and then increases
in size through this snowball effect. There are two aspects to this, one related to the
workers and the other related to the industries. The initial development of industry in a
location draws workers. The increase in the number of workers is also an increase in the
number of consumers. This in turn pushes up the local demand for the manufactured
good, triggering the installation of more firms in the region. This is termed backward
linkages.
11
The infusion of more firms pushes nominal wages upward. With the number of
firms located in the region increasing, the variety of products produced also increases.
The result of this is that the equilibrium price index of manufactured goods decreases in
this region. Together the increase in workers and nominal wages and the decrease in price
index results in an increase in real wages which then induces a new flow of workers who,
all else being equal, benefit from a higher standard of living. This is termed forward
linkages. If the effects of forward and backward linkages are combined, migration toward
the large market should continue until the whole industry is concentrated there (Combes,
Mayer, and Thisse 2008). As a result of these impacts, locations with a higher demand
for manufactures may pay higher nominal wages (Fujita, Krugman, and Venables 1999).
Jones and Kierzkowski (2004) point out that increases in the outsourcing of
economic activity, whether nationally or globally may lead to new forms of
agglomeration. Industries may have fragments that are similar from one industry to
another even though the integrated whole is not the same. This fragmentation can lead to
a horizontal spread wherein the similarity between fragments across industries promotes
technological progress to make fragments even more similar and thus encourages new
forms of agglomeration An example of this might be the outsourcing of call centers to
India across a number of industries, which has resulted in a specialization of the call
center service industry, whether supporting technology industry or retail industry.
There are a number of impacts of distance related to economic geography, and
these impacts are assessed with a model that identifies those impacts in relation to
international trade, the gravity model. Trade costs are frequently approximated with data
12
on the distance between trading partners. This information is intended to combine
transport costs, transport time (addressing perishability, adaptability to market conditions,
and irregularities in supply), and psychic distance (addressing familiarity with laws,
institutions, and habits) (Junius 1999).
In order to understand barriers to trade and their impacts, the gravity model
analyzes costs of doing business over geographical space. The theory includes the costs
of doing business at a distance (including lack of face-to-face contact, more complex and
expensive communications and information gathering, and different languages, legal
systems, product standards and cultures). If amount of trade between two countries
decreases with distance further apart they are, this could be due to increased costs of
trade. The way to evaluate that is with the gravity model (Fujita, Krugman, and Venables
1999).
The gravity model has both robust empirical results and simplicity of functional
form. Helpman (1987) derives a structural form of a gravity equation out of an
international trade model with imperfect competition. He shows that the larger the
similarity in factor compositions, the larger the share of intra-industry trade, and changes
over time in relative country size can explain rising trade-to-income ratios. Bergstrand
(1989) combines Heckscher-Ohlin elements, Chamberlinian monopolistic competition
and the Linder hypothesis in his model (Junius 1999).
Given that distance does not change over time, seeing a declining regression
coefficient of geographic distance over time would indicate a decline in the marginal
effect of distance between trading partners (Junius 1999). According to Pomfret (1997),
13
distance is a strong factor in the size of bilateral trade flows, but he also says that distance
is largely ignored in trade theory. Pomfret (1997) also provides a succinct summary that
the gravity model depends upon the size of two economies, measured by total output and
the distance between them.
Though the costs of transportation have declined dramatically over the past 200
years, the costs of trade continue to be important explanatory variables for trade between
countries. Effects of contiguity and common language are large and are estimated to
multiply trade by 2 and 2.6 respectively. Contiguity and common-language effects have
increased over time, so that sharing a common language or border has an impact on trade
that is higher now than it was thirty, forty, or fifty years ago (Combes, Mayer, and Thisse
2008).
When assessing impacts to trade, these are the types of trade barriers are generally
considered:
1. Natural barriers related to physical geography – distance, mountains, coastal – are
grouped together in what we will call transport costs
2. Trade policy measures, or those with an environmental focus, as well as exchange
rate transaction costs for countries that do not share the same currency.
3. Information costs
4. Cultural differences
These four factors are all considered to impact trade costs and trade preference.
All, in fact, restrict trade, but they also simplify the actuality of trade barriers by not
accounting for more complex differences such as those from protectionist policies,
14
transportation infrastructure and insufficient explanation for border effects (Combes,
Mayer, and Thisse 2008). There is an alternative method of assessing trade costs without
using the gravity model which is based on the Dixit-Stiglitz-Krugman model. This model
defines the overall degree of market segmentation indirectly through price differentials
and the gap between actual trade flows (Combes, Mayer, and Thisse 2008).
Regions are an important part of economic geography because as nations have
evolved tariffs, taxes and other protectionist strategies, formal regional agreements have
been implemented to attempt to eliminate national borders as a barrier to trade. There are
a couple of different uses of the term region. A region could be a trading bloc of
countries who have set up a formal agreement that includes reducing or eliminating trade
barriers between them that are within their control. In this instance of the term, labor is
usually considered immobile due to the challenges of immigration. The term region may
also be used to refer to an area where a particular product is produced – so there could be
certain types of manufacturing or agricultural regions that have many characteristics in
common, but do not have an agreement in place and are not necessarily trying to work
together across the common industry. Frequently this type of region resides within a
single country, in which case labor resources would be considered mobile.
During the 1800s, the economic development of Europe decreased at the same
time that distance-related costs were declining. In this situation, though costs to engage in
trade were lower, increased trade did not occur in the European economic region,
demonstrating that regional economics are uneven. (Combes, Mayer, and Thisse 2008).
15
When transport costs are high, interregional trade will generally be low. Wages
will depend primarily on local competition. When transport costs are low, a typical firm
sells extensively in both regions. Proximity to regions with larger population improves
access to markets. In that situation, it can then afford to pay higher wages. Consequently,
the purchasing power of the wages is also higher because workers have better access to
consumer goods. Real wages would be increasing in the region’s population. (Combes,
Mayer, and Thisse 2008).
There are two types of regional economic activities: those that are the region’s
economic base, also called the export base, because it is the region’s response to demands
from outside of the region, and the second type are those activities that supply goods and
services locally. There are some instances where the goods provided locally and
regionally are one and the same, which serves only to increase the economies of scale,
but sometimes a firm may have completely different or sometimes slightly different
products that they offer at home and outside of their region. When the regional economy
is strong, a firm may offer many more goods and services in their local market because
the market can afford it and helps with developing new products that can be sold outside
of the region (Fujita, Krugman, and Venables 1999).
As a result of the advantages to having a firm set up in the larger region with a
strong local market, firms that are established in the larger regions, generally see higher
profits than those in the smaller region. Again, this dynamic could also be self-fulfilling;
the core region would therefore attract more firms, increasing inequalities between the
core and the periphery. At the same time, when firms choose to set up in the core region,
16
they are choosing to compete and should expect to find competition waiting for them
(Combes, Mayer, and Thisse 2008). Remarkably, spatial structure can evolves by itself
(Fujita, Krugman, and Venables 1999)
Combes, Mayer, and Thisse (2008) also point out that economic geography is
differentiated from international trade by the fact that in international trade, all factors are
considered immobile, but in the study of economic geography, some factors are mobile
(non-labor). As a result, regional differences can arise due to firm location based on
market size and market access. The Dixit-Stiglitz model indicates that market-size effects
are a result of changes in variety (Fujita, Krugman, and Venables 1999).
Krugman’s model of immobility versus mobility helps to develop understanding
of regional unevenness of development. Fujita, Krugman, and Venables (1999) pose a
mechanism by which regional markets develop at the international level. Beginning with
one region or country that has established a self-reinforcing market in a manufacturing
sector, this region is now able to offer higher wages than other parts of the world. The
demand for this particular product rises and the region that has the expertise in this
commodity benefits from agglomeration and from the ability to increase wages. As the
wage gap increases, firms are able to establish and be competitive in a second region.
That region then begins to develop its own self-reinforcing market, and is able to increase
wages as well. This can continue to happen in additional regions.
Puga and Venables (1999) propose that with agglomeration, all countries do not
develop at the same pace. They suggest that poor countries transition to a developed state
one at a time, at a rapid pace. It is posed by Fujita, Krugman, and Venables (1999) that
17
such uneven development may have been a predictable result over time of the increase in
world integration. They also suggest that some recent narrowing of income gaps (China,
for example), might also be related to the same phenomena – once a country is able to
participate effectively in world trade, the dynamic changes dramatically because as trade
costs decrease, global inequalities also decrease (Fujita, Krugman, and Venables 1999).
Thus economic geography also becomes a mechanism to begin to understand the impacts
of economic integration on spatial inequality. According to Combes, Mayer, and Thisse
(2008), trade costs actually can serve as a proxy for economic integration, and a reduction
in these costs means that there is a greater degree of integration.
SSA countries have some of the greatest natural resources and opportunities in the
world, yet development continues to stagnate, corruption is endemic in most countries,
and industrialization and specialization have not taken off. Between 1995 and 2005, there
have been changes in Africa. According to Arbache, Go, and Page (2008), SSA leaders
are taking increasing control of their economic destiny. After 45 years of stagnation,
improvements are being made but the authors do not believe these changes are resulting
in significant improvements in economic development. This dissertation examines
whether these changes may be starting to drive increased trade over time by specifically
examining the impacts of regional trading blocs and institutional quality.
This paper focuses on countries in sub-Saharan Africa, and the ECOWAS
regional trading block within West Africa. According to the Centre for Democracy and
Development (2002), a regional trade agreement, or regionalism, is earmarked by a group
18
of countries with geographic proximity, and a common purpose seeking to improve an
economic or political objective of the countries within the agreement.
Recent studies indicate that trade liberalization efforts made in the 1980s may
have resulted in increased intra-ECOWAS trade in the 1990s, and additional efforts to
reduce trade barriers, specifically tariffs, in the 21st century, may have brought further
improvements among ECOWAS countries (Arbache, Go, and Page 2008).
African countries now numerous, sometimes overlapping regional trade
agreements that likely dilute the impact of any one trade agreement. In fact, only seven of
the 53 African countries are members of just one regional agreement. Twenty-seven
countries are members of two regional integration agreements, and 18 are members of
three. One country, the Democratic Republic of Congo, is a member of four different
organizations (Nyirabu 2004).
The development of regional trade agreements in Africa has been considered a
preliminary step to unification of Africa. Unifying Africa into something of the sort of
the United States of Africa has been a topic of discussion since it was introduced by
Kwame Nkrumah of Ghana and others immediately after colonial independence
(Nkrumah 1970). While Nkrumah sought immediate unity, Tanzanian leader Julius
Nyerere proposed that unity must be phased in over time via regional alignments, and
that is the approach that was taken. In 1980, the Lagos Plan of Action encouraged
African countries to create regional agreements as a step towards the ultimate creation of
an African Economic Community. The Lagos Plan identifies three steps to achieving an
integrated African economy: 1) trade liberalization, 2) customs unions, and 3) a single
19
economic community. In 1990, the Abuja treaty was signed as a commitment to
achieving the African Economic Community by 2025 (Nyirabu 2004).
In 2008, three trade blocs, COMESA, SADC and EAC joined together to form the
African Free Trade Zone (AFTZ). The AFTZ includes 26 countries that span the length
of Africa, but has not been in place for sufficient time to identify changes as a result of
the agreement, and there is almost no available documentation beyond the original
announcement on October 22, 2008 (BBC 2008).
While there has always been interest in a unified Africa, there are many reasons
why this has not happened yet which seem to align with potential reasons that SSA
countries are not developing. These reasons include, “lack of grassroots support,
excessive external dependency, institutional weakness, multiplicity of organisations (sic),
politics, underdeveloped economies, the international economic structure and distribution
of the benefits of integration” (Nyirabu 2004, 23).These justifications for lack of unity
are also cited as the argument for why economic integration is needed in Africa.
This dissertation focuses on the following trade agreements, SADC, EAC,
COMESA, ECCAS and ECOWAS. The SADC was formed in August 1992 with 14
countries (see Table 2.1) and a goal to introduce free trade in 2008 and liberalize country-
specific products by 2012. According to Nyirabu (2004), the SADC has established four
centralized directorates, negotiated 21 protocols, implemented 10 of them, and has
experienced a small increase in trade among the member states. Within the SADC, five
countries, Botswana, Lesotho, Namibia, Swaziland and South Africa, have successfully
established a customs union with no internal tariffs.
20
Table 2.1
Members of Selected Sub-Saharan African Regional Trade Agreements
COMESA*
EAC*
ECCAS
ECOWAS
SADC*
Burundi
Comoros
D.R. Congo
Djibouti
Egypt **
Eritrea
Ethiopia
Kenya
Libya **
Madagascar
Malawi
Mauritius
Rwanda
Seychelles
Sudan
Swaziland
Uganda
Zambia
Zimbabwe
Burundi
Kenya
Rwanda
Tanzania
Uganda
Angola
Burundi
Cameroon
Central African
Republic
Chad
Congo
(Brazzaville)
D.R. Congo
Equatorial
Guinea
Gabon
Rwanda
Sao Tome et
Principe
Benin
Burkina Faso
Cape Verde
Cote d’Ivoire
The Gambia
Ghana
Guinea
Guinea-Bissau
Liberia
Mali
Mauritania
Niger
Nigeria
Senegal
Sierra Leone
Togo
Angola
Botswana
D.R. Congo
Lesotho
Malawi
Mauritius
Mozambique
Namibia
Seychelles
South Africa
Swaziland
Tanzania
Zambia
Zimbabwe
* Members of Africa Free Trade Zone (established 2008)
** Not SSA countries
COMESA was established in 1993 with 20 member countries and goals to
establish a customs union, a common external tariff (CET) and to align policies of
member countries by 2004. As of Nyiarabu’s (2004) assessment, only nine of the
countries had ratified the free trade protocol, but they have seen a 30 percent increase in
intra-COMESA trade after the initiation of free trade in October 2000.
21
The EAC began in 2000 with the objective of implementing a customs union by
2003 to be followed by a common market, monetary union and then a political federation.
As of 2004, they were still finalizing their customs protocol (Nyirabu 2004).
ECCAS was signed as a treaty in October 1983 and was inactive from 1992 until
recently, primarily due to lack of member fee payment and the conflict in the D.R.
Congo. ECOWAS was created on May 28, 1975 in Lagos, Nigeria. There are 16 member
states: Benin, Burkina Faso, Cape Verde, Ivory Coast, Ghana, Gambia, Guinea, Guinea
Bissau, Liberia, Mali, Mauritania, Nigeria, Niger, Senegal, Sierra Leone and Togo. The
primary objective of ECOWAS according to its treaty is: “…accelerated and sustained
economic development of [the member] states and the creation of a homogeneous
society, leading to the unity of the countries of West Africa, by the elimination of all
types of obstacles to the free movement of goods, capital, and persons” (Hanink and
Owusu 1998, 365). ECOWAS did not implement its comprehensive trade liberalization
approach until 1990, and these changes did not result in the elimination on internal tariffs
completely or the creation of a common external tariff (CET).
Most of the ECOWAS countries participated in Structural Adjustment
Programmes (SAP) in the late 1980s. The SAP reforms contributed to a reductions in
tariff and in non-tariff trade barriers. These reforms included lower tariffs, reduction of
prohibitive tariffs, and application of low tariffs on certain import goods (Gbetnkom
2006). These reforms resulted in the following policy changes:
more flexible exchange rate,
reduced export taxes,
22
lower import controls,
less variability import tariffs,
privatization of state-run entities, and
removal of price control and restrictions on private business (though
Ghana eliminated restrictions and licenses on exports and imports in
January 1989).
Musila (2005) analyzes trade creation and trade diversion in COMESA, ECCAS
and ECOWAS. Musila observes that prior analyses do not fully address South-South
trade and whether it is creating trade (countries start importing a product previously
produced locally by inefficient producers) or diverting trade (with imports moving from
lower-cost non-member states to higher cost member states). He cites that the Economic
Commission for Africa argues that regional trade can promote market diversification and
reverse deindustrialization and marginalization because it provides wider trading and
investment environment that leads to economies of scale, induces backward and forward
linkages and promotes diversification within regional markets to build experience before
entering global markets. This contradicts the prediction of Yeats (1990) that these blocs
will have a negative impact on industrialization and will divert imports from low to
higher cost sources.
Gbetnkom (2006)summarizes that ECOWAS continues to be effective in creating
trade after the Structural Adjustment Programme (SAP) liberalization reforms were
enacted. His analysis contradicts Musila’s posit that trade diversion does not occur in
ECOWAS, and he says that trade creation and trade diversion occur simultaneously.
23
Table 2.2
Comparison of Gravity Model Studies
Author
Scope of
Trade
Model
Analysis
Categories/ Variables
Countries
Foroutan and
Pritchet (1993)
Sub-Saharan
Africa
1980-82
Gravity
Model
Tobit
Regression
(replace 0s
with small
positive
values)
African
African Oil Export
Distance
Adjacent Border
GDP
GDP per capita
Area
Island
ASEAN
LAFTA
CACM
Lime
CEAO
ECOWAS
UDEAC
Language
Hanink and
Owusu (1998)
ECOWAS
countries
1973 and 1993
Gravity
Model
Trade
Intensity
Index and
Tobit
Regression
GNP exporter
GNP importer
Distance
Adjacent Border
Language
CFA (pegged
currency)
Arab
CEAO
ECOWAS
ECOWAS
Algeria
Cameroon
Central
African Rep
Chad
Congo
Equatorial
Guinea
Gabon
Libya
Morocco
Musila (2005)
COMESA,
ECCAS and
ECOWAS
Countries
1991-1998
Gravity
Model
WLS
GNP exporter
GNP Importer
Population
Distance
Adjacent Border
Language
CFA
COMESA/ ECCAS/
ECOWAS – Intra-
trade, Extra-imports,
Extra-exports
24
Table 2.2 (continued).
Author
Scope of
Trade
Model
Analysis
Categories/ Variables
Countries
Gbetnkom(2006)
Uses 28
reporting
countries, 21
in SSA and 33
partner states
1994-2003
(uses avg 94-
98 and 99-03)
Gravity
Model
GDP exporter
GDP importer
Distance
Population
Difference in per
capita income
CEMAC
COMESA
ECOWAS
ECOWAS importer
Language
Subramanian and
Tamirisa (2003)
Non-EU
world 73
country
sample
1980, 1990,
2000
Gravity
Model
Non-Linear
estimation
GDP
Population
Distance
Remoteness
Anglophone
Francophone
Intra-Anglophone
trade
Intra-Francophone
trade
Non-EU Adv.-
Anglophone Trade
Non-EU Adv.-
Francophone Trade
Developing Country
– Anglophone Trade
Developing Country-
Francophone Trade
CFA
Free Trade Agmt
Primary Commodity
Exporter
Time Trend
Studies have shown that ECOWAS may be effective only because of
longstanding relationships between the countries, with or without the ECOWAS
agreement. Recent studies indicate that trade liberalization efforts made in the 1980s may
have resulted in increased intra-ECOWAS trade in the 1990s, and additional efforts to
25
reduce trade barriers, specifically tariffs, in the 21st century, may have brought further
improvements among ECOWAS countries (Arbache, Go and Page 2008).
Research on trade impacts of regional trading blocs in SSA countries includes
Hanink and Owusu (1998) who took data snapshots from 1980 and 1990. They find small
significant increase in trade from the start of the Economic Community of West African
States (ECOWAS) regional trade agreement through 1990.
Bagwell and Staiger (2002) propose an economic model to explain why trade
agreements are needed. Their model is based on the premise that governments, in setting
tariffs, seek to escape from a Prisoner’s Dilemma. They state that, in setting tariffs,
governments respond to a terms-of-trade externality which means that they set import
tariffs to maximize welfare, though some tariff costs fall to exporters whose products sell
at a lower price.
The Prisoner’s Dilemma faced by governments, is that in the absence of
assurance regarding the actions of other nations, the government must look out for their
own welfare and defect or take a defensive stance, thus putting in place higher tariffs than
might be efficient or optimal.
The Prisoner’s dilemma is usually drawn in a two by two matrix, describing that
the least optimal quadrants are when one country cooperates and the other does not. The
Nash equilibrium equates to the lower right hand quadrant where neither country
cooperates. The optimal solution, is when both countries cooperate. (Spencer and
Brander, 2008), and agree jointly to give up something, such as tariffs, under a managed
agreement so that all involved countries will benefit.
26
Table 2.3
Prisoner’s Dilemma Matrix Hughes (2007)
Country B – Cooperates
Country B – Defects
Country A – Cooperates
Optimal Scenario
Good for A/Bad for B
Country A - Defects
Good for B/ Bad for A
Suboptimal outcome for
both = Nash Equilibrium
Bagwell and Staiger (2002) pose that achieving the optimal solution, or escaping
from a terms-of-trade-driven prisoner’s dilemma, happens through a trade agreement.
Their trade model describes this effect in comparison to the Nash outcome, where no one
benefits. The trade model is based on the general equilibrium model, using two countries
that trade two goods. They first model the inefficient dynamic of the non-cooperative
Nash equilibrium. In this model, Bagwell and Staiger define the contract curve, which
are the tariffs that are efficient and yield greater-than-Nash welfare for each country.
Thus, the goal of a trade agreement would be to negotiate tariffs that would fall on the
contract curve, or efficiency frontier. This curve may include the reciprocal-free-trade
point (where both countries eliminate all tariffs and engage in free trade), but includes
many other pairs as well.
27
Non-cooperative Nash
Equilibrium
N
E
E
Contract Curve
(Bold Portion)
1
1Ŧ*
Ŧ
R
Reciprocal-Free-Trade
Point
Figure 2.1. Non-cooperative Nash Equilibrium (Ŧ = domestic tariff +1 and Ŧ* = foreign
tariff + 1)
According to Bagwell and Staiger (2002), governments seek to address their
interests in both economic-efficiency (national welfare) consequences of local-price
movements as well the distributional or political consequences when negotiating a trade
agreement. Overall, the goal of a trade agreement is to move from a situation approaching
the Nash Equilibrium, where trade policy is set unilaterally and costs are shifted from one
country to another via change in world price caused by the tariffs, to one that approaches
the efficiency frontier through reciprocal trade liberalization. Reciprocal means that both
governments must agree to reduce their tariffs below the Nash equilibrium in order to
have a successful outcome, which is frequently the intent of a PTA. If not reciprocal, the
28
result is that the pair ends up one of the two quadrants of the matrix in which the net
outcome is worse than the Nash outcome for one of the countries.
E
E
Contract Curve
(Bold Portion)
1
Ŧ
R
Reciprocal-Free-Trade
Point = Agreement
Point
Figure 2.2. SSA Countries with no tariffs (Ŧ = domestic tariff +1 and Ŧ* = foreign tariff
+ 1)
The benefit of the perspective shown in Figure 2.2 is that it frees up the efficiency
locus, by stating that the most efficient tariff point may not be the point of reciprocal free
trade, which provides many more outcomes that are possible. In addition, by
demonstrating this, Bagwell and Staiger have shown that while a government may have
both nation.al welfare and political motivations, political considerations are not the only
driver for a trade agreement and can be incorporated with an economic motivator in a
single economic trade model.
29
N
E
E
1
1Ŧ*
Ŧ
R
Reciprocal-Free-
Trade
Point
Politically Optimum (PO)
Efficiency Locus
W
W*
PwPO
Politically Optimum
World Price
Range of possible
locations for scenario 2 –
both SSA countries still
retain tariffs
Figure 2.3. Trade negotiations – both countries have tariffs (Ŧ = domestic tariff +1 and
Ŧ* = foreign tariff + 1)
Figure 2.3 helps to describe the impact of a trade agreement. Here again, the
Nash Equilibrium (N) equates to a non-cooperative state. The use of a trade agreement
can help governments to move their tariffs closer to the contract curve (E). This graphic
model demonstrates that the efficiency locus on the contract curve does not necessarily
pass through the point of free trade, which would occur when the tariffs are reciprocal.
W represents the welfare of the domestic government, and W* the foreign government.
Bagwell and Staiger (2002) additionally take the perspective that in moving from
the Nash Equilibrium to the contract curve, the governments are securing additional
market access from their trading partner. By reducing import tariffs, a government is
increasing its partner’s access to that country’s markets, because there is a simultaneous
30
lowering of local prices. If a government does not obtain additional market access from
its partner, then the result would be a terms-of-trade loss for the domestic country.
In applying Bagwell and Staiger’s model to SSA PTAs, there are several points to
consider. The first point is that since this model only applies to trade that passes through
official channels, there may be impact from unrecorded trade. The second point is, given
the relationships among the SSA countries and their colonial backgrounds, it could be
possible that the countries might not start from the Nash Equilibrium point when a trade
agreement is created. If the SSA countries start from another point, it is possible that it
could be a point closer to their politically optimal point. A final point is that if the SSA
PTAs have not had any impact on trade and do not result in reaching a more optimal
tariff point, then what is the impetus for the agreement?
Non-cooperative Nash
Equilibrium
N
E
E
Contract Curve
(Bold Portion)
1
1Ŧ*
Ŧ
R
Reciprocal-Free-Trade
Point
Potential starting point for
agreements with
unrecorded trade between
countries
Figure 2.4. Trade negotiations with Unrecorded Trade (Ŧ = domestic tariff +1 and Ŧ* =
foreign tariff + 1)
31
Regarding the first point, Bagwell and Staiger’s model shows that the politically
optimal point would be the point of reciprocal free trade. Unrecorded Transborder Trade
(UTT) could be as much as or more than the recorded trade. To take a more conservative
perspective, according to Edoho (1997), UTT accounts for approximately 20 to 35
percent of the trade among ECOWAS countries, and it is likely that there is similar UTT
in other regions of SSA. This means that 20 to 35 percent of trade is technically
reciprocal free trade, because there are no tariffs or taxes paid on this unrecorded trade.
Therefore, only 65 to 80 percent of trade must be addressed by a regional trade agreement
to reach a politically optimal point.
This model only addresses two countries, so clearly the picture will be different
for different countries. There are three scenarios at this point: the first is where both
countries have eliminated all tariffs between them, in which case they have reached a
reciprocal free trade state as shown in Figure 2.2. This is the optimal scenario described
in the top left box of the prisoner’s dilemma matrix.
The second scenario is when both countries still have tariffs between them. If
those tariffs have changed over time, then it would be important to assess where those fall
in relation to the contract curve. This scenario might look similar to Figure 2.3 above.
Assuming the countries seem have started with some free (undocumented) trade in place,
they begin in a better position than the Nash equilibrium. At the same time, they have not
negotiated together and their existing tariffs have been arbitrarily applied, so it is unlikely
that they have achieved a politically optimum point described in Figure 2.2, therefore
32
they are probably floating somewhere between the suboptimal Nash point and a point on
the contract curve.
In the third scenario, one country has eliminated tariffs and the other has not.
This is the worst of the scenarios, and is depicted by either of the two boxes in the
Prisoner’s Dilemma matrix where one country defects and the other country cooperates.
Looking at figure 2.4, it is all good for the foreign country, and all bad for the domestic
country. The higher the tariff is, the worse off they are.
N
E
E
1
Ŧ
R
W
W*
PwPO
Scenario where one SSA
country has eliminated
tariffs and the other has
not.
Figure 2.5. SSA trade negotiations – one country has tariffs, the other does not
(Ŧ = domestic tariff +1 and Ŧ* = foreign tariff + 1)
Since these countries have such significant undocumented trade, it is clear that
they did not start from the Nash equilibrium, but must have started at a point that would
be 20 to 35 percent closer to the politically optimal point of reciprocal free trade.
33
Non-cooperative Nash
Equilibrium
N
E
E
Contract Curve
(Bold Portion)
1
1Ŧ*
Ŧ
R
Reciprocal-Free-Trade
Point
R
Potential Starting Point for
SSA trade negotiations
Figure 2.6. Trade negotiations with Unrecorded Trade (Ŧ = domestic tariff +1 and Ŧ* =
foreign tariff + 1)
The third point addresses the fact that if the SSA countries were not trying to
escape the Prisoner’s Dilemma of the Nash equilibrium, why were they interested in a
trade agreement and how has the agreement benefited them? Looking at the background
of the SSA trade agreements, the SSA countries were likely pursuing political as much
as, if not more than, economic objectives with their alliances.. For example, ECOWAS
has been extensively involved in resolving conflicts in West Africa. While this is
considered to be part of their original mission, it is also true that the extent of this effort
required in this area caught the organization by surprised, and ECOWAS has only
recently become more effective at resolving conflict (Chambas 2007). There is an
unstated objective of the ECOWAS agreement to “overcome the established legacies of
colonial inheritance of the West African region’s political Balkanisation and its inferior
34
locus in the world economic system” (Hanink and Owusu, 1998, 365). Essentially, a
purpose of the agreement is to reinstate the ability for interaction based on pre-colonial
ethnic boundaries, which still exist. Thus, ECOWAS allows free movement of persons
across these arbitrary political borders that have impeded interpersonal relationships that
include trade, which should provide the ability for labor to be available where needed.
Examining the impacts of the Bagwell and Staiger model as applied to SSA
agreements identified three primary areas of focus.
1. Impact of unrecorded trade
2. Given the relationships among the ECOWAS countries and their colonial
background, did they really start from the Nash Equilibrium when the trade agreement
was created, or did they start from another point, and if they started from another point,
might it have been a point closer to their politically optimal point?
3. If the SSA trade agreements have not had any impact on trade and did not
result in reaching a more optimal tariff point, then what was the impetus for the
agreement?
Unrecorded trade could be as low as 20 to 35 percent, which means that the
ECOWAS agreement would only need to address 65 to 80 percent of trade. As a result,
the SSA countries could not have started at the NASH equilibrium, and after more than
30 years under various agreements, their current positions vary across the continent
depending on the three possible scenarios described above. Given any two SSA countries
today, they either have no tariffs between them, they both have tariffs, or one country has
tariffs and one does not.
35
Ideally, a regional trade agreement should place all countries in an equal
relationship with each other. Clearly, SSA PTAs have not yet created an equal
environment for all participating countries. While they have objectives to achieve
consistent external tariffs and customs unions, that there continue to be barriers to
achieving these goals. Bagwell and Staiger have proposed that the true reason for a trade
agreement is to escape the Prisoner’s dilemma. The SSA trade agreements have placed
several of its countries exactly in that dilemma.
De Groot et al (2004) analyze international trade data with the incorporation of
institutional quality indicators and determine that the institutional indicator is an
important component of a trade model and should be included in future studies, but do
not specifically examine Africa or the related trade agreements. None of the prior works
that examine trade in SSA countries incorporates an institutional variable as an
explanatory variable.
Economic institutions, property rights, and open economies, are determining
factors of growth (Acemoglu, Johnson, and Robinson 2004) (D. North 1990). Acemoglu
Johnson, and Robinson point out that economic institutions not only help to make the
overall pie bigger, but also determine how that pie is distributed (2004). They define
good economic institutions as those with property rights and access to economic
resources across society.
Baliamoune-Lutz and Ndikumana (2007) state that institutions have a robust and
positive impact on growth, and they also enhance the effectiveness of trade liberalization.
Dollar and Kraay (2003) have a similar finding that good institutions are critical for a
36
country to experience long term growth gains from trade liberalization. North (1994, 3)
defines institutions as “the humanly devised constraints that structure human interaction.
They are made up of formal constraints (e.g., rules, laws, constitutions), informal
constraints (e.g., norms of behavior, conventions, self-imposed codes of conduct), and
their enforcement characteristics. Together they define the incentive structure of societies
and specifically economies.”
A state with effective institutions should be able to implement good governance
practices, and with those in place, their country should experience growth (Baliamoune-
Lutz and Ndikumana 2007). The question at hand is whether trade is impacted by good
governance. In sub-Saharan Africa, many countries have not developed, and, in fact are
worse off than they were at the time of independence from colonization. This paper
examines whether the effects of good governance and regional trade have impacted trade
in African states, with specific focus on the countries that are members of the major
regional trading blocs.
A nation with effective institutions and an open economy should experience
economic growth and development (Baliamoune-Lutz and Ndikumana 2007). In Africa,
effective institutions are lacking, self-serving dictators are prevalent, even in countries
that claim to be democracies, and many protective trade barriers are in place. Between
1980 and the beginning of the twenty-first century, the world experienced increasing
overall growth. African countries, which have also had various protectionist policies in
place, started to implement significant changes in their economic structures in the latter
half of this time period (Arbache, Go, and Page 2008).
37
Acemoglu, Johnson, and Robinson (2004) and North (1990) identify that
economic institutions are the determining factor for whether a country is able to develop
and grow, and they focus particularly on the importance of such institutions as property
rights and open markets. Acemoglu, Johnson and Robinson additionally point out that
economic institutions can both support development and support the policy for how
development is managed or implemented. They define good economic institutions as
“those that provide security of property rights and relatively equal access to economic
resources to a broad cross-section of society” (Acemoglu, Johnson, and Robinson 2004,
9). Diop, Dufrenot, and Sanon (2010) believe poor governance and institutions’ failures
explain the lack of growth in ECOWAS.
Overall, SSA countries face the challenge of an abundance of labor, shortage of
capital and availability of the same raw goods with similar relative factor costs in each
country, which reduces opportunity for trade among these countries. In general, lack of
variability or specialization, means that no countries have a comparative advantage
(Hanink and Owusu 1998).
Foroutan and Pritchett (1993) discuss the issue of unrecorded trans-border trade
(UTT), noting that research from the 1990s has shown that UTT might be equal to or
higher than recorded trade. They also discussed another study that mentioned that the
discrepancy was more than 60 percent. This data contrasts somewhat with the findings of
Edoho (1997) who defined UTT in West Africa as ranging from 20 to 35 percent.
Whatever the number, it is clear that UTT exists and this information is an important
consideration when analyzing trade data for the SSA countries. For the Foroutan and
38
Pritchett study, the fact that trade might be higher than their data showed would only
further validate their finding that trade is higher than expected among SSA countries.
As O’Connell (1997) points out, however, trade policies (including those
governing RTAs) in Africa are often endogenous: they are largely influenced by
macroeconomic variables (e.g., the balance of payments and the exchange rate) and, in
turn, affect countries’ macroeconomic policy positions. Apart from the endogeneity
problem, specification errors in the gravity model could lead to large margins of error in
capturing the residual effect of RTAs. This problem may be compounded by the
inaccuracy of African trade data (1990)
According to the International Monetary Fund (IMF) (Yang and Gupta 2005),
African Regional Trade Agreements (RTAs) have not been effective in promoting trade
or foreign direct investment (FDI). As a result, the IMF recommends significant changes
in the Doha round of World Trade Organization (WTO) which include broad-based trade
liberalization, domestic tax mobilization, infrastructure development for transport among
African countries, reduction of the number of RTAs and focusing them on development,
trade facilitation, investment protection, institution building and regulatory reforms.
Yeats (1990) examines the quality of African trade data primarily from 1962 to
1983. He determines the quality of data to be low though issues on both the export and
import sides may mask the inconsistencies. Particular issues he notes are overvalued
exchange rates, over-invoiced shipments and under-reported imports, time lags in
reporting, differences in classification and valuation procedures, and not specifying the
precise product detail. In addition, the routing country (or intermediate stop) may be
39
documented as the importing country. Except when the data is misreported to the routing
country, which then is over reported, the remaining discrepancies should result simply in
underreporting. Yeats (1998) provides a detailed discussion of the deficiencies of African
trade data, which, as reported to the United Nations COMTRADE system, are often
incomplete, out of date, missing, or even contradictory. On the other hand, data reported
in the IMF’s Direction of Trade Statistics on Africa are up to date but contain data on
aggregate trade only.
Yeats (1994) proposes aggressive reform of African trade regimes in order to
offset the changes that were planned as part of the Uruguay round of the WTO (expected
as a result of removing tariff preferences). Offsetting policies could include removing
quotas and variable import levels on sugar and having African countries eliminate their
own trade barriers, high tariffs. In subsequent analysis, Yeats (1998) notes that the SSA
per capita domestic product decreased annually by 0.5% from 1973 to 1990. The 1998
analysis examines trade complementarity and comparative advantage to assess
opportunity for growth relate to regional trade agreements.
In describing the lack of effective data to review opportunity for growth, Yeats
notes that the most recent reporting years for the SSA countries varies from as early as
1972 up to 1994. Yeats looks at non-oil exports and determines that SSA exports lack
diversity and do not meet local needs which would drive regional trade. SSA countries
have little to trade with each other. This non-complementarity means that trade does not
cross regions. Yeats also notes that there is a scarcity of manufacturing and capital goods
40
along with high tariff barriers. Yeats recommends policy changes related to trade and
structural adjustment. He does not assess the impact of government quality.
According to Landman (2003), the concept of good governance arose within the
late 1980s to early 1990s at the World Bank, with an initial economic focus, but was then
expanded to include a political dimension, (including human rights protection) by the
United Nations Development Programme (UNDP). The World Bank (2000) explains that
good governance has become more important due to demands of increasing competition,
accountability and transparency worldwide. This has posed the challenge of defining
good governance and what sets one country apart from another.
In 2003, the European Commission commissioned an analysis of indicators for
democracy, human rights and good governance. They defined good governance as, “the
transparent and accountable management of all a country’s resources for its equitable and
sustainable economic and social development” (Landman 2003, 2). The concept is that a
good government is one that complies with Adam Smith’s guidance that institutions
enable individuals to pursue interests that drive overall growth (Smith 1776).
The World Bank (Kaufmann, Kraay, and Mastruzzi 2011) developed six
indicators of good governance:
1. Voice and Accountability –political, civil and human rights
2. Political Instability and Violence –likelihood of violent threats to, or changes
in, government, including terrorism
3. Government Effectiveness –competence of the bureaucracy and the quality of
public service delivery
41
4. Regulatory Burden – incidence of market-unfriendly policies
5. Rule of Law – quality of contract enforcement, the police, and the courts, and
likelihood of crime and violence
6. Control of Corruption –the exercise of public power for private gain,
including both petty and grand corruption and state capture
The institutional quality indicator in the analysis is conducted from an average of
the six indicators of good governance. Countries included in the trade bloc analysis are
all SSA countries for which export trade data is available for the selected years. Countries
included in the institutional quality analysis are the same countries utilized in the trade
bloc analysis.
These six indicators are developed from a reduction technique called unobserved
components model which is a variant of factor analysis that extracts components common
to the 300 different measures in the original data (Kaufmann, Kraay and Mastruzzi,
Governance Matters VII: Aggregate and Individual Governance Indicators 1996-2010
2011). There is some criticism of the accuracy of these indicators, which is
acknowledged by the team who developed the indicators. Because their data comes from
so many different sources (the six indicators are a synthesis of 300 data elements), they
do not have complete data for all the countries, and it is not always framed in the same
manner. Landman (2003) says that it is like combining apples and oranges and calling the
result pears. Kaufman (2005) in this explanation of the data explains that this is a
problem, and that he continues to try to improve the data each year. He provides detail on
which data is available from which source and when, and he states that with the large
42
margin of error, the data is not meant to be used to compare individual countries with
each other, especially those countries that fall towards the middle. They firmly believe
that those countries that are high, low and in the middle are where they belong, but the
subtleties of the individual rankings are not so firm.
43
CHAPTER III
METHODS
Gravity Model
Credit is given to both Jan Tinbergen and Pentti Pöyhönen in the early 1960s for
concurrently and independently developing the gravity model of international trade
(Sandberg 2006). Tinbergen (1962) identifies the gravity model in order to examine the
flow of trade between two countries utilizing three primary indicators: the gross national
product (GNP) of the exporting country, the GNP of the importing country, and distance.
Tinbergen further explains the justification for these three indicators, the export GNP
indicates the size of a country’s market (or the amount that can be sold), the import GNP
indicates the amount a country can buy (or the country’s market) and distance is a proxy
for transportation costs.
Sandberg (2006) sums up the work on gravity model theory that has been
conducted over an almost 40 year period and notes that the gravity model can be derived
from both a partial general equilibrium framework and from an international expenditure
system. He says that it has some basis in Heckscher-Ohlin trade theory and theories of
increasing return.
While the Gravity Model had been developed decades previously, Anderson
(1979, 108) was the first to document theoretical foundation for the gravity model. He
says the simplest view of the gravity model is a rearrangement of a Cobb-Douglas
expenditure system with one good for each country, “The gravity equation is a silly
specification from an econometric standpoint since it substitutes out the share (which in
44
the Cobb-Douglas case is the only parameter).” Deardorff (1998), considered the gravity
model with the Heckscher-Ohlin trade theory. Deardorff assumes that bilateral trade
preferences are similar to Cobb-Douglas. He uses two extreme cases of the Heckscher-
Ohlin model, one with frictionless trade and the second with countries that each produce
different goods. He determines that it is not difficult to justify the gravity equation with
standard trade theories such as Heckscher-Ohlin and Cobb-Douglas. He also says that
because the gravity equation characterizes a large class of models, using it for empirical
testing of any of the models is suspect.
Though costs of transportation have declined dramatically over the past 200 years,
costs of trade continue to be important explanatory variables for trade between countries.
Distance does not change over time. Therefore, seeing a declining regression coefficient
of geographic distance over time indicates a decline in the marginal effect of distance
between trading partners (Junius 1999). According to Pomfret (1997), distance is a
strong factor in the size of bilateral trade flows, though at the same time distance is
largely ignored in trade theory.
As the gravity model has evolved, it now includes variables to address effects
beyond size and distance. Four factors that are all considered to impact trade costs and
trade preference are: natural barriers related to physical geography – distance, mountains,
coastal - grouped together as transport costs; trade policy measures, or those with an
environmental focus, as well as exchange rate transaction costs for countries that do not
share the same currency; information costs and cultural differences. All of these factors
restrict trade, but they also simplify the actuality of trade barriers by not accounting for
45
more complex differences such as those from protectionist policies, transportation
infrastructure and insufficient explanation for border effects (Combes, Mayer and Thisse
2008).
Effects of contiguity and common language are usually large and are estimated to
multiply trade by 2 and 2.6 respectively. Contiguity and common-language effects have
increased over time, so that sharing a common language or border has an impact on trade
that is higher now than it was thirty, forty, or fifty years ago (Combes, Mayer, and Thisse
2008).
In the gravity model, relationships are examined between each country and each
of the other countries being examined, so that for the n countries there are (n*(n-1))
relationships. Aitken (1973) describes the gravity model as a cross-sectional trade flow
model. He explains that the gravity model of bilateral trade is inspired by Newton’s
equation of gravity in physics, which relates the gravity force with which two bodies
attract each other proportionately to the product of their masses, and inversely to the
square of their distance. The analogy of this to trade is that countries that are larger (by
GDP and population), and those that are closer to each other, have a larger gravity, and
the greater the gravitational pull or attraction, the greater the trade. The gravity model
analyzes trade between a pair of countries as an increasing function of incomes (size) and
a decreasing function of their distance.
A seminal paper on the use of the gravity model is Aitken’s (1973) analysis of the
integration effects of the EEC and EFTA. Utilizing the gravity model to analyze each
year between 1951 and 1967, Aitken is able to pinpoint when the effects of integration
46
began for each preferential trade arrangement, and is able to identify gross trade creation
within each. His analysis uses the following variables:
Log Xij = log 𝛽0 + 𝛽1 log Dij + 𝛽2 log Yi + 𝛽3 log Yj + 𝛽4 log Ni + 𝛽5 log Nj +
𝛽6 log Aij + 𝛽7 log Pij EEC + 𝛽8 log Pij EFTA + log 𝑒ij
X – country i’s exports to country j measured according to country j’s import data
Dij is distance between commercial centers
Yi and Yj are nominal GNP
Ni and Nj are the populations
Aij is dummy for adjacent
Pij EEC and Pij EFTA are dummies for trade agreements with EEC and EFTA
Sandberg’s 2004 analysis utilizes the gravity model to assess bilateral trade
impacts of the CARICOM agreement in the Caribbean. His model contains the following
variables:
Xijt – value of bilateral exports from country i to country j in period t.
Dij - the physical distance between the two countries.
Yit and Yjt - measures of economic sizes of countries i and j in time t,
respectively, measured by their GDPs.
Nit and Njt - measures of the trading partners’ ‘physical’ sizes, measured by their
respective populations in time.
W captures any resistance or enhancement factors to trade.
47
Sandberg’s analysis found that exporters with larger populations have a larger
productive base, have more opportunities for scale economies, and thus have a greater
ability to export goods to the world market than do smaller countries.
Research Questions
This dissertation studies trade in Africa, particularly as it is impacted by regional
trading blocs over time. It also examines impacts of institutional quality indicators which
have begun to be utilized with the gravity model in trade analysis only in recent years.
The research questions are as follows:
1. Does membership in an SSA trade bloc increase trade in SSA countries?
2. Does membership in an SSA trade bloc increase trade between member
countries of the same bloc?
3. Does institutional quality affect trade in SSA countries?
Hypotheses
H1: Membership in an SSA trade bloc increases trade for SSA countries.
Countries that are members of SADC, COMESA, EAC or ECOWAS
experience higher trade than countries that are not members of one of these trade
blocs.
H2: Membership in an SSA trade bloc increases trade between members of the
trade bloc.
Trade with countries within the trading blocs (member-to-member) is increasing
among SSA member countries more than trade with countries outside of the trading
bloc. Wonnacott and Wonnacott (1981) indicate customs unions, which are components
48
of trading bloc agreements, may benefit member countries more than the benefits of a
unilateral tariff reduction due to better terms of trade. A number of studies have utilized
the gravity model to assess impact of trading blocs on trade, including Sandberg (2006),
which is specific to the CARICOM (Caribbean Community and Common Market)
trading group in the Caribbean, and Foroutan and Pritchett (1993) and Hanink and
Owusu (1998), which both assess sub-Saharan Africa.
Sandberg (2006) shows that membership in the CARICOM trading bloc has a
strong intra-CARICOM effect on trade. Foroutan and Pritchett (1993) find that the low
degree of trade among SSA countries explained by their low GDPs and that intra-SSA
trade is actually higher than expected given their determinants. At the same time, they
only find one trading bloc with significant influence, CEAO, in their 1980-82 analysis.
The intent of Hanink and Owusu (1998) is specifically to determine whether there is
intra-regional trade potential in the least developed countries of Sub-Saharan Africa by
asking whether ECOWAS had promoted trade among its members. They determine from
their study that ECOWAS had not had an impact on trade among its members. They were
looking at trade from 1973 and 1993, and it is possible that drivers of trade have evolved
since then.
Hanink and Owusu (1998) state that customs unions are generally able to create
trade due to decreases in transaction costs that result from either comparative advantage
or increasing returns in differentiated markets. They choose to examine ECOWAS
because relative factor costs are identical due to abundance of labor and shortage of
capital, thus they should have weak opportunity for trade growth based on comparative
49
advantage or differentiated markets. Additional challenges are the uneven sizes of
member countries, high transaction costs, lack of common language or currency. In their
research, they determine that trade has not been much influenced by the ECOWAS
agreement.
Musila (2005) uses the gravity model to assess trade creation and trade diversion
in COMESA, ECCAS and ECOWAS and finds that trade creation is statistically
significant in both COMESA and ECOWAS for the years 1991-1998. Gbetnkom (2006)
utilizes the gravity model to determine that belonging to ECOWAS fosters trade,
positioning this analysis after Structural Adjustment Programme (SAP) liberalization
reforms were enacted.
H3: Institutional quality affects trade in SSA countries.
The average of the World Bank’s Governance Matters indicators of
institutional quality is a significant effect of low trade in Africa.
Variables
The gravity model is the basic tool of analysis within this dissertation. This is the
basic model, and variations are conducted based on the hypothesis and utilizing a single
variable to show participation in any trade bloc versus a variable for each trade bloc, or
utilizing the individual institutional quality indicators versus the average variable.
Log Xij = β0 + β1 logYi + β2 logYj + β3 logNi + β4 logNj + β5 logDij + β6 logAij +
β7 logLOij + β8 logLEij + β9 logEDi + β10 logEDj + β11 logPTAi + β12 logPTAij + β13
logGOVAi + β14 logGOVAj + log 𝑒ij
50
Yi and Yj and Ni and Nj are GDP and population of the exporting and importing
countries respectively. Dij is distance between commercial centers of the exporting and
importing countries. Aij is a dummy for adjacent borders between the exporting and
importing countries. LOij is a dummy for common official language between the
exporting and importing countries. LEij is a dummy for common ethnic language. PTAi is
a dummy for when exporting country is a member of an SSA PTA. PTAij is a dummy for
when both exporter and importer are members of an SSA PTA. GOVAi is the average
institutional indicator for the exporting country. GOVAj is the average institutional
indicator for the importing country. The last term is the stochastic error term, which
captures all other (omitted) effects on trade and is assumed to be well-behaved.
Dependent Variable – Bilateral Exports
The dependent variable Xij for this analysis is bilateral exports from each
exporting country i to each importing country j. Export data between countries was
collected from the International Monetary Fund’s (IMF) Direction of Trade Statistics
(DOT) database (International Monetary Fund 2014).
51
Table 3.1
Total Exports to Sub-Saharan Africa and to World by Country, 2001-2005
Country
Exports to Sub-
Saharan Africa
% Exports
to SSA
Exports to
World
Angola
125,258,363
1.16%
10,757,638,241
Benin
141,058,034
41.59%
339,162,506
Burkina Faso
147,888,686
56.31%
262,643,713
Burundi
6,050,201
13.98%
43,268,747
Cameroon
240,850,824
9.48%
2,540,620,470
Cape Verde
170,089
1.30%
13,081,290
Central African Republic
12,965,874
8.70%
148,988,753
Chad
4,140,603
0.63%
657,418,561
Comoros
414,211
1.28%
32,484,781
Congo, DR
71,171,067
5.98%
1,189,978,048
Congo, Republic of
55,669,956
2.38%
2,336,845,189
Cote d'Ivoire
1,541,956,138
27.39%
5,629,020,863
Djibouti
55,535,343
22.68%
244,834,406
Equatorial Guinea
9,050,430
0.28%
3,214,570,171
Ethiopia
1,498,219
0.30%
505,503,605
Gabon
108,740,886
2.75%
3,948,096,816
Gambia, The
2,658,278
9.29%
28,613,460
Ghana
168,524,266
8.54%
1,973,686,360
Guinea
39,641,104
5.11%
775,778,956
Guinea-Bissau
11,502,381
12.42%
92,583,405
Kenya
952,465,217
37.42%
2,545,441,823
Liberia
11,140,999
1.12%
998,014,094
Madagascar
49,791,666
5.68%
877,275,886
Malawi
97,232,566
19.84%
489,978,922
Mali
9,891,079
4.52%
218,752,216
Mauritania
111,799,816
16.17%
691,431,942
Mauritius
161,700,045
8.97%
1,802,905,201
Mozambique
254,790,917
22.43%
1,135,749,267
Niger
75,558,397
35.97%
210,077,376
Nigeria
2,359,425,023
8.58%
27,512,594,467
Rwanda
5,580,924
2.84%
196,383,234
Sao Tome and Principe
335,184
2.83%
11,848,768
Senegal
377,055,765
33.60%
1,122,334,740
Seychelles
10,629,490
3.99%
266,387,832
Sierra Leone
7,659,526
5.62%
136,300,368
52
Table 3.1 (continued).
Country
Exports to Sub-
Saharan Africa
% Exports
to SSA
Exports to World
Somalia
8,190,158
5.30%
154,595,501
South Africa
5,004,907,349
12.97%
38,578,972,990
Sudan
28,653,560
0.97%
2,967,352,818
Tanzania
169,208,484
15.71%
1,077,214,954
Togo
206,723,825
62.32%
331,732,961
Uganda
175,173,884
30.00%
583,819,185
Zambia
515,987,939
41.50%
1,243,379,599
Zimbabwe
681,904,005
31.71%
2,150,167,151
Sub-Saharan Africa
13,820,949,587
11.92%
115,995,712,196
World
104,779,724,134
1.32%
7,925,594,516,763
Independent Variables
The independent variables for this model are: the distance between the major
cities of the exporting and importing countries; the GDP and the population of the
exporting and importing countries; and the following possible resistant or enhancement
factors to trade: dummy variables for whether or not the countries share a common
border, whether or not the countries have a common official or ethnic language, whether
or not the exporting or both countries are SADC, COMESA, EAC, ECCAS or ECOWAS
members; and quality of institutions of the exporting and importing country.
1. Distance between countries (Dij)
Transaction costs in the gravity model are represented by distance between the
two countries, dist, calculated as the distance between the major cities of the two
countries, as the crow flies, in kilometers, The shorter the distance between trading
partners, the lower the resistance or impediments to trade, which would be associated
53
with higher levels of trade, and vice versa (Gbetnkom 2006). Distance is expected to
increase transport costs and impede the flow of trade across countries (M. A. Babatunde
2006). Data on distance between all country pairs were obtained from the Centre
d’Etudes Prospectives et d’Informations Internationales (CEPII) (2014)which has a
downloadable file that contains these variables and relationships between all country
pairs.
2. GDP (Yi /Yj) and population (Ni / Nj) of both the exporting and importing
countries.
GDP and population represent the economic mass of the two countries (Burger,
van Oort and Linders 2009). Higher income level indicates greater potential supply from
the exporting country and increased demand of the importing country, which would lead
to a positive effect on trade. If the population is large, it would be expected to increase
the ratio of domestic to foreign market production, and thus allow greater output
diversification. With greater output diversification, there would be lower potential
demand in the importing country and lower potential supply from the exporting country,
and an overall decrease in imports. (Sandberg 2006). In using GDP as a variable,
Gbentkom (2006) says a high level of income indicates a high level of production for the
exporter and increases availability of export goods. In this instance, GDP is generally
positively related to trade, though there are scenarios when larger countries trade less
because they have resources available domestically.
Anderson (1979) stresses the importance of using both GDP and population by
stating that GDP and population together determine the potential export supply of a
54
country – GDP determines economic capacity and population determines domestic
market/foreign market production ratio. It has also been argued (Bergstrand 1985) that
GDP per capita can be used, as a proxy for the capital-labor ratio of the exporting country
and a demand factor for the importing country. GDP and population were accessed from
the World Bank World Development Indicators (2014) in current U.S. dollars.
Table 3.2
GDP for Selected Countries, 1980, 1990, 2000, 2010 with percent growth 1980 to 2010
Country
1980
1990
2000
2010
Growth
1980-2010
Algeria
42,345
62,045
54,790
161,207
281%
Angola
10,000
9,130
82,471
Armenia
2,257
1,912
9,260
Australia
149,679
310,945
414,987
1,141,268
662%
Bangladesh
18,115
31,219
53,370
115,279
536%
Belize
195
413
832
1,397
617%
Benin
1,405
1,960
2,359
6,562
367%
Brazil
461,952
657,216
2,209,400
Burkina Faso
1,929
3,101
2,633
8,980
366%
Burundi
920
1,132
870
2,027
120%
Cabo Verde
142
307
539
1,664
1070%
Cambodia
3,654
11,242
Cameroon
6,741
11,152
9,287
23,622
250%
Canada
273,437
592,015
739,456
1,614,014
490%
Central African
Republic
797
1,441
915
1,986
149%
Chad
1,033
1,739
1,385
10,658
932%
China, Mainland
189,650
358,973
1,205,261
6,039,659
3085%
Comoros
124
250
202
530
330%
Congo, DR
19,088
20,523
Congo, Republic of
1,706
2,799
3,220
12,008
604%
Cote d'Ivoire
10,176
10,796
10,717
24,885
145%
Djibouti
452
551
1,129
Egypt
22,912
43,130
99,839
218,888
855%
Equatorial Guinea
51
112
1,046
11,564
22735%
55
Table 3.2 (continued).
Country
1980
1990
2000
2010
Growth
1980-2010
Ethiopia
12,175
8,242
29,934
Gabon
4,280
5,952
5,068
14,476
238%
Gambia, The
241
317
783
952
295%
Ghana
4,445
5,889
4,983
32,175
624%
Guinea
2,667
2,995
4,736
Guinea-Bissau
111
244
370
847
666%
Kenya
7,265
8,572
12,705
40,000
451%
Lebanon
2,838
17,260
38,010
Liberia
855
384
529
1,293
51%
Libya
28,905
38,271
74,773
Madagascar
4,042
3,081
3,878
8,730
116%
Malawi
1,238
1,881
1,744
5,399
336%
Mali
1,787
2,421
2,422
9,422
427%
Mauritania
709
1,020
1,294
4,338
512%
Mauritius
1,137
2,653
4,583
9,718
755%
Mexico
194,357
262,710
683,648
1,051,129
441%
Morocco
21,079
28,839
37,021
90,770
331%
Mozambique
3,526
2,512
4,310
10,119
187%
Nicaragua
2,144
1,009
5,107
8,741
308%
Niger
2,509
2,481
1,798
5,719
128%
Nigeria
64,202
30,757
46,386
369,062
475%
Rwanda
1,255
2,550
1,735
5,699
354%
Sao Tome and
Principe
77
201
Saudi Arabia
164,542
116,778
188,442
526,811
220%
Senegal
3,503
5,717
4,680
12,932
269%
Seychelles
147
369
615
970
558%
Sierra Leone
1,101
650
636
2,578
134%
Somalia
604
917
-100%
South Africa
80,547
112,015
136,362
375,349
366%
Sudan
7,617
12,409
12,257
65,634
762%
Tanzania
4,259
10,186
31,105
Thailand
32,354
85,343
122,725
318,908
886%
Togo
1,136
1,628
1,294
3,173
179%
Uganda
1,245
4,304
6,193
18,804
1411%
United Kingdom
564,948
1,093,169
1,548,664
2,407,857
326%
United States
2,862,510
5,979,580
10,284,779
14,964,372
423%
56
Table 3.2 (continued).
Country
1980
1990
2000
2010
Growth
1980-2010
Zambia
3,885
3,288
3,601
20,266
422%
Zimbabwe
6,679
8,784
6,690
9,457
42%
Source: World Development Indicators (WDI). Data extracted from WDI Database on 7/19/15
3. Resistant or Enhancement Factors to trade
a) Shared border and language and colonizer (Aij, LOij, LEij)
Variables which capture other common impacts on trade are captured by the
following: Aij, adjacent, a dummy variable for whether or not the countries share a
common border, LOij, a dummy variable for whether the countries share a common
official language, LEij, a dummy variable for whether the countries share a common
ethnic language. The dummy variables regarding shared language and shared border were
obtained from the Centre d’Etudes Prospectives et d’Informations Internationales (CEPII)
(2014).
b) Membership in a preferential trading agreement (PTAi, PTAij,)
PTAi is a dummy variable for whether or not the exporter is a member of any of
the selected African trading agreements, which include the Economic Community of
West African States (ECOWAS), Common Market for Eastern and Southern Africa
(COMESA), Southern African Development Community (SADC), Economic
Community of Central Africa States (ECCAS), and East African Community (EAC).
PTAij is a dummy variable for whether both countries are members of an SSA trading
agreement. These variables are assumed to stimulate trade (Gbetnkom 2006). PTAi has a
57
value of one if the exporter belongs to a regional integration organization and zero if the
exporter does not belong. PTAij is zero unless both countries belong to an SSA trade
organization, in which case the value is one. Therefore, the second variable assesses trade
creation within the community, and the first assesses impact on trade outside of the
community, or trade diversion.
c) Good Governance
GOVAi and GOVAj are variables for the average of the six good governance
indicators for the quality of institutions of the exporting and importing country. Data on
good governance was taken from Good Governance Matters IV (Kraay, Kaufman, and
Mastruzzi 2005). The institutional quality indicator is an average of the six indicators of
good governance developed by the World Bank (Kraay, Kaufman, and Mastruzzi 2005).
1. Voice and Accountability – measuring political, civil and human rights
2. Political Instability and Violence – measuring the likelihood of violent threats
to, or changes in, government, including terrorism
3. Government Effectiveness – measuring the competence of the bureaucracy and
the quality of public service delivery
4. Regulatory Burden – measuring the incidence of market-unfriendly policies
5. Rule of Law – measuring the quality of contract enforcement, the police, and
the courts, as well as the likelihood of crime and violence
6. Control of Corruption – measuring the exercise of public power for private
gain, including both petty and grand corruption and state capture
58
These six indicators are made up of components from 300 measures. Kaufman
(2005) provides detail on which data is available from which source and when, and he
states that with the large margin of error, the data is not meant to be used to compare
individual countries with each other, especially those countries that fall towards the
middle. He believes those countries that are high, low and in the middle are where they
belong, but the subtleties of the individual rankings are not as firm.
Table 3.3
Descriptive statistics 2010
Variable
Obs
Mean
Std. Dev.
Min
Max
Exports (M)
3401
708.97
9,977.41
0.00
290,054.00
Aij
3401
0.05
0.22
0.00
1.00
Loij
3401
0.36
0.48
0.00
1.00
Leij
3401
0.30
0.46
0.00
1.00
Dij
3401
5,938.57
3,965.94
10.48
18,008.29
Yi (M)
3401
577,498.10
2,152,542.00
0.00
15,000,000.00
Ni (M)
3401
58.89
181.41
0.09
1,337.71
EDi
2000
60.54
30.60
13.83
130.84
Yj (M)
3401
577,869.70
2,152,526.00
0.00
15,000,000.00
Nj (M)
3401
58.83
181.41
0.09
1,337.71
Edj
2017
60.39
30.58
13.83
130.84
GOVAi
3401
-0.49
0.73
-2.11
1.61
GOVAj
3401
-0.49
0.73
-2.11
1.61
PTAi
3401
0.70
0.46
0
1
PTAij
3401
0.70
0.46
0
1
Participants
Participants for the purposes of this paper will be countries. Countries included in
the trade bloc analysis are 42 SSA countries for which export trade data is available for
59
the selected years, and 20 non-SSA countries. The dataset excludes cases where no
bilateral trade is reported or values are missing so that the final sample includes 3,401
observations for the 62 SSA and non-SSA countries. With 62 countries, there are a
possible 3,782 observations, but the DOT data (2014) frequently returns no values when
trade is not reported, so the total observations are 3,401. The time period is 2010 with
comparison to data from 1980, 1990, and 2000.
Table 3.4
Countries included in the analysis
SSA Countries
1. Angola
2. Benin
3. Burkina Faso
4. Burundi
5. Cameroon
6. Cape Verde
7. Central African
Republic
8. Chad
9. Comoros
10. Congo
(Brazzaville)
11. Congo (Democratic
Republic)
12. Côte d'Ivoire
13. Djibouti
14. Equatorial
Guinea
15. Ethiopia
16. Gabon
17. The Gambia
18. Ghana
19. Guinea
20. Guinea-
Bissau
21. Kenya
22. Liberia
23. Madagascar
24. Malawi
25. Mali
26. Mauritania
27. Mozambique
28. Niger
29. Nigeria
30. Rwanda
31. Sao Tome
and Principe
32. Senegal
33. Seychelles
34. Sierra Leone
35. Somalia
36. South Africa
37. Sudan
38. Tanzania
39. Togo
40. Uganda
41. Zambia
42. Zimbabwe
Additional Partner Countries
1. Algeria
2. Armenia
3. Australia
4. Bangladesh
5. Belize
6. Brazil
7. Cambodia
8. Canada
9. China
10. Egypt
11. Libya
12. Lebanon
13. Mauritius
14. Mexico
15. Morocco
16. Nicaragua
17. Saudi Arabia
18. Thailand
19. United Kingdom
20. United States
60
The data cover the period from 1980 to 2010 in ten year increments. Each country
pair has two observations, one with each country as exporter. Bilateral Data includes:
dependent variable, Exports (Xij), Distance (Dij), Common border (Aij), common
language (LOij and LEij), and preferential trading agreement of both trading partners
(PTAij). Country-specific data includes: GDP, Population, preferential trading agreement
of the exporter only (PTAi), and Government Indicator average (GOVAi and GOVAj).
Methodology
This dissertation utilizes regression analysis to assess the impacts of regional
trading blocs and institutional quality on trade. The model uses OLS and Tobit regression
in ten year increments, beginning with the most recent data in 2010.
Limitations
There is some debate over the appropriate analytical method for analyzing trade
data with the gravity model. There are zeroes in the export data, sometimes when there
truly is trade but it is unreported. The zeroes result in challenges in analyzing the data.
Zeroes could be discarded, but might compromise the economic analysis since it may be
important that there was no data for these particular entries. Hanink and Owusu (1998)
and others addressed this by using Tobit regression, which treats zero or unrecorded trade
flows as unobserved data points (Gbetnkom 2006). Tobit has its ownchallenges because
certain elements of analysis, such as the R-squared, are missed. Sandberg (2006), on the
other hand, chose to address this by adding ‘1’ to all values. His dataset contained 247
zero observations out of 3,454 total observations. In adding 1, he defines exports = X*ijt =
(Xijt + 1) which makes empirical observation possible. The perception is that any
61
distortion is insignificant, because the variation in data is in the thousands, so adding one
to each number is not going to impact the results but not including the one is likely to
bias the results.
While some propose that it is necessary to utilize Tobit regression for gravity
model analysis where the data contains many zero values, Subramanian and Tamirisa
(2003) suggest that the approach of using Tobit regression confers unduly large weights
on the adjusted zero-value observations. For example, Cape Verde, which had no trade at
all, should be reflected as zero. Sierra Leone and Liberia, which were undergoing
extenuating circumstances of physical conflict during the trade year selected (1993),
could potentially have been removed from the sample altogether, and another option
would be to conduct at least two separate analyses, with and without those countries.
In addition, another limitation would be that this analysis is reflective of officially
reported trade and would not include any unrecorded trans-border trade (UTT). The
assumption is that the official data represent relative values of trade accurately even if
they do not provide the complete picture. Thus, they would provide relative values within
and across SSA countries, where it can be assumed that levels of UTT are fairly similar
across countries. Differences would only be more marked increases in export values if the
UTT could be estimated. As noted in Chapter II, the UTT serves to move the SSA
countries away from the undesired Nash equilibrium and closer to a mutually beneficial
contract point.
62
CHAPTER IV
ANALYSIS
The gravity model for this paper states that the trade between two countries is a
function of the GDP and population of both the exporting and importing countries, the
distance between the major cities of the two countries, existence of shared border,
commonality of official or ethnic language, membership in different sub-Saharan Africa
(SSA) trade agreements such as ECOWAS, COMESA, EAC, SADC or ECCAS, and the
quality of institutions of the exporting and importing countries. Each country pair is
reflected with two observations, one with each trading partner as exporter. Bilateral data
includes: Distance, Common border, common language, common colonizer, and
Preferential Trading Agreement (PTA). Country-specific data includes: Exports, GDP,
population, participation in preferential trade agreement by the exporter or both exporter
and importer, and average governance indicator for the six individual governance
indicators for the exporter and the importer.
The gravity model is a standard model used to analyze international trade among
countries. The gravity model is intended to assess barriers to trade such as the transaction
cost of trade. In order to understand barriers to trade and their impacts, it is necessary to
understand all of the costs involving trade over geographical space. Thus, the theory
should be inclusive of all costs, including costs associated with lack of face-to-face
contact, more complex, expensive communications, information gathering, and different
languages, legal systems, product standards, and cultures. Fujita, Krugman, and Venables
(1999) believe that these costs of trade are evident in the trade data. The gravity model
63
begins from the information that trade decreases with distance, and this implies that costs
are higher with greater distance. Distance becomes a proxy for transport costs, transport
time (addressing perishability, adaptability to market conditions, and irregularities in
supply), and psychic distance (addressing familiarity with laws, institutions, and habits)
(Junius 1999). Junius states that the closer two countries are to each other, the greater are
the regional cultural similarities that make trade easier.
A number of studies have utilized the gravity model to assess the impact of
trading blocs on trade, including Sandberg (2006), who examines CARICOM (Caribbean
Community and Common Market) trading group in the Caribbean, Foroutan and Pritchett
(1993), and Hanink and Owusu (1998), both assess SSA countries. Sandberg (2006)
shows that membership in the CARICOM trading bloc has a strong intra-CARICOM
effect on trade. Foroutan and Pritchett (1993) find that the low degree of trade among
SSA countries is explained by their low GDPs and that intra-SSA trade is actually higher
than expected given their determinants. At the same time, they only find one trading bloc
with significant influence, CEAO (Communaute Economique de l’Afrique de l’Ouest) a
West African monetary community which no longer exists and was succeeded in 1994 by
UEOMA (Union Economique et Monetaire Ouest Africaine). In their 1980-82 analysis.
Hanink and Owusu (1998) state that customs unions are generally able to create trade due
to decreases in transaction costs and tariffs that result from either comparative advantage
or increasing returns in differentiated markets. Musila (2005) uses the gravity model to
assess trade creation and trade diversion in COMESA, ECCAS and ECOWAS and finds
that trade creation is statistically significant in both COMESA and ECOWAS for the
64
years 1991-1998. Gbetnkom (2006) utilizes the gravity model to determine that
belonging to ECOWAS, after Structural Adjustment Programme (SAP) liberalization
reforms were enacted, fosters trade. The gravity model is utilized to test the hypotheses in
this dissertation.
H1: Membership in an SSA trade bloc increases trade for SSA countries.
Countries that are members of active trading blocs in Sub-Saharan Africa (SADC,
COMESA, EAC or ECOWAS) experience more trade than countries that are not
members of one of these trade blocs. As stated by the Centre for Democracy and
Development (2002) a regional trade agreement, or ‘regionalism’, is defined by
geographically proximal countries with common objectives for economic or political co-
operation among themselves to improve their standing.
Between 1995 and 2005, there have been changes in African economics.
According to Arbache, Page, and Go (2008), SSA leaders are taking increasing control of
their economic destiny. After 45 years of stagnation, improvements are being made but
the authors do not believe these changes are resulting in significant improvements in
economic development. H1 examines whether these changes may be driving increased
trade by examining the overall impacts of regional trading with countries outside the
trading bloc. This paper focuses on countries and active trading blocs in sub-Saharan
Africa.
H2: Membership in an SSA trade bloc increases trade between members of the trade
bloc.
65
Trade among countries within the trading blocs (member-to-member) is
increasing among SSA member countries more than trade between countries within the
trading bloc and countries outside of the trading bloc. This would indicate that
membership in the trade bloc creates trade, which is an intent for a trade bloc.
This analysis focuses on whether the exporter is a member of a selected African
trading agreement (including Economic Community of West African States (ECOWAS),
Common Market for Eastern and Southern Africa (COMESA), Southern African
Development Community (SADC), East African Community (EAC) or Economic
Community of Central Africa States (ECCAS) ), or both countries are members of the
same trading agreement. Membership in a trading bloc is intended to stimulate trade
(Gbetnkom 2006). Utilizing membership in a trading bloc for analysis is considered a
method to assess trade creation and trade diversion. Trade creation is a net increase in
trade where more expensive domestic goods are now purchased at a lower price via the
terms of a trade agreement (generally due to the reduction in tariffs).
When the preferential trade agreement variable in which both trading partners are
members of an SSA trade agreement (PTAij) is positive, it reflects trade creation. Trade
creation is when there is more trade between trading partners than predicted by the other
variables in the model, and where the cost of the good is now lower because of the tariff
reduction. Trade diversion, on the other hand, is when trade moves from a lower cost
partner who is not a member of the trading bloc to a higher cost partner who is a member
of the trading bloc. When PTAi is negative, then that shows that trade is diverted from a
66
lower cost external partner to a lower cost member partner, thus reflecting trade
diversion. (Kanda and Jordaan 2010).
H3: Institutional quality affects trade in SSA countries. The average of the World
Bank’s Governance Matters indicators of institutional quality has a significant
effect on trade in Africa.
A nation with effective institutions and an open economy should experience
economic growth and development (Baliamoune-Lutz and Ndikumana 2007). Acemoglu,
Johnson, and Robinson (2004) and North (1990) identify that economic institutions are
the determining factor for whether a country is able to develop and grow, and they focus
particularly on the importance of such institutions as property rights and open markets.
Acemoglu, Johnson, and Robinson (2004) point out that economic institutions are
important for growth, and can guide how that growth is managed. They define good
economic institutions as ‘those that provide security of property rights and relatively
equal access to economic resources to a broad cross-section of society’. Diop, Dufrenot,
and Sanon (2010) state that failures of governance and institutions can explain lack of
growth in the ECOWAS (West African) countries.
Effective governance is analyzed utilizing the gravity model with the Good
Governance Matters indicators developed by the World Bank beginning in 2005 (Kraay,
Kaufman, and Mastruzzi 2005). De Groot et al (2004) find that institutional quality has a
significant, positive and substantial impact on bilateral trade flows. De Groot also
determines that countries with more similar institutions are more likely to trade with each
other, so that institutions with lower governance quality might do more trade with each
67
other, and lower quality would drive increased trade in that instance. Lavallee (2005)
utilizes data from 21 OECD and 95 developing countries to analyze the effects of
governance on trade, also including membership in a free trade agreement. Lavallee
utilizes this analysis to demonstrate that poor governance reduces the amount of trade
that a country has with an industrialized nation.
Data on good governance is obtained from Good Governance Matters IV (Kraay,
Kaufman, and Mastruzzi 2005). The institutional quality indicator is the average of the
six indicators of good governance developed by the World Bank (Kraay, Kaufman, and
Mastruzzi 2005): Voice and Accountability – political, civil and human rights; Political
Instability and Violence – likelihood of violent threats to, or changes in, government;
Government Effectiveness – competence of the bureaucracy and quality of public service;
Regulatory Burden –market-unfriendly policies; Rule of Law – quality of contract
enforcement, police, and courts; Control of Corruption – exercise of public power for
private gain. These six indicators are developed from extracts of components common to
300 different measures (Kaufmann, Kraay, and Mastruzzi 2011). Kray, Kaufman, and
Mastruzzi (2005) provide detail on the data sources and timing. They state that with the
margin of error, the data is not meant to be used to compare individual countries with
each other, especially those countries that fall towards the middle. The overall structure
showing top, middle and bottom countries, is valid, but that comparing two countries
ranked next to each other, particularly in the middle, is not refined to that degree.
This analysis includes 42 SSA countries and 20 non-SSA countries. Exports are
used as an indicator of quantity of trade. In the gravity model, those exports are utilized
68
bilaterally – from each country in the analysis to each other country in the analysis. .
Export data between countries are from the International Monetary Fund’s (IMF’s)
Direction of Trade Statistics (DOT) database.
This paper utilizes regression analysis to assess the impacts of regional trading
blocs and institutional quality on trade. The model conducts OLS regression on the
exports from 1980 to 2010 in 10 year increments, and follows with comparison to a Tobit
analysis. Data in 1980 and 1990 only apply to trade agreement data, as the good
governance dataset begins in the mid-1990s. Export data have numerous instances of
zero trades between two individual countries. Some of these may be truly zero trade
between countries, but others may reflect either poor reporting or specific instances of
turmoil in a given country, such as civil war which may only last for one to two years.
Several authors have recommended Tobit analysis, which replaces the zeroes with a
small number (i.e., 0.1) to address the lack of trade values (Foroutan and Pritchett 1993)
(Hanink and Owusu 1998).
With 62 countries (SSA and non-SSA), there are a possible 3,782 (n*(n-1))
observations because there are 62 countries each exporting to 61 partner countries), but
the DOT data frequently returns no values when trade is not reported, so the total
observations extracted from the DOT database for exports between these countries are
3,402. The data covers 2010, with comparison to data from 1980 to 2000 in ten year
increments. The data include the countries listed in table 3.4. Selected countries include
SSA countries and countries that were included in similar studies that included non-SSA
countries by Kanda and Jordaan (2010), and Subramanian and Tamirisa (2003).
69
During the 1980 to 2010 time period of this study, world growth (GDP, current
US$) increased 487%, growth in the United States was 422%, and growth across all SSA
countries was 396%. Growth across the countries in the study ranged from a low of 42%
(Zimbabwe) to an extreme high of 22,735% (Equatorial Guinea) and the next highest of
1,411% (Uganda) (World Bank 1980-2010). Several countries were not included in the
growth analysis because they had no data reported in either 1980 (Angola, Democratic
Republic of Congo, Djibouti, Ethiopia, Guinea, Sao Tome and Principe and Tanzania) or
2010 (Somalia). When looking at average growth in just the developing countries in SSA,
the growth rate was only slightly lower than the total SSA growth rate, at 392%, which is
an indicator that the lowest income countries are keeping pace with the higher income
SSA countries. In the most recent ten years of the study, from 2000 to 2010, the SSA
growth was 268% and the developing countries only two percent lower at 266%. This
shows that two-thirds of the growth in that 30 year period was in the most recent ten year
period.
70
Chart 4.1. Average GDP vs Growth, SSA Countries, 2001 to 2005 (removed outlier
Equatorial Guinea with 7000% growth)
In Chart 4.1, countries are clearly clustered in the lower left hand corner of the
chart. There is an amount of variation in the 100 to 300 percent growth range. Then there
are a few anomalies that exceeded 400% growth.
The SSA countries with the highest current GDP are South Africa (375B) and
Nigeria (369B), and those with the lowest GDP are Sao Tome and Principe (201M) and
Comoros (530M). When population is incorporated to assess growth in GDP per capita,
the highest growth is in Equatorial Guinea (7,136%) and Cape Verde (624%). The lowest
GDP growth rates are negative, including Liberia (-28%), Zimbabwe (-21%) and Niger (-
16%). In the last 10 years examined, 2000 to 2010, the highest growth was in Angola
(543%) and Equatorial Guinea (723%), and the lowest growth was in the Gambia (-11%)
and the Democratic Republic of the Congo (-19%). Of note, all other growth rates from
71
2000 to 2010 were positive, whereas in the longer period from 1980 to 2010 there were
five countries with negative growth rates.
Table 4.1
Lowest and highest growth SSA countries, 1980 to 2010
Lowest Growth
Growth 1980-2010
Trade Agreement
Somalia
-100%
none
Zimbabwe
42%
COMESA & SADC
Liberia
51%
ECOWAS
Madagascar
116%
COMESA
Burundi
120%
COMESA, EAC, ECCAS
Highest Growth
Over 200%
Trade Agreement
Sudan
762%
COMESA
Chad
932%
ECCAS
Cape Verde
1070%
ECOWAS
Uganda
1411%
COMESA
Equatorial Guinea
22735%
ECCAS
Arbache, Go, and Page (2008) identify an acceleration of growth in African
countries that begins after the mid-1990s. Arbache, Go, and Page examine whether
increase in African growth is an indicator of reversal of decades of development and
growth failures. The analysis determines that sustainability is precarious due to stagnant
savings, investment, productivity, and export diversification. Arbache, Go, and Page also
note that growth is seen in almost every category of country that he examines – oil-
importing and oil-exporting, coastal and landlocked, with some growth in what he
describes as fragile countries as well. The fragile countries, those without significant oil
resources, showed the most growth between 2004 and 2006.
72
Chart 4.2. Average GDP per person and percent growth from 1980 to 2010 (removed
outlier Equatorial Guinea with 7000% growth)
In Chart 4.2 it is possible to see that growth is happening everywhere in SSA
countries and at different rates. With the countries arranged in order of average GDP PP,
the greatest growth is seen almost at the far left, but then there are spikes of growth along
almost the complete length of the chart.
African countries are benefitting from higher oil prices, rising non-oil commodity
prices, increased foreign aid, leadership reforms, trade improvements and tariff
reductions, and increased school enrollments and literacy rates (Arbache, Go, and Page
2008). Arbache, Go, and Page summarize that in order to sustain the growth, African
countries will need more exports, private sector growth, productivity, foreign investment
and remittances, and more. They claim these goals will be achieved with reform and
improvements in governance.
73
The gravity model equation for this analysis is as follows:
Log Xij = β0 + β1 logYi + β2 logYj + β3 logNi + β4 logNj+ β5 logDij + β6 logAij + β7
logLOij + β8 logLEij + β9 logEDi + β10 logEDj + β11 logPTAi + β12 logPTAij + β13
logGOVAi + β14 logGOVAj + log 𝑒ij
The dependent variable is represented by Xij, which represents total exports from
exporting country (i) to importing country (j).
Yi and Yj are GDP of the exporting and importing countries respectively.
Ni and Nj are Population of the exporting and importing countries respectively.
Dij is distance between capitals of the exporting and importing countries.
Aij is the dummy for adjacent borders between the exporting and importing
countries.
LOij is a dummy for common ofnficial language between the exporting and
importing countries.
LEij is a dummy for common ethnic language between the exporting and
importing countries.
EDi is secondary education level of exporting country
EDj is secondary education level of importing country
PTAi is a dummy for exporting country member of an SSA preferential trade
agreement (PTA).
PTAij is a dummy for both countries members of an SSA preferential trade
agreement (PTA).
GOVAi is the average institutional indicator for the exporting country.
GOVAj is the average institutional indicator for the importing country.
𝑒ij - the final term is the stochastic error term, which captures all other (omitted)
effects on trade and is assumed to be well-behaved.
Expected signs of the regression output are as follows:
+ +/- + +/- - + + + +/- +/- +/- +/-
Xij = F ( Yi, Ni, Yj, Nj, Dij, Aij, LOij, LEij, PTAi,PTAij, GOVAi, GOVAj )
GDP is expected to be positive, since larger economies would have more volume
and diversity of opportunities for trade. Population could be either positive or negative
because smaller countries may need to trade more because of lack of access to products,
or may trade less because of lack of resources. Some large countries may trade less
because they are able to generate their own goods, or may drive more trade because of
74
their resources. Aij (common border) is expected to be positive, as two countries that
share a border should have more opportunity for trade. Dij (Distance between two
countries) is expected to be negative because the greater the distance between two
countries, the higher the cost of trade and the lower the opportunity to trade. Common
language, official or ethnic, would be expected to be positive, because language
differences can be a barrier to trade. The direction of this paper and the results of prior
studies would indicate that the impact of the preferential tradse agreement as well as
quality of governance would be positive, but it is the objective of this analysis to assess
impacts of these variables.
The gravity model analysis evaluates the level of trade among each set of trading
partners. Results from this analysis begin with Table 4.3. Hypotheses are repeated for
reference and analysis below relates to the listed hypotheses.
Test of H1: Membership in an SSA trade bloc increases trade for SSA countries
and H2: Membership of the exporting country in an SSA trade bloc increases trade for
the trade bloc members.
Regression analyses were conducted to examine the relationship between exports
and various potential predictors. The tables below summarize the descriptive statistics
and analysis results. The variable PTAi is tested here, which evaluates effects on trade
when the exporter is a member of an SSA trading bloc.
75
Table 4.2
Descriptive Statistics of the Dependent and Independent Variables for Trade Analysis,
2010
Variable
Obs
Mean
Std. Dev.
Min
Max
Exports (M)
3401
708.97
9,977.41
0.00
290,054.00
Aij
3401
0.05
0.22
0.00
1.00
Loij
3401
0.36
0.48
0.00
1.00
Leij
3401
0.30
0.46
0.00
1.00
Dij
3401
5,938.57
3,965.94
10.48
18,008.29
Yi (M)
3401
577,498.10
2,152,542.00
0.00
15,000,000.00
Ni (M)
3401
58.89
181.41
0.09
1,337.71
EDi
2000
60.54
30.60
13.83
130.84
Yj (M)
3401
577,869.70
2,152,526.00
0.00
15,000,000.00
Nj (M)
3401
58.83
181.41
0.09
1,337.71
Edj
2017
60.39
30.58
13.83
130.84
GOVAi
3401
-0.49
0.73
-2.11
1.61
GOVAj
3401
-0.49
0.73
-2.11
1.61
PTAi
3401
0.70
0.46
0
1
PTAij
3401
0.70
0.46
0
1
Per discussion from Kanda (2010) , there is both trade creation and trade
expansion if both PTAi (the exporter is a member of a PTA) and PTAij (both trade
partners are member of a PTA) variables are positive. There is trade diversion when
PTAij is positive and PTAi is negative, and there is trade expansion if PTAij is negative
and PTAi is positive.
76
Table 4.3
Impact of SSA Preferential Trading Agreements (PTA) on International Trade, 2010 and
2000, OLS Regression
2010
2000
Variable
Coef.
t
P>t
Coef.
t
P>t
Aij
12994.99
6.49
0.000
*
9086.96
8.41
0.000
*
LOij
-1590.31
-1.12
0.262
-642.24
-0.96
0.336
LEij
1695.63
1.15
0.250
835.63
1.19
0.235
Dij
-0.559
-4.19
0.000
*
-0.278
-4.22
0.000
*
Yi
0.001
4.98
0.000
*
0.001
5.53
0.000
*
Ni
5.46
2.34
0.020
**
0.968
0.85
0.398
EDi
38.98
1.92
0.055
***
15.48
1.90
0.058
***
Yj
0.002
9.14
0.000
*
0.001
8.20
0.000
*
Nj
-2.69
-1.15
0.249
-0.041
-0.04
0.971
EDj
31.27
1.53
0.126
13.80
1.68
0.093
***
PTAi
-2303.80
-1.78
0.075
***
-1325.88
-2.17
0.030
**
PTAij
-2139.78
-1.64
0.101
-1084.77
-1.76
0.078
***
_cons
684.57
0.24
0.807
1328.42
1.18
0.239
Number of obs
1168
1562
F( 12, 1155)
25.02
F( 12,1549)
21.89
Adj R-squared
0.1858
0.1384
*Statistically Significant at the 0.01 level.
** Statistically Significant at the 0.05 level.
*** Statistically Significant at the 0.10 level.
For 2010 and 2000 the F-statistic is significant for both models, and the adjusted
R-Squared is 18.58% for 2010 and 13.84% for 2000, which says that the model predicts
19% and 14% of the variation in exports respectively. In 2010, this model is significant in
seven of 12 variables. Four variables are significant at 0.01, one at 0.05 and two at 0.1.
77
Five variables are negative: LOij, Dij, Nj, PTAi, and PTAij. Dij and PTAi are the only
negative variables that are significant.
Sharing a common official language (LOij) is negative in this and every model,
and not significant in this and every model. LOij is negative in every model, even though
it is expected to be positive. This could be due to multi-collinearity. To assess this,
analysis was conducted using each of the other dependent variables as the independent
variable, and was run with and without the LOij variable.
One mechanism to assess collinearity is to run the regressions in sequence using
each independent variable as the dependent variable one at a time and evaluate for a low
resulting R-squared. When this is conducted, the resulting R-squareds ranged from 31%to
87%, indicating that these results are not of concern.
When the model is run with all of the core variables plus the preferential trade
agreement variables, but without common official language, the F-statistic is significant,
the R-Squared is 18.6%, both PTA variables are significant and negative, and several
other variables are not significant including population of the exporter, secondary
education of the exporter, common ethnic language. If this same analysis is done with
Tobit instead of OLS, both of the PTA variables are significant, though common ethnic
language and population of the exporter are again not significant. Running the model
with the governance variables, the F-Statistic is significant, the R-squared is 18.4% and
neither governance variable is significant, nor is the common ethnic language variable or
the population of the exporter.
78
Table 4.4
Analysis of multicollinearity
Dependent Variable
Adjusted R-squared
Aij
87%
Loij
43%
Leij
43%
Dij
60%
Yi
53%
Ni
60%
EDi
31%
Yj
53%
Nj
60%
Edj
31%
Govi
39%
GOVj
39%
PTAi
49%
PTAij
48%
Analyzing shared border (Aij), distance (Dij), GDP of importer and exporter (Yi
andYj) and Population of importer and exporter (Ni and Nj), the R squared is 11% and all
of the variables except distance and population of the importer are positive. All of the
variables are significant except population of the importer. The R-squared increases to
18.4% when the secondary education variables are added, and those are both positive and
significant.
If Shared Border is combined with Official Language, the F-statistic is significant,
but the R-squared is only 1%. In addition, official language is negative and not
significant. When distance is combined with official language, then the F-statistic is not
significant and both variables are negative and not significant. When the model is run
79
with the two GDP variables and Official Language, the F-Statistic is significant and the
R-squared is 9%. Official language is positive and not significant. When Official
Language is run with GDP and Population of the importer and the exporter, LOij is
positive but not significant. When it is run with ethnic language, both variables are
significant but Common official language is again negative.
This same analysis under Tobit regression shows that the Governance indicator of
the exporter is significant and positive, and the variables that are not significant are the
common ethnic language, and the governance indicator for the importer. The results
when running all of the variables (including governance and preferential trade agreement)
without the common official language variable have a significant F-statistic and R-
squared of 18.8%. Of the base variables, the ones that were not significant are common
ethnic language, secondary education of the exporter and the importer, population of the
importer, governance of the importer and the Y-intercept. When this same analysis is
done with Tobit, the chi-squared is significant and the only variable that is not significant
is the common ethnic language variable and the Y-intercept. The only variables that are
negative are the preferential trade agreement variables and the Y-Intercept. As a result of
this assessment, the Common Official Language variable was retained. Significant
variables in both time periods include whether the countries share a border (Aij), distance
between the two countries (Dij), and GDP of both the exporter and the importer (Yi/Yj),
secondary school enrollment of both the exporter and importer (EDi/EDj), and whether
the exporter is a member of an SSA trade agreement (PTAi). Population of the exporter
80
(Ni) is only significant in 2010, and whether both trading partners are member of an SSA
trade agreement (PTAij) is only significant in 2000.
Aij, Cij, Dij, Yi, Ni, Yj, and Nj are all basic variables in the gravity model and
expected to drive trade. Aij (common border) is positive, which is expected, as two
countries that share a border should have more opportunity for trade. Aij increases trade
by $13B. Dij (Distance between two countries) is negative because the greater the
distance between two countries, the lower the trade. The fact that Dij is negative aligns
with the concept that the farther apart two countries are geographically, the less they are
likely to trade. Increasing Dij (the distance between two countries) decreases trade by
about $560k for every kilometer of distance between the countries.
GDP (Yi/Yj) and Population (Ni/Nj) indicate the size of the economy of the
exporting and importing countries respectively and indicate that countries with larger
economies drive more trade, both as importer and exporter. Higher income level indicates
greater potential supply from the exporting country and increased demand of the
importing country, which would lead to a positive effect on trade. If the population is
large, it would be expected to increase the ratio of domestic to foreign market production,
and thus allow greater output diversification. With greater output diversification, there
would be lower potential demand in the importing country and lower potential supply
from the exporting country, and an overall decrease in trade. (Sandberg 2006). In both
models both the exporting and importing countries’ GDPs are significant at 0.01. An
increase in Yi or Yj (GDP of either the exporter or importer) by $1M increases trade by
about $1,000 for the exporter in both 2010 and 2000, and by $2,000 for the importer in
81
2010 and by $1,000 for the importer in 2000. An increase in Ni (population of the
exporter) by 1M will increase trade by $5.5M in 2010, but only by $968,000 in 2000. The
coefficient on the population of the importer is not significant in either model, but is
negative in both.
Table 4.3 indicates that membership of only one partner in an SSA preferential
trade agreement is significant and has a negative impact on trade in both 2010 and 2000,
specifically that membership in a PTA decreases trade by $2.3B in 2000 and $1.3B in
2000. PTAij is negative but not significant in 2010 and significant in 2000, showing that
in 2000, membership of both trading partners in an SSA PTA results in a $1.1B decrease
in trade. The guidance from Kanda and Jordaan (2010) indicates that there is neither trade
creation nor trade expansion when both PTAi and PTAij are negative. In situations like
the one shown here, with both PTAi and PTAij negative, there is trade contraction.
Table 4.5
Impact of SSA Preferential Trading Agreements (PTA) on International Trade, 1990 and
1980, OLS analysis
1990
1980
Variable
Coef.
t
P>t
Coef.
t
P>t
Aij
2668.03
6.01
0.000
*
748.93
5.34
0.000
*
LOij
-64.26
-0.18
0.859
-84.64
-0.78
0.437
Dij
-0.158
-4.66
0.000
*
-0.073
-6.05
0.000
*
Yi
0.001
4.63
0.000
*
0.00
4.99
0.000
*
Ni
0.281
0.50
0.620
0.065
0.29
0.774
EDi
12.08
2.15
0.032
**
5.54
3.05
0.002
*
Yj
0.001
5.08
0.000
*
0.001
7.63
0.000
*
Nj
0.299
0.52
0.604
0.193
0.8
0.424
EDj
10.83
2.04
0.042
**
3.713
2.01
0.045
**
82
Table 4.5 (continued).
1990
1980
Variable
Coef.
t
P>t
Coef.
t
P>t
PTAi
-767.65
-2.43
0.015
**
-349.94
-3.69
0.000
*
PTAij
-736.66
-2.39
0.017
**
-293.63
-3.06
0.002
*
_cons
1015.46
1.69
0.091
***
498.44
2.83
0.005
*
Number of obs
1294
1707
F( 12, 1281)
17.31
F( 12,1694)
23.31
Adj R-squared
0.1315
0.1356
*Statistically Significant at the 0.01 level.
** Statistically Significant at the 0.05 level.
*** Statistically Significant at the 0.10 level.
The analysis in Table 4.5 looks further back in time. The 1980 data primarily
reflects a time period prior to effect of formal trade organizations. A simple view of the
effectiveness of PTAs would say that from the time period before the organization
formally exists to the subsequent time periods later, the expectation would be to see
higher trade than before the PTA. The first PTA to be implemented, ECOWAS, was
initiated in 1975. ECCAS began in 1983, SADC in 1992, COMESA in 1993 and EAC in
2000. In both 1990 and 1980 the F-statistic is significant. In 1990, the model explains
13.2% of the variance in trade, and the 1980 the results are slightly higher at 13.6% of
trade. The only variables that are not significant in either model are language (official or
ethnic) and population (both exporter and importer). PTAi and PTAij are again both
negative, showing that members of preferential trade arrangements do not trade more
either with other members of their group or with other countries outside of the agreement.
Both variables in 1990 are significant at the 0.05 level and result in around a $750M
83
decrease in exports. In 1980, both PTAi and PTAij are negative and significant at the
0.01 level. Exporter membership in a trade group shows a $350M decrease in trade, and
membership of both parties shows a $300M decrease in trade. In 1980, ECOWAS was
the only trade group that was in place, and only for five years.
Looking back over the four decades in Tables 4.3 and 4.5, the PTAi and PTAij
variables appear to have decreasing significance over time. The trade in countries that
became members of these agreements was most significant in 1980, at 0.01, when only
ECOWAS had an agreement in place, and then in 1990, as more organizations were
established, went to 0.05 significance. In 2000, PTAi was at 0.05 significance and PTAij
at 0.1 and in 2010, only PTAi is significant at 0.1. So while there are consistent negative
trade relationships among the countries in these groups, results demonstrate that the
ability to tie the results in 2010 back to the influence of the PTAs is at its lowest point in
the history of the organizations.
Table 4.6
Impact of SSA Preferential Trading Agreements (PTA) on International Trade, 2010 and
2000 – TOBIT ANALYSIS
TOBIT
2010
2000
Variable
Coef.
t
P>t
Coef.
t
P>t
Aij
21926.83
8.49
0.000
*
17132.78
10.89
0.000
*
LOij
-1215.38
-0.64
0.524
322.00
0.29
0.772
LEij
2411.82
1.23
0.220
2508.41
2.16
0.031
**
Dij
-1.123
-6.11
0.000
*
-0.895
-7.80
0.000
*
84
Table 4.6 (continued).
TOBIT
2010
2000
Variable
Coef.
t
P>t
Coef.
t
P>t
Yi
0.001
5.01
0.000
*
0.001
5.84
0.000
*
Ni
10.86
3.78
0.000
*
10.453
6.28
0.000
*
EDi
154.22
5.43
0.000
*
120.01
8.53
0.000
*
Yj
0.002
8.86
0.000
*
0.002
8.60
0.000
*
Nj
2.73
0.94
0.350
7.665
4.45
0.000
*
EDj
107.19
3.76
0.000
*
71.77
4.99
0.000
*
PTAi
-5450.10
-3.13
0.002
*
-1644.18
-1.61
0.108
PTAij
-4272.55
-2.39
0.017
**
-2258.10
-2.13
0.033
**
_cons
-14435.33
-3.72
0.000
*
-12624.55
-6.53
0.000
*
1168
Number of obs
1562
460
left-censored obs at exports <= .1
849
708
uncensored observations
712
*Statistically Significant at the 0.01 level.
** Statistically Significant at the 0.05 level.
*** Statistically Significant at the 0.10 level.
The tables above (Table 4.6) and following (Table 4.7) conduct the same analysis
of PTAi and PTAij as Tables 4.3 and 4.5, this time using Tobit regression instead of
OLS, in order to address the large number of zero values in the export data. In reviewing
the Tobit results, there are two noticeable differences. First, these results show an
increased coefficient size of some of the variables. For example, in the OLS analysis,
having a common border increased trade by about $13B and in the Tobit analysis of the
same data set, it shows an impact of increasing trade by almost $22B. Second, the Tobit
data show differences in level of significance. In the Tobit analysis, PTAij is significant
in all four of the models. It is significant at 0.05 in 2000 and 2010 and at 0.01 in 1990 and
1980. In the OLS models, PTAij is significant at 0.01 in 1980 (pre-trade agreements), at
85
0.05 in 1990, at 0.01 in 2000 and not at all in 2010. In the Tobit analysis for the PTAi
variable, 2010 results are significant at 0.01 and in OLS at 0.1. In 2000, the Tobit data is
not significant, but the OLS data was significant at 0.05. The Tobit data was significant at
0.01 in both 1990 and 1980, whereas in the OLS data those two years were significant at
0.05 and 0.01. As in the OLS models, there is again, some evidence of decreasing
significance over time. The two variables remain negative in all four of the models.
GDP, population and education level of the exporter are all significant at 0.01 in
all models except for 2000 for education level of the exporter. GDP, population and
education level of the importer is significant and at 0.01 in all models except in 2010 for
the population of the importer. This model does not support the hypothesis that
membership in an SSA trade block increases trade for SSA countries, because the results
are significant and negative on exports.
Table 4.7
Impact of SSA Preferential Trading Agreements (PTA) on International Trade, 1990 and
1980. TOBIT ANALYSIS
TOBIT
1990
1980
Variable
Coef.
t
Variable
Coef.
t
Variable
Coef.
Aij
6770.07
9.39
0.000
*
2346.31
9.53
0.000
*
LOij
-192.63
-0.29
0.774
-184.97
-0.84
0.399
LEij
1117.52
1.59
0.112
719.91
3.19
0.001
*
Dij
-0.302
-4.94
0.000
*
-0.192
-7.82
0.000
*
Yi
0.001
3.81
0.000
*
0.00
5.18
0.000
*
Ni
3.846
4.39
0.000
*
-0.300
-0.62
0.535
EDi
48.59
4.79
0.000
*
16.77
4.63
0.000
*
Yj
0.001
4.64
0.000
*
0.001
6.96
0.000
*
86
Table 4.7 (continued).
TOBIT
1990
1980
Variable
Coef.
t
Variable
Coef.
t
Variable
Coef.
Nj
2.648
2.77
0.006
*
1.755
4.02
0.000
*
EDj
41.29
4.22
0.000
*
16.905
4.5
0.000
*
PTAi
-2758.72
-4.81
0.000
*
-1722.32
-9.04
0.000
*
PTAij
-2181.70
-3.89
0.000
*
-773.78
-3.94
0.000
*
_cons
-2945.57
-2.74
0.006
*
-581.34
-1.66
0.097
***
1294
Number of obs
1707
777
left-censored observations at
exports<=.1
1109
517
uncensored observations
598
*Statistically Significant at the 0.01 level.
** Statistically Significant at the 0.05 level.
*** Statistically Significant at the 0.10 level.
Test of H3: Institutional quality affects trade in SSA countries.
The third hypothesis examines quality of governance in SSA countries. This first
model utilizes the average of the six good governance indicators both for the exporter and
for the importer. The institutional quality indicator is an average of the six indicators of
good governance developed by the World Bank (Kraay, Kaufman and Mastruzzi 2005) :
Voice and Accountability (VA)– political, civil and human rights, Political Instability and
Violence (PS) – likelihood of violent threats to, or changes in, government, Government
Effectiveness (GE)– competence of bureaucracy and quality of public service delivery,
Regulatory Burden(RB) – incidence of market-unfriendly policies, Rule of Law (RL) –
quality of contract enforcement, police, and courts, and likelihood of crime and violence,
and Control of Corruption (CC) – exercise of public power for private gain.
87
In this model of good governance, the data for the governance variables only goes
back as far as 1996. Therefore, for this analysis by decade, data is only available for 2010
and 2000. This first analysis in Table 7 utilizes OLS regression, and neither the quality of
the exporter or the importer is significant in either 2010 or 2000. Utilizing OLS
regression, this would indicate that this hypothesis is not supported by this data, and that
quality of government does not have a significant affect on trade.
Table 4.8
Impact of government quality on International Trade (average of six Government Quality
Indicators), 2010 and 2000, OLS Analysis
OLS
2010
2000
Variable
Coef.
t
P>t
Coef.
t
P>t
Aij
13,325.38
6.65
0.000
*
9,070.97
8.36
0.000
*
LOij
-2,107.90
-1.47
0.141
-876.52
-1.31
0.192
LEij
1,749.00
1.18
0.236
798.02
1.13
0.257
Dij
-0.52
-4.00
0.000
*
-0.25
-3.89
0.000
*
Yi
0.00
3.95
0.000
*
0.001
4.63
0.000
*
Ni
7.68
3.02
0.003
*
1.76
1.55
0.122
EDi
33.27
1.49
0.137
***
14.57
1.58
0.115
Yj
0.00
7.87
0.000
*
0.000
7.24
0.000
*
Nj
-0.95
-0.37
0.710
0.000
0.49
0.622
EDj
30.63
1.37
0.170
13.34
1.42
0.155
GOVAi
1,566.40
1.70
0.090
***
684.56
1.64
0.101
GOVAj
1,196.41
1.30
0.194
511.07
1.21
0.226
_cons
-1,193.11
-0.51
0.608
41.36
0.05
0.963
Number
of obs
1168
1562
F( 12,
1155)
23.07
F( 12, 1549)
21.57
Adj R-
squared
0.1850
0.1365
88
Table 4.8 continued.
*Statistically Significant at the 0.01 level.
** Statistically Significant at the 0.05 level.
*** Statistically Significant at the 0.10 level.
The F-Statistic is significant for the model, and the adjusted R-Squared is 18.5%
in 2010 and 13.65% in 2000, indicating that the model predicts 19%and 14% of the
variation in exports respectively. In 2010, the model is significant in seven of 12
variables. Five variables are significant at 0.01, and two at 0.1. In 2000, the model is
significant in four of 12 variables all at 0.01.
Significant variables in 2010 include whether the countries share a border (Aij),
distance between the two countries (Dij), GDP of both the exporter (Yi) and the importer
(Yj), population of the exporter (Ni) and secondary school level of the exporter (EDi).
The Governance quality indicator of the exporter is significant in 2010 but the importer is
not, and both are positive. Looking at 2000 results, only shared border, the GDP of
exporter and importer, and population of the exporter are significant. The Governance
Matters indicators are positive and not significant. Overall, the null hypothesis can only
be rejected for the exporter in 2010. For the importer in 2010 and both partners in 2000,
the null cannot be rejected utilizing this OLS model, and does not show that quality of
government significantly impacts trade. Table 4.9 will assess the same data using Tobit
regression.
89
Table 4.9
Impact of government quality on International Trade (average of six Government Quality
Indicators), 2010 and 2000. TOBIT Analysis
TOBIT
2010
2000
Variable
Coef.
t
P>t
Coef.
t
P>t
Aij
22,341.18
8.67
0.000
*
16,607.32
10.54
0.000
*
LOij
-2,311.67
-1.30
0.234
-469.92
-0.42
0.677
LEij
2,388.89
1.25
0.226
2,677.53
2.31
0.021
**
Dij
-1.08
-5.87
0.000
*
-0.94
-8.12
0.000
*
Yi
0.00
3.68
0.000
*
0.001
3.70
0.000
*
Ni
15.31
4.95
0.000
*
12.35
7.54
0.000
*
EDi
151.17
4.69
0.000
*
92.95
5.80
0.000
*
Yj
0.00
7.63
0.000
*
0.000
6.83
0.000
*
Nj
5.56
1.76
0.084
***
0.000
5.66
0.000
*
EDj
115.48
3.65
0.000
*
63.13
3.88
0.000
*
GOVAi
3,150.03
2.88
0.012
**
2,695.74
3.79
0.000
*
GOVAj
1,895.96
1.69
0.130
1,855.14
2.60
0.009
*
_cons
-19,666.02
-5.75
0.000
***
-11,689.38
-7.48
0.000
*
1168
Number of
obs
1562
460
left-censored obs
850
708
uncensored obs
712
*Statistically Significant at the 0.01 level.
** Statistically Significant at the 0.05 level.
*** Statistically Significant at the 0.10 level.
In both of these Tobit analyses, the chi-squared is significant at 0.01. In 2010, the
only variables that are not significant are the two common language variables and the
government quality of the importer. Of the 10 variables that are significant, 7 are at 0.01,
one at 0.05 and two at 0.1.In 2000, all variables except for Common official language
(LOij) are significant.
90
Sharing a common border has a $22.4B impact in 2010 and a $17B impact in
2000. , which is $9B and $7B greater than what was shown using OLS. Distance is
negative as expected, with more than double the impact on trade as shown in the OLS
model. GDP, population and level of secondary education of both the exporter and
importer are all positive and significant. Government quality of the exporter is positive at
0.05 in 2010 and at 0.01 in 2000, and increases trade by $3.1B and $2.7B respectively.
Government quality of the importer is positive and not significant in 2010 and significant
in 2000 at 0.1, increasing trade by $2.7B.
One of the governance indicators under review is significant in 2010, and both are
significant in 2000, while only the 2010 exporter is significant in the model utilizing
OLS. Not only are the results showing higher levels of significance using Tobit analysis
(0.01 vs 0.1), the impact shown for each value is also greater, sometimes by close to
400%. For example, quality of governance of the importer is significant with Tobit, not
significant with OLS, and the coefficient is about 60% higher in 2010 using Tobit and
more than three times greater in 2000 with the Tobit model. Utilizing Tobit analysis, we
can conclude that the null hypothesis can be rejected for the impact of governance quality
on the exporter. For the importer, the results are inconsistent and would require further
analysis, such as conducting the same analysis for additional years to assess patterns.
The next analysis examines the individual quality of governance indicators,
utilizing both OLS and Tobit, to understand how these six different factors are impacting
trade.
91
Table 4.10
Impact of government quality on International Trade six Government Quality Indicators),
2010 and 2000. OLS Analysis
OLS
2010
2000
Variable
Coef.
t
P>t
Coef.
t
P>
t
Aij
12,689.05
6.20
0.000
*
10,127.92
8.28
0.000
*
LOij
-2,188.05
-1.49
0.137
-934.85
-1.27
0.206
LEij
1,816.12
1.19
0.235
776.68
0.99
0.321
Dij
-0.73
-5.06
0.000
*
-0.363
-4.79
0.000
*
Yi
0.00
3.56
0.000
*
0.001
3.53
0.000
*
Ni
8.53
2.87
0.004
*
2.19
1.60
0.109
EDi
-3.03
-0.10
0.917
15.96
1.42
0.155
Yj
0.00
7.28
0.000
*
0.001
6.08
0.000
*
Nj
-0.03
-0.01
0.992
0.105
0.77
0.444
EDj
3.94
0.14
0.892
14.82
1.29
0.199
VOCi
1,677.87
1.38
0.169
867.42
1.23
0.218
POLi
236.19
0.27
0.787
-93.41
-0.17
0.867
GOVEFi
7,072.27
2.40
0.017
**
2,428.62
2.17
0.030
**
REGi
797.93
0.39
0.694
-1,003.06
-1.26
0.207
ROLi
-7,877.67
-2.92
0.004
*
-1,918.72
-1.73
0.085
CCi
1,500.42
2.18
0.030
**
851.65
2.27
0.024
**
VOCj
1,601.80
1.33
0.185
834.32
0.81
0.416
POLj
153.64
0.18
0.861
-60.60
0.14
0.886
GOVEFj
5,196.82
1.76
0.079
***
1,932.23
1.57
0.117
REGj
873.27
0.42
0.673
-933.92
0.43
0.670
ROLj
-6,323.84
-2.34
0.020
**
-1,632.90
1.27
0.204
CCj
1,192.03
1.70
0.089
***
774.86
2.05
0.041
**
_cons
1333.00
0.48
0.632
-43.05
-0.05
0.963
Number of
obs
1168
Number of
obs
1561
F( 12,
1147)
14.34
F( 12, 1548)
21.38
Adj R-
squared
0.1861
Adj R-
squared
0.1355
92
Table 4.10 continued.
*Statistically Significant at the 0.01 level.
** Statistically Significant at the 0.05 level.
*** Statistically Significant at the 0.10 level.
Looking back at Table 4.8, assessing the average governance variables with OLS
regression, the only governance variable that is significant is the exporter in 2010 at 0.1.
In this analysis the six governance indicators are assessed individually using OLS, as
shown here in table 4.10. In 2010, six of the 12 governance quality indicators are
significant, the same three variables of Government Effectiveness, Rule of Law and
Control of Corruption, for both the importer and the exporter. Rule of Law of the
exporter is significant at 0.01, and 0.05 for the importer. Rule of law shows as having a
negative impact on trade for both exporter and importer. This could be one of the
situations where similar countries that lack rule of law are trading together and therefore
lack of rule of law drives trade. Government Effectiveness of the exporter in 2010 is
significant at 0.05 and for the importer is significant at 0.1. In both scenarios it is
positive. Control of corruption is significant for the exporter at 0.05 and at 0.1 for the
importer. Both of these variables are positive.
In 2000, Government Effectiveness of the exporter is significant at 0.05, control
of corruption of the exporter and the importer are both significant at 0.05 and all are
positive.
The average governance quality of the exporter is significant in 2010 and yet
three of the individual indicators for both the exporter and the importer are significant in
this individual analysis. The reason that the average governance quality of the importer
93
(GOVAi in prior analysis) was not significant in 2010 may be a combination of the
positive and negative impacts of the individual variables, as well as the fact that the level
of significance on the importer variables is less strong than the level of significance on
the exporter variables.
Table 4.11
Impact of government quality on International Trade six Government Quality Indicators),
2010 and 2000. TOBIT Analysis
TOBIT
2010
2000
Variable
Coef.
t
P>t
Coef.
t
P>t
Aij
22,314.34
8.54
0.000
*
19,595.73
10.97
0.000
*
LOij
-1,562.61
-0.79
0.430
-1,360.08
-1.08
0.282
LEij
446.01
0.22
0.826
2,003.64
1.53
0.127
Dij
-1.24
-6.3
0.000
*
-1.11
-8.16
0.000
*
Yi
0.00
2.69
0.007
*
0.000
1.88
0.060
**
Ni
14.84
3.9
0.000
*
12.61
6.10
0.000
*
EDi
81.33
1.98
0.047
**
71.06
3.67
0.000
*
Yj
0.00
6.18
0.000
*
0.001
5.25
0.000
*
Nj
6.16
1.62
0.105
1.16
5.42
0.000
*
EDj
45.74
1.12
0.262
54.51
2.74
0.006
*
VOCi
69.56
0.04
0.967
1,004.03
0.80
0.422
POLi
-2,919.07
-2.26
0.024
**
-3,122.15
-2.98
0.003
*
GOVEFi
5,714.29
1.38
0.168
10,359.93
5.29
0.000
*
REGi
9,197.01
3.1
0.002
*
-684.52
-0.47
0.636
ROLi
-5,092.66
-1.38
0.169
-3,822.48
-1.94
0.053
***
CCi
-966.27
-1.02
0.310
363.96
0.58
0.564
VOCj
998.45
0.6
0.552
2,769.11
2.22
0.027
POLj
-3,442.09
-2.78
0.005
*
-2,465.68
-2.49
0.013
**
GOVEFj
6,640.25
1.62
0.106
5,627.30
2.84
0.005
*
REGj
4,024.50
1.36
0.174
-341.56
-0.24
0.814
ROLj
-3,015.42
-0.82
0.415
-3,049.61
-1.57
0.118
CCj
-699.24
-0.72
0.474
536.88
0.84
0.400
_cons
-11,375.61
-2.57
0.010
**
-10,795.58
-5.85
0.000
*
94
Table 4.11 (continued).
TOBIT
Variable
2010
2000
1168
Number of obs
1403
460
Left-censored
observations
772
708
Uncensored
observations
631
0
Right-censored
observations
0
*Statistically Significant at the 0.01 level.
** Statistically Significant at the 0.05 level.
*** Statistically Significant at the 0.10 level.
Looking back at, Table 4.9, which assesses the average governance variables with
OLS regression, the average governance quality of the exporter is significant in 2010 and
the average governance quality of both the exporter and the importer are significant in
2000. In this Tobit analysis in Table 4.11 with the six indicators broken out, there are
quite a few differences from the OLS analysis. Political instability of the exporter and
importer is significant in both 2010 (.05) and 2000 (0.1), and was not significant at all
with OLS, and Regulatory Burden of the exporter is also significant in 2010, but was not
significant with OLS.
In the OLS analysis, Government Effectiveness, Rule of Law and Control of
Corruption of both exporter and importer are all significant in 2010 and in 2000, only
Government effectiveness of the exporter is significant, none of the other five variables
are significant. In this Tobit analysis in 2010, Government Effectiveness is not
95
significant at all, neither for the exporter nor the importer, though it is significant for both
in 2000, and it is positive in all instances.
What is unusual about this Tobit analysis is that in every other comparison
between OLS and Tobit, the Tobit results have all the variables significant that are
significant with OLS, plus additional variables, and at a higher degree of significance. In
this analysis, the Tobit results have a couple of significant variables that were not
significant with OLS, but does not show all the variables as significant that were
significant with OLS. This is unusual compared to all of the other findings in this study,
and it would be helpful to run this same analysis for the other years for which the data is
available to see how these results compare.
The analysis that follows examines combined variables that impact all three of the
hypotheses for this dissertation. The governance variables are only available in 2000 and
2010, so these results only cover those two time periods.
Table 4.12
Impact of preferential trade agreements (PTA) and government quality on International
Trade (average of six Government Quality Indicators), 2010 and 2000. OLS Analysis
OLS
2010
2000
Variable
Coef.
t
P>t
Coef.
T
P>t
Aij
13,069.74
6.53
0.000
*
8,871.32
8.17
0.000
*
LOij
-g,058.10
-1.44
0.150
-800.97
-1.19
0.233
LEij
1,867.77
1.27
0.205
947.52
1.34
0.179
Dij
-0.63
-4.60
0.000
*
-0.304
-4.53
0.000
*
Yi
0.00
3.53
0.000
*
0.001
4.33
0.000
*
Ni
7.47
2.94
0.003
*
1.31
1.13
0.259
EDi
7.64
0.29
0.768
5.73
0.57
0.570
96
Table 4.12 (continued).
OLS
Variable
2010
2000
Coef.
T
P>t
Coef.
T
P>t
Yj
0.00
7.43
0.000
*
0.001
6.96
0.000
*
Nj
-1.11
-0.44
0.663
0.023
0.20
0.844
EDj
7.30
0.28
0.779
6.59
0.65
0.517
GOVAi
1,841.16
1.98
0.048
**
710.05
1.70
0.089
GOVAj
1,453.40
1.57
0.117
527.39
1.25
0.210
PTAi
-2,674.95
-2.05
0.040
**
-1,350.17
-2.21
0.027
**
PTAij
-2,454.04
-1.87
0.062
***
-1,108.78
-1.80
0.072
***
_cons
5,937.92
1.68
0.093
***
2,782.14
2.09
0.037
***
Number of obs
1168
1562
F( 14, 1153)
20.38
F( 14, 1147)
19.1
Adj R-squared
0.1887
0.1397
*Statistically Significant at the 0.01 level.
** Statistically Significant at the 0.05 level.
*** Statistically Significant at the 0.10 level.
The F-Statistic is significant for the model, and the adjusted R-Squared is 18.87%
in 2010 and 13.97% in 2000, indicating that the model predicts 19% and 14% of the
variation in exports respectively. In 2010, the model is significant in nine of 14 variables.
Five variables are significant at 0.01, two at 0.05, and two at 0.1. In 2000, the model is
significant in seven of 14 variables. Four variables are significant at 0.01 and one at 0.05
and two at 0.1.
Significant variables in 2010 include whether the countries share a border (Aij),
distance between the two countries (Dij), GDP of both the exporter (Yi) and the importer
(Yj), population of the exporter (Ni), quality of governance of the exporter (GOVAi),
SSA PTA membership of the exporter (PTAi) and PTA membership of both the exporter
and importer (PTAij). Three of the four indicators being tested in this model are
97
significant, and only the governance quality of the importer (GOVAj) is not significant.
The PTA variables continue to be negative.
The results of this analysis show that with both PTAi and PTAij negative and
significant, then there is confirmed trade contraction (Kanda and Jordaan 2010). This
model shows that membership in an SSA PTA decreases trade by $2.7B when the
exporter is a member of an SSA PTA and by $2.6B when both trading partners are
members of a PTA. In addition, this shows that the impact of government quality of the
exporter (GOVAi) is significant, as it also is when tested without the PTA variables
(Table 4.7). The results again disprove hypotheses one and two and validate hypothesis
three on the part of the exporter.
Looking at 2000 results, the major differences with the 2010 data are that the
variables for the population of the importer (Ni) and the governance quality of the
exporter (GOVAi) are not significant. The PTAi and PTAij variables are significant and
negative, thus are unable to reject the null hypothesis. When the exporter is a member of
an SSA PTA, trade is reduced by $1.4B and by $1.1B when both partners are members of
an SSA PTA.
Kuncic (2013) conducts analysis similar to that in this study, utilizing regional
trade agreements and institutional quality. He also includes membership in world trade
organization and indicators for institutional distance that he developed himself (Kuncic
2012). The focus of his study is on Asian-Caribbean Pacific trade with the European
Union (EU). He determines that the institutions of both the exporter and the importer are
important, and says that good economic and political institutions on the exporter side
98
seem to encourage trade. This is similar to what is seen in this analysis – results of the
governance indicators for the exporter are stronger than those for the importer. De Groot
et al. (2004) also indicate that institutional similarities increase trade.
Table 4.13
Impact of Sub-Saharan Africa (SSA) preferential trade agreements (PTAs) and
government quality on International Trade (average of six Government Quality
Indicators), 2010 and 2000. TOBIT Analysis
TOBIT
2010
2000
Variable
Coef.
t
P>t
Coef.
t
P>t
Aij
21,952.35
8.53
0.000
*
16,423.85
10.44
0.000
*
LOij
-2,372.07
-1.23
0.221
-462.85
-0.41
0.681
LEij
2,870.32
1.46
0.145
3,037.35
2.60
0.009
*
Dij
-1.24
-6.64
0.000
*
-0.982
-8.42
0.000
*
Yi
0.00
3.03
0.002
*
0.001
3.49
0.000
*
Ni
14.71
4.65
0.000
*
11.86
6.98
0.000
*
EDi
90.84
2.54
0.011
**
83.11
4.80
0.000
*
Yj
0.00
7.00
0.000
*
0.001
6.54
0.000
*
Nj
5.14
1.60
0.109
0.876
4.99
0.000
*
EDj
69.32
1.93
0.054
***
47.95
2.70
0.007
**
GOVAi
3,636.66
2.90
0.004
*
2,670.08
3.76
0.000
*
GOVAj
2,277.54
1.82
0.069
***
1,801.67
2.53
0.011
**
PTAi
-6,036.77
-3.46
0.001
*
-1,401.37
-1.37
0.172
PTAij
-4,600.16
-2.57
0.010
*
-2,130.42
-2.01
0.045
**
_cons
-5,046.91
-1.05
0.294
-7,884.64
-3.55
0.000
*
1168
Number of
obs
1562
460
left-censored
observations
850
708
uncensored
observations
712
99
Table 4.13 continued.
*Statistically Significant at the 0.01 level.
** Statistically Significant at the 0.05 level.
*** Statistically Significant at the 0.10 level.
In this Tobit analysis, the chi-squared is significant at 0.01, and significance of the
results obtained in the two time periods differs. In 2010, common official common ethnic
language, population of the importer and the intercept are not significant. Eleven of the
15 variables are significant, eight of those at the 0.01 level. These variables include
sharing a common border (Aij), distance (Dij), GDP of the importer (Yi) and the exporter
(Yj), population of the exporter (Ni), governance quality of the exporter (GOVAi), and
SSA PTA membership of the exporter (PTAi) or of both trading partners (PTAij). One
variable, secondary education of the exporter, is significant at the 0.05 level. . Two final
variables are significant at the 0.1 level, the secondary education level of the importing
country (EDj) and governance quality of the importer (GOVAj). The signs on the
significant variables are all positive except for Distance, which is expected to be
negative, and the PTA variables, PTAi and PTAij, which again shows that the hypothesis
that preferential trade agreements in SSA countries improve trade is not valid. This model
shows that the hypothesis that good governance improves trade is valid for both the
exporter and the importer.
Sharing a common border is significant and positive with a $22.0B impact, which
is $9B greater than what was shown using OLS. Distance is significant and negative as
expected, with about double the impact shown in the OLS model. Government quality of
the exporter is positive, increases trade by $3.6B and is roughly double the impact
100
identified in the OLS model. The government quality of the importer is positive and
increases trade by $2.3B. PTAi decreases trade by $6.0B, twice the result from the OLS
model, and PTAij decreases trade by $4.6B, about two million more than the results from
the OLS model.
In 2000, every variable except common official language and membership in an
SSA PTA by the exporter is significant, including the intercept. There are nine variables
significant at the 0.01 level and three at 0.05. Sharing a border (Aij) increases trade by
$16.4B, which is almost double the results from the OLS model. Sharing a common
ethnic language impacts trade by 3B, which is more than triple the OLS results. Other
notable differences from the OLS model are distance (Dij) which is more than triple the
OLS output, population of the exporter (Ni) which is about eight times more than the
OLS model, secondary education level of the exporter (Edi) which is about 15 times
more than the OLS model, population of the importer (Nj) is more than 30 times bigger
than the OLS output. For the variables being assessed, the two government quality
indicators are more than three times greater than with OLS, and PTAi is slightly higher,
while PTAij is about double. Overall, this model shows that the governance quality
hypothesis is valid, that good governance of both the exporter and the importer improve
trade. As seen throughout this document, PTA membership serves only to reduce trade.
101
CHAPTER V
CONCLUSION
This paper assesses the impact of regional trade agreements (RTAs) and
government quality on international trade. The results of this analysis demonstrate that
SSA RTAs have a negative impact on trade while government quality has a positive
impact. There are three hypotheses in this paper which assess trade by exporting
countries who are part of an SSA trade agreement, trade among countries when both
trading partners are members of an SSA trade agreement, and trade impacts of
government quality.
This paper analyzes SSA trade agreements: South African Development
Community (SADC), East African Community (EAC), the Common Market for Eastern
and Southern Africa (COMESA), the Economic Community of West African States
(ECOWAS) and the Economic Community of Central Africa States (ECCAS).
Theories examined in this paper are based on economic geography and the effect
of institutions on development. Combes, Mayer, and Thisse (2008) state that economic
geography addresses location, such as where producers are located in relation to
consumers, and benefits and costs of location-related choices. Puga and Venables (1999)
suggest that poor countries transition to a developed state one at a time, at a rapid pace.
Acemoglu, Johnson, and Robinson (2004) and North (1990) identify that economic
institutions, which align resources with most efficient resources, are the determining
factor for a country’s ability to develop and grow. Acemoglu, Johnson, and Robinson
102
say that economic institutions define both an economy’s potential for economic growth
and distribution of wealth resulting from economic growth.
De Melo and Tsikata (2014) state that PTAs and regional integration are the right
direction, but the integration approach taken to date has not proven successful. They
describe the integration model as linear, beginning with goods, labor and capital markets,
through fiscal integration in a monetary union. De Melo and Tsikata also note the lack of
trade infrastructure and full integration that makes the borders between partner countries
more permeable, allowing a greater flow of labor and development of infrastructure
where they are needed. Lack of compensation mechanisms and inconsistent distribution
of gains has obstructed progress, as has a focus on trade in goods rather than trade in
services, which is growing worldwide at a higher level. De Melo and Tsikata (2014) point
out that the PTAs in Africa are made up of countries with extremely varied access to
resources, and this is more than in other parts of the world. Permeable borders also
increase the likelihood of UTT, which results in a lower need for formal trade agreements
and moves the countries away from the undesirable Nash equilibrium with out any
negotiation required.
A 2012 technical paper on potential growth in SSA countries (International Trade
Centre 2012) identifies lack of value-added goods as a key issue for SSA countries and
makes three recommendations to improve SSA growth. The International Trade Centre
(2012) proposes improving SSA infrastructure to reduce transportation costs by half (at
the cost of $13B per year for 10 years, assumed not to be funded by SSA countries),
improving customs procedures to reduce processing time by half within two years, and
103
reducing SSA-Asian trade costs. As a result of these changes, the International Trade
Centre proposes that SSA exports would increase by as much as 50%, with improvement
in terms of trade, increases in imports and transformed products. They forecast an
increase in real income between one percent and 1.7% and a GDP gain of $15B by 2025.
In a case study of Bharti Airtel, an Indian cellular company which expanded
operations to SSA in 2010, Palepu and Biljani (2012) note the high costs of
transportation in SSA countries. With operations in 16 countries, Bharti Airtel could not
reach landlocked countries via ship and would have to make deliveries via helicopter,
elephant, or even build a road for access. The time required for transportation only served
to increase the costs, taking 15 days up to two months from port to final destination. They
also describe that air transportation from one country to another could take as much as 30
hours, frequently with routing through Europe.
H1: Membership in an active SSA trade bloc increases trade in SSA countries:
SSA countries that are members of SADC, COMESA, EAC, ECCAS or ECOWAS
experience higher trade than countries that are not members of one of those five trade
blocs. Hypotheses one utilizes a preferential trade agreement (PTA) variable. PTAi
represents when the exporting country is a member of an SSA trade group.
Results of analysis show that this these variable is significant, but has a negative
impact on trade. When only one trading partner is a member of a PTA, then trade
declines. This variable is significant and negative in all four time periods (1980, 1990,
2000 and 2010). In the earliest time period of 1980, before there are formal trade
agreements in place for sufficient time to have impact, the significance is strongest (yet
104
still negative) at 0.01. In 1990 it is significant at 0.05 and in both 2000 and 2010, it is
significant at 0.1. Thus significance decreases over time. With Tobit analysis, both 1980
and 1990 are at 0.01 significance. 2000 is not significant, and 2010 is again significant at
0.01. The next hypothesis assesses whether there are benefits within the trading groups.
H2: Membership in an SSA trade bloc increases trade between members of the
trade bloc: Trade is higher when both countries are members of an SSA trade bloc
(member-to-member trade). The second hypothesis also uses a trading bloc variable,
PTAij, which reflects when both the exporting and importing countries are members of
an SSA preferential trade agreement. This variable is significant in three of the four time
periods using OLS, and is negative in all time periods.
In 1980, the earliest time period, before there are formal trade agreements in
place, the significance is strongest at 0.01. In 1990 it is significant at 0.05 and in 2000 it
is significant at 0.1, and in 2010 it is not significant at all. With Tobit, PTAij is
significant in 1980 and 1990 at 0.01 and in both 2000 and 2010 at 0.05. Again, the level
of significance and the size of the coefficients is greater using Tobit. Thus significance
decreases over time, which indicates that influence of regionalism in general is
decreasing, and given that the variable is always negative, these trading blocs have not
been able to successfully create trade with their partner countries.
Across these two hypotheses, both PTAi and PTAij are negative in all four
models. If the dependent variable analysis is positive, it reflects trade creation because it
shows there is more trade among the countries who are both members of the trading
group. If the coefficient is positive, it is evidence of more trade than expected between a
105
member of the trading bloc and a non-member. If the coefficient is negative, it indicates
trade diversion because there is less trade with non-PTA countries than is expected
(Kanda and Jordaan 2010). In this study, with both PTAi and PTAij negative, trade is
neither being created nor diverted, trade is contracting according to Kanda (2010).
H3: Institutional quality of SSA countries affects trade. This third hypothesis
assess the impact of good governance on international trade. The analysis in Table 47
utilizes OLS regression. The quality of the exporter but not the importer is significant in
2010, and neither the quality of the exporter or the importer is significant in 2000. If the
OLS analysis were sufficient to asses this information, then results indicate that this
hypothesis is not supported by this data, and that quality of government does not have a
significant affect on trade.
With Tobit regression in 2010, the governance indicator of the exporter is
significant at the 0.05 level and of the importer is not significant. In 2000, Government
quality of both the exporter and the importer is positive at 0.01.In 2010, government
quality of the exporter is positive and increases trade by $3.1B and government quality of
the importer is positive and increases trade by $1.9B. In 2000, government quality of the
exporter is positive and increases trade by $2.7B and the government quality of the
importer is positive and increases trade by $1.9B. This model demonstrates the benefit of
utilizing Tobit analysis to address the large number of zero values found in international
trade data. With the Tobit regression, zero values are replaced with a small number, 0.1
in this instance, to address the fact that in most instances the zeroes reflect unreported
rather than nonexistent trade.
106
With Tobit, both of the governance indicators under review are significant in
2000, where neither is significant in the model utilizing OLS. Results show not just
significance, but high significance, with Tobit analysis. Each coefficient is also greater,
sometimes by almost 400%. Utilizing Tobit analysis, we can conclude that the null
hypothesis can be rejected and we can see the significant impact of good governance on
trade.
The combined model with both the trade agreement variables and the good
governance variables shows that both of the PTA variables are significant in 2010 and
2000 with OLS, more significant than when the model is run without the governance
variables. For PTAi, in the model without governance, it is significant at 0.1 and 0.05 and
in the model that includes governance it is significant at 0.05 in both 2010 and 2000. For
PTAij, in the model without governance it is not significant in 2010 and significant at 0.1
in 2000, and with the governance variable it is significant at 0.01 in 2010 and not
significant at all in 2000
Utilizing Tobit regression to analyze both the PTA variables and the governance
variables, significance and size of the coefficients again increases. In 2010, governance
quality of the exporter (GOVAi), and SSA PTA membership of the exporter (PTAi) or of
both trading partners (PTAij) are all significant, but PTAi is not significant in 2000.
Governance quality of the importer (GOVAj) is significant at 0.1 in 2010 and 0.05 in
2000. For the governance variables, where the only significant variable was the exporter
in 2010 at 0.1, with Tobit both governance variables are significant in both years.
107
Contributions of this Research
This research contributes to understanding related to both trade blocs and good
governance, and to use of Tobit regression with the gravity model. In relation to trade
blocs, this research demonstrates that SSA trading agreements are not positively
impacting trade. On the other hand, the government quality variables show that trade is
impacted by the quality of government. The analysis shows that Tobit regression
provides stronger results with the gravity model than OLS.
Limitations of this Research
There are a number of limitations to this research, beginning with the trade data
itself. As has been noted within this document, there is a great deal of unreported trade
and the trade data itself frequently has zero values even where some trade exists. There
are also limitations to the governance variables, since they are not available before 1996..
As noted by Kaufmann, there are limitations to the governance data itself indicating that
the data should not be utilized at a detailed level to compare one country to another, but
only to compare relative positions.
Implications for Future Research
Future research should include analyzing the impacts of individual PTAs over
time and further assessing other studies of these SSA PTAs and their impacts on the
region. Several of the groups have peacekeeping functions, so that may have more
influence than the trade impacts, but if they are not driving growth, then there is still need
for more understanding of development and growth in SSA. Deeper analysis of what
makes a PTA successful might be of use, otherwise if interventions are implemented like
108
those suggested by the International Trade Centre (2012) they should be carefully
monitored for impact.
Further analysis could be conducted on the negative results of the common
official language variable. With good governance, there is opportunity to conduct further
analysis on the six individual variables and understand their impact on countries like the
SSA countries to further identify impacts on growth. With regard to the governance
variables, more analysis of data across all available years would be helpful in better
understanding the implications of the findings in this paper.
Another area of research, to further understand impacts on growth at a more
granular level, would be to augment variables data related to size of tariff reductions,
exchange rate management and access to imported inputs. Babatunde (2009) notes that
there a contradictory findings on whether trade liberalization has improved export
performance and examines SSA countries after the trade liberalization efforts that were
introduced subsequent to the IMF programs of Import Substitution Industrialization (ISI)
and Structural Adjustment Programme (SAP) in the 1980s and 1990s. Babatunde utilizes
panel least squares estimation technique and analyzes the SSA countries by region (West,
East, South and Central). What Babatunde determines is that trade liberalization is not
sufficient to increase exports. He finds no significant effect of reductions in trade
protections on exports, but that there is significance with access to imported inputs and
effective valuation of exchange rates. Utilizing these variables with the gravity model and
Tobit analysis would provide additional understanding of SSA growth.
