Exporting to Fragile States in Africa: Firm Level
Evidence
21/04/2017
Peter W. Chacha and Lawrence Edwards
Abstract
This paper examines the effect of destination country fragility on a firm’s decision to serve
that destination with exports. We find that fragility has a negative effect on bilateral trade
between countries. This negative effect occurs mainly along the extensive margin because
fragility reduces the number of firms exporting to a given country. In particular, there is a
large and significant negative effect of fragility on the probability to export to a fragile state.
This negative effect is however, counteracted by two firm attributes; an increase in the firm’s
size and belonging in a network, which are associated with an increase in the probability to
serve a given destination country.
Keywords: Fragility, Firm’s size and network, International trade.
JEL Classification: F51, F12, F14, D81
This draft paper is submitted as a framework paper under the AERC project on Growth in Fragile States in Africa. We would
like to thank Prof. Anke Hoeffler for comments on an earlier draft of the paper and comments received from participants in
the Final Review Workshop held between 4-5th March 2017, in Nairobi Kenya. The paper also benefited from comments
received at the school seminar, on 3rd April, 2017.
2
Contents
Abstract 1
1.0 Introduction 3
2.0 Related Literature 5
3.0 Theoretical Framework 9
3.1 The Model ........................................................................................................................ 9
3.2 Estimation strategy......................................................................................................... 15
3.3 Data and Data sources .................................................................................................... 17
3.4 Stylized facts .................................................................................................................. 19
4.0 Empirical Results 23
4.1 Destination fragility and firm’s export status ................................................................ 23
4.2 Firm’s size and exports to fragile states ......................................................................... 25
4.3 Firm’s network and exporting to fragile states .............................................................. 26
4.4 Destination fragility and Trade margins ........................................................................ 28
4.5 Robustness tests ............................................................................................................. 30
5.0 Conclusion 34
References 35
Appendix 1A: Country rank by average share of export and fragility ............................. 38
Appendix 1B: Measurement of Key variables 39
Appendix 1C: Decomposition of Kenya’s exports into intensive and extensive margins 40
Appendix 2: Mathematical derivations 41
LIST OF TABLES
Table 1: Summary statistics for the key variables ................................................................................ 19
Table 2: Top 15 export destinations and the respective fragility index, 2005 and 2012 ...................... 21
Table 3: Firm’s export status and destination country fragility ........................................................... 23
Table 4: Firm’s size and export status to a given country .................................................................... 25
Table 5: Firm network and exports to a given destination ................................................................... 27
Table 6: Destination fragility and trade margins to destination j ........................................................ 29
Table 7: Firm’s export status to a given destination and other proxies for fragility ............................ 31
Table 8: Robustness to alternative measurement of fragility using ICRG Index .................................. 32
Table 9: Alternative measurement of Fragility and trade margins ...................................................... 33
Table 10: Export share and designation fragility index (mean values 2004-2013) .............................. 38
Table 11: decomposition of Kenya’s export trade to destination j, 2012 ............................................. 41
LIST OF FIGURES
Figure 1: Ranking of African countries by their fragility index, 2005 and 2012 ................................. 20
Figure 2: Average exports and mean destination fragility over time ................................................... 22
Figure 3: Share of Kenya’s total exports to Rwanda and Madagascar, 2004-2013 ............................ 22
3
1.0 Introduction
Exports flow into fragile states in Africa faces severe constraints due to weak supporting
institutions and in general, a tougher environment for doing business (Brenton et al., 2012,
Cadot et al., 2013). Fragility in a destination elevate uncertainty regarding the sunk costs of
entry for the potential exporters and may curtail export activity of small and less efficient
firms. Understanding the effect of destination fragility on firm’s entry decision and the
mechanisms through which firms mediate this effect is important for trade within sub-
Saharan African countries. This paper examines the effect of destination country fragility on
a firm’s decision to serve that destination with exports. It also evaluates firm attributes that
mediate the effect of fragility on exporter’s entry decision.
Destination country fragility is a multi-dimensional concept and hard to define and measure.
There is no globally accepted definition of fragility1 and nations do not like this label (see
Stewart and Brown, 2009 for a review of existing definitions). Chauvet and Collier (2008)
provide an economic definition of the concept to constitute a low-income country in which
economic policies, institutions and governance are so poor that growth is highly unlikely. In
the World Bank’s Development Report on Conflict, Insecurity and Development (World
Bank, 2011), fragile situations are defined as “periods when states or institutions lack the
capacity, accountability or legitimacy to mediate relations between citizen groups and
between citizens and the state, making them vulnerable to violence” (p.18).
Despite the challenge of definition and measurement, fragility remains a common feature
among countries in Africa, across the various proxies and indicators used to measure this
phenomenon. An example of the common index used to measure this phenomenon include
the Worldwide Governance Indicators (WGI) by Kaufmann et al (2011); the Country Policy
and Institutional Assessment (CPIA) Indices (World Bank, 2014); the International Country
Risk Guide (ICRG) index (PRS Group, 2011) and Fragile States Indices (Messner,2005). Our
measurement is derived from different combinations of the Worldwide Governance
Indicators (WGI) by Kaufmann et al (2011), using the Principal Component Analysis (PCA).
1 Our definition of fragility situations is taken from the World Bank’s World Development Report (2011).
4
There are numerous studies that have investigated the economic consequences of spatial
diffusion of fragility and conflict on neighbouring countries (Sambanis, 2001; Fearon and
Laitin, 2003; Murdoch and Sandler, 2004; Collier and Hoeffler, 2004, Iqbal and Starr, 2008
and Qureshi, 2009). These papers have found large negative spillovers of fragility on
economic growth of neighbouring states. Chauvet et al (2007) show that the bad neighbour
effect accounted for approximately 0.6% shortfall on economic growth in countries
neigbouring a fragile state. One potential channel through which fragility hinders economic
growth is in reduction of bilateral trade between neighbouring countries (Martin et al., 2008).
This paper examine the effect of destination country fragility on a firm’s decision to serve a
give destination with exports. Specifically, we ask three main research questions:
What is the effect of destination country fragility on an exporter’s decision to
maintain export relationship to that country?
What firm attributes mediate the effect of destination country fragility?
What is the effect of destination country fragility on bilateral trade margins?
In addressing the above questions, this paper contributes to existing literature in three main
ways. Firstly, an increase in destination fragility can be viewed as an exogenous shock to the
cost of exporting to affected countries. Existing models in the new trade literature currently
provide little insight into how fragile situations in a destination affect a firm’s export
decision. By working with a frictionless environment that permits efficient allocation of
resources across and within firms, the new trade models (Melitz, 2003, Bernard et al.,2003,
Chaney, 2008) ignore the market conditions in destinations that are important for low income
countries exporters. We extend and simplify a monopolistic competition model of trade in an
approach closely related to Crozet et al (2007) to account for a specific type of market
friction, namely fragility of a destination country and evaluate its effect on the firm’s decision
to serve a given destination with exports.
Secondly, we test the hypothesis from the model using Kenya’s firm level transaction data
and evaluate the role of firm size and network in mediating the effect of destination fragility.
Kenya trades with several countries plagued by weak institutions and largely considered
fragile such as Burundi, Somalia, Sudan, South Sudan and Democratic Republic of Congo.
Firms from Kenya, it would appear, bear the burden of additional costs associated with
5
exporting to fragile states. This makes it an excellent case study on the effect of destination
country fragility on firm’s decision to maintain export relations to a given destination.
Thirdly, a study on the effect of destination country fragility on firm export behavior is
extremely important for Sub-Saharan Africa (SSA). Although a marginalized topic in
international economics and treated sometimes as an indirect/or a hidden tax on trade
(Anderson and Marcouiller, 2002; Blomberg and Hess, 2006) fragility might play a role in
explaining low bilateral trade among neighbouring countries (Martin et al., 2008) and may be
a source of low intra-African region trade. Yeats (1998) argues that the potential for intra-
African trade is yet to be fully exploited, especially in food products. This is puzzling despite
popular embracing of regional integration initiatives as a core part of trade policy in Africa
(Carrere, 2004). Studies looking at the effect of fragility on firm level export decisions are
very few in the international trade literature (Crozet et al., 2007). This paper adds to this
literature.
The rest of this paper is organized as follows; section 2.0 reviews related literature, section
3.0 discusses the theoretical model, estimation strategy and describes the dataset, section 4.0
presents and discusses the empirical results, while section 5.0 concludes.
2.0 Related Literature
This paper is motivated by several strands of literature, including heterogeneous firms and
trade, trade under uncertainty and information asymmetry, literature on the consequence of
destination conflict and fragility on trade and literature from the field of political science2.
Since the late 1990s, understanding the relationship between firm’s export decisions and their
characteristics has been central to research in international trade (Greenaway and Kneller,
2007; Wagner, 2007). Melitz (2003) develops a model that is able to explain the behaviour of
firms breaking into international markets in light of fixed costs of entry. The model
emphasizes that entry into the export market involves huge fixed costs and only the most
productive firms self-select into exports because of ability to overcome costs. A number of
papers seeks to investigate further, what constitute this fixed costs and what mechanisms do
firms use to reduce them (Chaney, 2008, Crozet et al., 2007, Choquette and Meinen, 2015).
2 The literature in political science examines the reverse causality. The liberal view argues that increasing trade flows limits
the incentives of conflict, while the realist view argues that asymmetric trade links and competition lead to conflict (see Martin
et al. 2008). We abstract from this literature and focus on the effect of fragility on trade.
6
The Melitz model is highly flexible and explains well firm level export behaviour in both
developed and developing economies. However, the model has been criticized by Crozet et al
(2007) for being blind to critical market frictions in trading with conflict and insecure
countries. Crozet et al (2007) modify the basic Melitz model to incorporate insecurity. They
show that ex-ante, insecurity affects all firms since all of them face the same risk, however,
ex-post some of the firms are not affected. Only a random subset of firms is subject to
predation while others are lucky and export without misfortune. In addition, a high level of
insecurity may dissuade unlucky productive firms from exporting, while some lucky
unproductive ones may succeed. There are two testable hypotheses in this framework. Firstly,
the prevalence of insecurity will decrease bilateral exports by reducing the number of
exporters. Secondly, a higher level of insecurity may dissuade unlucky productive firms from
exporting to that destination while some lucky unproductive ones may succeed in exporting
to insecure destinations.
This paper is also related to the literature examining firm’s strategies in overcoming
destination country risks and maintaining export relationships. In Rauch and Watson (2003)
and Araujo and Ornelas (2007), the behaviour of exporters is summarized in a sequential
equilibrium. In the first step, exporters from a home country form a partnership with
distributors in a foreign country that has weak institutions and imperfect contract
enforcement. Exporters do not know the type of distributors they are dealing with, some of
whom care little about the future (impatient). In the second stage, exporters choose an initial
volume of goods to be shipped to importers in risky destinations. The presence of search
costs and information asymmetry causes firms to start small in order to test the credibility of
their foreign trade partners. Once this credibility is learned and networks established, firms
increase their volume of shipments as well as the number of products exported to that
destination.
This literature, in turn, points to the importance of networks and the building of a reputation
to maintaining trade relationships in destinations that may be considered fragile or with poor
quality of institutions3. To test for the importance of reputation and networks in overcoming
risks, McMillan and Woodruff (1999) use cross-sectional survey evidence from the
Vietnamese private sector to show that, in an environment characterized by the absence of
3 see Acemoglu et al., 2005, and Easterly, 2005 for detail analysis on the quality of institutions and economic growth.
7
formal contract enforcement, long term trade relationships (or reputation) facilitate access to
trade credit. Making use of the amount of trade credit granted as a measure of business trust,
they found that credit was offered when it was difficult to find an alternative supplier, or the
supplier had information about the customer’s reliability following previous interaction and if
the supplier belonged to a business or a social network. This network provided both
information about a customer’s reliability and as a means to sanction those who reneged on
deals.
Similarly, Macchiavello and Morjaria (2015) study the importance of reputation in the
context of Kenya’s rose flower export sector. The nature of flower products renders them
highly perishable meaning that relationships between sellers and buyers of roses from Kenya
are not based on a written and enforceable contract. This means trading partners rely on
repeated transactions and the building of a reputation and trust. They test for the endurance of
this reputation by examining the reaction of exporters in the midst of the December 2007 post
general election violence. The authors model violence as an unanticipated and observable
shock that makes it impossible to deliver roses unless the exporter was willing to undertake
additional costs, such as hiring a security escort. Their results show that indeed, there was an
incentive for exporters to deliver, even during the incidents of violence, in order to protect a
valuable reputation. These findings may also be at play in trade with fragile states, where
some exporters make additional investment to maintain their trade relationships in the midst
of conflict and fragility situations.
Our paper also relates to numerous cross-country studies that estimate gravity type models on
the effect of conflict and fragility on bilateral trade, but results remain mixed. A large number
of studies using gravity-type models of trade find that fragility and conflict in particular,
affects bilateral trade negatively (Pollins 1989, Mansfield and Bronson, 1997, Anderson and
Marcouiller, 2002, Blomberg and Hess, 2006, Martin et al., 2008, and Glick and Taylor,
2010). Yet, another set of papers, also deploying gravity, finds a negative but insignificant
relationship between conflict and bilateral trade (Morrow et al, 1998, Mansfield and
Pavehouse, 2000 and Penubarti and Ward, 2000). In most of these, a dummy variable
indicating the presence of bilateral conflict or destination country fragility is used to capture
the impact of fragility and conflict on trade. This variable is complemented by a set of
traditional gravity variables such as GDP, geographic distance, common border, common
colonial history, common language, preferential trade arrangements among others.
8
Very briefly, we review the findings from some of the macro-based studies. In Pollins (1989),
international commerce is often affected by the vagaries of interstate conflict and
cooperation. They show that trade flows are significantly influenced by broad political
relations of amity and enmity between nations. Similarly, Mansfield and Bronson (1997)
examined the effects of conflict on bilateral trade flows using a modified gravity regression.
Their results show that conflict substantially reduced trade by as much as 6.5 times between
countries in conflict relative to countries not in conflict.
In a different approach, Anderson and Marcouiller (2002) use a structural model to estimate
the impact of corruption and imperfect contract enforcement on international trade. They
found a reduction in import demand with insecurity acting as a hidden tax on trade. Related,
Blomberg and Hess (2006) broaden the concept of conflict and its impact on bilateral trade
by obtaining a synthetic measure of violence through factor analysis that includes terrorism,
external war, revolutions and inter-ethnic fighting. They find that the presence of conflict is
equivalent to a 30% tariff on trade, which is larger than traditional tariff barriers.
Martin et al (2008) use the correlates of war project (COW) data set from the University of
Michigan to analyze the relationship between conflict and trade using a large data set of
military conflicts over the 1950-2000 period. The key hypothesis in their model is that the
absence of peace disrupts trade and therefore puts trade gains at risk. They test this using a
gravity model of trade. Their results show that in both the traditional gravity and the
theoretical gravity by Anderson and Van Wincoop (2003), the impact of bilateral conflicts
was negative and statistically significant. For example during a military conflict, trade fell by
about 22% relative to the traditional gravity predictions.
Related, Glick and Taylor (2010) examined the effects of conflict on bilateral trade for almost
all countries (172) with a large sample of data over the period 1870-1997. Their results
showed that trade was still low by 42% below the pre-conflict level five years after cessation
of conflict and 21% after eight years. In the context of SSA countries, Bussmann et al (2005)
find evidence to support the fact that trade liberalization led to durable peace once the
restructuring of the economy of a fragile state is completed.
A common weakness in cross-country studies is in the level of their analysis of the effect of
fragility, which is based on aggregate trade data. While this conveys useful information on
the average effects of fragility on trade, it conceals the potential for richer adjustments taking
9
place at the firm and product level, such as the entry and exit of exporters from a given
destination and the reduction number of products and volume of shipments due to destination
fragility. Understanding the effect of fragility along these margins may help explain the
ambiguity in results using aggregate data (Crozet et al., 2007).
To summarize, the recent new trade theory models provide invaluable insights into the
behavior of firms breaking into foreign markets. Empirically, the model yields a standard
gravity estimation capturing the fact that large countries in size tend to trade more, while
geographic distance lowers trade the further the trade partners are from each other. Conflict
and fragility is often included in the gravity estimations through the resistance term.
Intuitively, destination fragility raises trade costs (see Martin et al., 2008) but does not affect
all exporters in a way similar to trade tariffs (Crozet et al., 2007).
Our framework differs from Crozet et al (2007) in two important ways. Firstly, we emphasize
on firm productivity as key in explaining firm’s ability to overcome fixed costs of entry in
foreign markets as in Melitz (2003). Crozet et al explain the same concept but focus on
marginal costs-that’s firms with least marginal costs are more likely to select into export
markets. Our approach, therefore, provide an alternative formulation to take into account
market friction related to destination fragility. Secondly, in Crozet et al, the empirical tests
for the prediction of the model is applied to France-a developed country. On the contrary, we
look at Kenya, a country that neighbours some of the most fragile states such as Somalia and
South Sudan. Furthermore, a lot of fragile states are in Africa (Bussmann et al., 2005) and
since we know from the gravity model of trade that neighbouring countries are likely to trade
more, we believe Kenyan firms are more likely to trade with fragile states relative to French
firms.
3.0 Theoretical Framework
3.1 The Model
The model environment is for a world of two countries, home (H) and foreign (F) with
consumers in each country (LH and LF) seeking to maximize utility. Each of the consumers
has one unit of labour and a single share of perfectly diversified portfolio of all firms in the
two country world. Profits earned by firms are repatriated as dividends in terms of a
homogenous good ( 𝑄1) which is also the numeraire.
10
Consumer utility and demand functions
The consumers in both countries share an identical two tier utility function. In the first tier, a
Cobb Douglas function form captures the substitution of consumption between good 1 (the
numeraire) and good 2 (a horizontally differentiated variety). In the second tier, a constant
elasticity of substitution (CES) form represents the substitution of consumption among
varieties of good 2. Consumer utility is given as:
11
1
(1 )1
1
n
i
i
U Q q di
(1.)
where 𝑄1 and 𝑄2 = 1n
i
i
q di
are consumption of good 1 and good 2. 𝜎(> 1) = constant
elasticity of substitution (CES) between varieties of good 2. 𝜇1 and (1 − 𝜇1 ) are shares of
each good and must sum to unit. n = is the number of varieties of good 2.
Consumers in country f ∈ [H F] earn income Yf, of which, a share (1 − μ1 = 𝜇2) is spent on
good 2. From the consumption function of good 2, we can obtain its dual, which yields the
perfect competition price index for the good as:
𝑃𝑓 = [∫ 𝑝𝑖𝑓1−𝜎
𝑛
𝑖
𝑑𝑖]
11−𝜎
(2.)
where 𝑝𝑖𝑓 is the price of variety 𝑖 in country 𝑓 ∈ [𝐻 𝐹]. Total expenditure of good 2 in
country 𝑓 ∈ [𝐻 𝐹] is given by 𝜇2𝑌𝑓𝑃𝑓. Applying Shephard’s lemma on equation (2) yields
individual demand function for varieties of good 2, supplied by firms from country 𝑓 ∈ [𝐻 𝐹]
as (see appendix 2):
𝑞𝑖𝑓 = 𝜇2 𝑝𝑖𝑓−𝜎 𝑌𝑓 𝑃𝑓
𝜎−1 (3.)
Firm technology and profit maximisation
Good 1 is produced using constant returns to scale technology under a perfect competition
market structure. It uses one unit of labour per unit of output and is freely traded across the
countries. Differences in endowments between countries are sufficiently small to ensure that
good 1 is always produced in both countries and as such, the sector establishes the wages in
both countries equal to 1. Good 2 has a continuum of differentiated varieties produced under
increasing returns to scale technology (Melitz, 2003).
11
𝑇𝐶 = 𝐹𝑓𝐷 +
𝑞𝑖𝑓
𝜑𝑖
(4.)
where 𝐹𝑓𝐷 is the fixed costs of entry in the domestic market;
𝑞𝑖𝑓
𝜑𝑖 is the variable cost that
depends wholly on firm’s productivity draw 𝜑𝑖 after paying the fixed costs. Firms then
decide whether or not to export. Exporting good 2 is costly in two ways. Firstly, there is
iceberg trade costs, which implies that 𝜏 > 1 units of goods have to be shipped from home
(H) to ensure that one unit arrives in foreign market (F). Secondly, to enter the export market
𝑓 ∈ [𝐻 𝐹] , each firm must pay a fixed cost (𝐹𝑓𝑥) to export. The fixed costs cover basic
market research, identifying distributors, adjusting to foreign standards, among others. We
assume, for simplicity, that these costs are known with certainty to the potential exporter.
Incorporating additional costs to export to fragile states
Assuming an export destination market is fragile, there are additional costs associated to
serving this market. Risks may include delays in delivery, lost sales proceeds, unhonoured
contracts, among others. Firms must factor these risks in their decision to serve fragile
markets. Each firm has an exogenous probability(1−∝) of being directly affected by fragility
when trying to export to a fragile state (Crozet et al., 2007). The incident of loss due to these
risks affects only a sub set of firms. This sub-set of firms (1−∝) can either pay the additional
costs or avoid exporting to fragile states altogether, foregoing the fixed cost incurred to enter
export market in the first place. Let this additional cost be specified as 𝛽𝐹𝑓𝑋(𝛽 > 0) such
that the total fixed costs to export to a fragile state becomes:
(1 + 𝛽)𝐹𝑓𝑋 (5.)
𝛽 is a proportion of the fixed cost of entry into the destination. The smaller the proportion,
the easier it is for an exporter to pay and serve the destination. We assume that the probability
∝ and the payment 𝛽 are known by firms, ex-ante. This assumption simplify the solution
since we do not have information asymmetry.
Profits and firms-selection
The export market profit function may be specified as:
𝜋𝑓(𝜑𝑖) = 𝑝𝑖𝑓𝑞𝑖𝑓 − 𝜏𝑞𝑖𝑓
𝜑𝑖− 𝐹𝑓
𝐷 − 𝐹𝑓𝑋
(6.)
12
where 𝑞𝑖𝑓 is the demand function defined in equation (3), 𝜑𝑖 is the productivity draw, 𝐹𝑓𝐷
and 𝐹𝑓𝑋 are the fixed costs for entering domestic and foreign markets, respectively.
Maximising (6) with respect to quantity, yields the optimal price as:
𝑝𝑖𝑓 = 𝜏𝜎
𝜎 − 1
1
𝜑𝑖
(7.)
Using equation (7) we can rewrite the profit function for each of the two types of exporters.
Profit obtained by selling at domestic market
We can use Equation (7) and setting 𝜏 =1 to rewrite domestic demand as:
𝑞𝑖𝐻 = 𝜇2 (𝜎
𝜎 − 1
1
𝜑𝑖)
−𝜎
𝑌𝐻 𝑃𝐻𝜎−1
Substitute Equation(8) into the domestic market profit function and re-arrange:
(8.)
𝜋𝐻(𝜑𝑖) = 𝑝𝑖𝐻𝑞𝑖𝐻 −𝑞𝑖𝐻
𝜑𝑖− 𝐹𝐻
𝐷
𝜋𝐷𝐻(𝜑𝑖) =𝜇2
𝜎 𝑌𝐻 𝑃𝐻
𝜎−1 (𝜎
𝜎 − 1)
1−𝜎
(1
𝜑𝑖)
1−𝜎
− 𝐹𝐻𝐷
𝜋𝐷𝐻(𝜑𝑖) ≥ 0 ⟺ 𝜇2
𝜎 𝑌𝐻 𝑃𝐻
𝜎−1 (𝜎
𝜎 − 1)
1−𝜎
(1
𝜑𝑖)
1−𝜎
= 𝐹𝐻𝐷
𝜋𝐷𝐻(𝜑𝑖) ≥ 0 ⟺ 𝜇2
𝜎 𝑌𝐻 𝑃𝐻
𝜎−1 (𝜎
𝜎 − 1)
1−𝜎
(1
𝜑𝑖)
1−𝜎
= 𝐹𝐻𝐷
𝜑𝐷𝐻 = 𝜆1 [𝐹𝐻
𝐷
𝑌𝐻 𝑃𝐻𝜎−1]
1𝜎−1
where: 𝜆1 = [𝜎
𝜇2]
1
𝜎−1 (
𝜎
𝜎−1). Firms with productivity threshold less than 𝜑𝐷𝐻 are not
profitable enough to be active in the domestic market. The domestic productivity threshold
increases with fixed cost of entry and decreases with increase in income and/or increase in
aggregate price index. We now focus on the home country firms (H) shipping to foreign (F)
markets that may be considered fragile. There are two types of exporting firms: the lucky
ones ∝ that do not have to pay the additional fixed costs to serve a fragile destination and the
unlucky ones, a proportion (1−∝) that must pay additional costs to serve a fragile state.
Exporters that never pay additional costs
There is a proportion (∝) of exporters (𝑋) from home (𝐻) who export to a fragile state
without payment of additional costs. Their profits are given as:
𝜋𝑋𝐻(𝜑𝑖) = 𝑝𝑖𝑓𝑞𝑖𝑓 −𝑞𝑖𝑓
𝜑𝑖− 𝐹𝐻
𝐷 − 𝐹𝐻𝑋
(9.)
13
𝜋𝑋𝐻(𝜑𝑖) = (𝜑𝑖𝑝𝑖𝑓 − 1)𝑞𝑖𝑓
𝜑𝑖− 𝐹𝐻
𝐷 − 𝐹𝐻𝑋
Substituting for 𝑝𝑖𝑓 and 𝑞𝑖𝑓 using equations (3) and (7) and re-arranging we obtain4:
𝜋𝑋𝐻 (𝜑𝑖) =𝜇2
𝜎 (𝑌𝐻 𝑃𝐻
𝜎−1 + 𝜏1−𝜎 𝑌𝐹 𝑃𝐹𝜎−1) (
𝜎
𝜎 − 1)
1−𝜎
(1
𝜑𝑖)
1−𝜎
− 𝐹𝐻𝐷 − 𝐹𝐻
𝑋
(10.)
Equation (10) expresses the exporter’s profitability as a function of the share of expenditure
on good two, market size, iceberg trade costs, the price of good two and the sunk entry costs
to both domestic and foreign markets.
Exporters that pay additional costs due to fragility
There is a proportion (1−∝) of exporters (𝑋) from home (𝐻) who export to a fragile state
but must pay additional costs (unlucky exporters). Their profits are given as:
𝜋𝑋𝐻(𝜑𝑖) = 𝜋𝑋𝐻 (𝜑𝑖) − 𝛽𝐹𝐻𝑋
(11.)
where 𝛽 > 0. We assume firms are risk neutral and they decide to make their first
irreversible sunk cost 𝐹𝐻𝑋 if expected profit from exporting is greater than profit obtained at
the domestic market ( see appendix 2).
(1 − 𝛼)𝜋𝑋𝐻(𝜑𝑖) + 𝛼𝜋𝑋𝐻(𝜑𝑖) ≥ 𝜋𝐷𝐻(𝜑𝑖)
⇔ (1 − 𝛼)(𝜋𝑋𝐻(𝜑𝑖) − 𝛽𝐹𝐻𝑋) + 𝛼𝜋𝑋𝐻(𝜑𝑖) ≥ 𝜋𝐷𝐻(𝜑𝑖)
(12.)
𝜑𝑖 ≥ 𝜑𝑋𝐻 = 𝜆1 [𝐹𝐻
𝑋 + (1 − 𝛼)𝛽𝐹𝐻𝑋
𝑌𝐹]
1𝜎−1 𝜏
𝑃𝐹
All firms from (𝐻) with a productivity threshold greater than 𝜑𝑋𝐻 try to export. The
productivity threshold is increasing in the fixed costs, the probability of being unlucky and
the additional payments to serve a fragile market. It is also increasing in the iceberg trade
costs and decreasing in the aggregate price level and in the foreign market GDP. Taking the
derivative of Equation (12) with respect to 𝛽 and ∝ we get first testable hypothesis (see
appendix 2).
4 The same holds for exporters from the foreign market (F) to home market (H).
𝜕(𝜑𝑋𝐻
)
𝜕𝛽=
𝜕(𝜑𝑋𝐻
)
𝜕𝑈 𝜕𝑈
𝜕𝛽> 0
(13.)
14
𝜕(𝜑𝑋𝐻
)
𝜕 ∝=
𝜕(𝜑𝑋𝐻
)
𝜕𝑈 𝜕𝑈
𝜕 ∝< 0
Equation (13) shows that an increase in destination fragility increases the cost of serving that
market and discourages entry. This leads to the following hypothesis:
H1: An increase in destination country fragility has a negative effect on the decision to serve
that market with exports.
On the other hand, an increase in the share of lucky exporters that are able to serve a fragile
state without having to pay additional costs is associated with a higher probability of
maintaining an export relationship to that destination. We expect both 𝛽 and (1−∝) to be
positively correlated. The hard assumption implicit in this model is that there is no possibility
that 𝛽 and (1−∝) can be negatively correlated. Thus if 𝛽 > 0 then an increase in destination
fragility is correlated with low expected profit from exporting, which implies it is also true
that (1−∝) is negatively correlated with expected profit5.
Next, the volume of exports between the home country and a foreign fragile country depends
on the number of exporters and the average value of shipment per firm.
𝑋𝐻𝐹 = 𝑁𝐻𝐹 𝑝𝑖𝐻 𝑞𝑖𝐻
(14.)
where 𝑁𝐻𝐹 is the total number of exporters from a home country to a foreign country. At
equilibrium this is shown to be equal to 𝑁𝐻𝐹 = 𝛼𝐿𝐻 ∫ (1
𝜑)
1−𝜎𝑑
𝜑𝑋𝐻
0𝐺(𝜑)𝑑𝜑 + 𝐿𝐻 ∫ (
1
𝜑)
1−𝜎𝑑𝐺(𝜑)
1
𝜑𝑋𝐻 𝑑𝜑
i.e. the sum of all exporters from home serving destination 𝑗. 𝐺(𝜑)=(𝜑
��)
𝜌 is Pareto distributed
which is consistent with the empirical distribution of firm productivity in most economies
(see Melitz, 2003 for similar assumption). Productivity is normalized to 1( 0 < 𝜑 < 1 ) and 𝜌 =
1 − 𝜎 . 𝑝𝑖𝐻 is as defined in Equation (7) and 𝑞𝑖𝐻 is defined in Equation (3). The effect of
fragility is captured indirectly through 𝜑 and it can be shown that (see appendix 2):
𝜕(𝑋𝐻𝐹)
𝜕𝛽=
𝜕(𝑋𝐻𝐹)
𝜕𝑁𝐻𝐹
𝜕𝑁𝐻𝐹
𝜕𝛽< 0
𝜕(𝑋𝐻𝐹)
𝜕 ∝=
𝜕(𝑋𝐻𝐹)
𝜕𝑁𝐻𝐹 𝜕𝑁𝐻𝐹
𝜕 ∝> 0
(15.)
5 A scenario where 𝛽 < 0 will suggest a bonus in profits if a firm exports to a fragile state. Mathematically this is a possibility
but is not the focus of this paper.
15
An increase in destination country fragility has a negative relationship with the value of
exports to that destination. This drop in export flow is an outcome of reduced number of
exporters/products to a fragile destination. This leads us to the following hypothesis:
H2: An increase in destination country fragility has a negative effect on the home country’s
extensive margins of trade.
3.2 Estimation strategy
To test hypothesis (1) empirically, we regress the dummy variable 𝑒𝑥𝑝𝑖𝑗=1, if firm 𝑖
maintains an export status to a destination 𝑗, conditional on destination country fragility, firm
attributes and destination market characteristics. The average marginal effects of 𝛽 and 𝜑𝑖
can be estimated using a binary choice model as:
𝑃𝑟[𝐸𝑥𝑝𝑖𝑗𝑡 = 1|𝐗] = 𝛽0 + 𝛽1 𝑓𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗𝑡 + 𝛽2 ln(𝑆𝑖𝑧𝑒𝑖) + 𝛽3ln (𝑆𝑖𝑧𝑒𝑖) ∗
𝑓𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗𝑡 + (𝑿𝒋𝒕′𝜷) + 𝜸𝜹𝒕 + 𝜸𝜹𝒊 + 𝜸𝜹𝒋 + 𝜀𝑖𝑗𝑡
(16.)
The main explanatory variables include destination fragility (-), firm size (+), the interaction
of (Sizei) ∗ fragilityjt (+) and the usual gravity type variables 𝑿𝒋𝒕′ such as real GDP and
distance. We include year (𝜹𝒕), firm (𝜹𝒊) and destination country (𝜹𝒋)fixed effects. The time
effects controls for the time varying regressors that are common across firms but not
controlled for, while the firm and destination country fixed effects controls for time invariant
factors that affects the probability of serving destination j but not controlled for. 𝛽1 is
expected to be negative capturing the negative effects of both 𝛽 and (1−∝) in our theoretical
model. We estimate equation (16) over the sample period 2004-2013.
The role of firm network
The literature examining firm level trade to markets with uncertainty and information
asymmetry underscores the role of firm network in overcoming challenges to serve risky
markets. The definition of firm’s network and export spillovers in this paper follows that of
Cadot et al (2013) and does not include a region bounded definition as in Greenaway and
Kneller (2008). This is because we do not have geographic location of firms in our data.
Cadot et al (2013) find that if a large number of firms exported similar products to the same
16
destination, this gave rise to positive signalling on profitability of that product, which
enhanced survival of firms in foreign markets.
We constructed three different proxy measures for firm network. The first measure is the
average number of firms exporting similar products (HS6) to the same destination. This
variable constitute a narrow measure of firm’s network. Firm’s exporting similar products to
the same destination can be considered competitors. At the same time, there is a lot of
signalling effect and learning from the fortunes of other exporters, including potential for new
profit (Cadot et al., 2013, Eaton et al., 2009). A second proxy is the average number of firms
in a sector (HS2), exporting to the same destination. This variable is expected to capture
foreign market intra-industry spillovers (Choquette and Meinen, 2015). A third proxy is the
average number of firm-product to a given destination. This is used in the literature to capture
foreign market specific inter industry spillovers (Cadot et al., 2013, Choquette and Meinen,
2015). By the nature of construction, these variables are highly correlated and log
transformation of each is included in a separate regression. We estimate an augmented variant
of equation (16) as follows:
𝑃𝑟[𝐸𝑥𝑝𝑖𝑗𝑡 = 1|𝐗] = 𝛼0+𝛼1𝑓𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗𝑡+𝛼2 𝑓𝑖𝑟𝑚𝑛𝑒𝑡𝑤𝑜𝑟𝑘 + 𝛼3 ln (𝑆𝑖𝑧𝑒𝑖) +
𝛼4 (𝑓𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗𝑡 ∗ 𝑓𝑖𝑟𝑚𝑛𝑒𝑡𝑤𝑜𝑟𝑘) + 𝛼5(𝑓𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗𝑡 ∗ 𝑆𝑖𝑧𝑒𝑖) + (𝑿𝒋𝒕′𝜷) + 𝜗𝑖𝑗𝑡
(17.)
Expijt is as defined earlier. The additional explanatory variable is 𝑓𝑖𝑟𝑚𝑛𝑒𝑡𝑤𝑜𝑟𝑘 and its
interaction with destination country 𝑓𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗𝑡 . We expect the sign on 𝛼2 𝑎𝑛𝑑 𝛼4 to be
positive and statistically significant. The coefficient on the interaction term 𝑓𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗𝑡 ∗
𝑓𝑖𝑟𝑚𝑛𝑒𝑡𝑤𝑜𝑟𝑘 capture the fact that network effects help in overcoming hurdles in exporting to fragile
states. We also include year (𝜹𝒕), firm (𝜹𝒊) and destination country (𝜹𝒋) effects.
Finally, to test the second hypothesis we estimate the following panel regression.
𝐸𝑥𝑝 𝑗𝑡 = 𝛽0 + 𝛽1 𝑓𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗𝑡 + (𝑿′𝒋𝒕 𝜷) + 𝜸𝜹𝒕 + 𝜸𝜹𝒋 + 𝜗𝑗𝑡
(18.)
The dependent variable is the extensive and intensive margins of trade as proposed by
Bernard et al (2007, 2014) from Kenya to each destination j over the period 2004-2013. The
main explanatory variable is the destination country fragility. The vector 𝑿′ contains the
gravity variables such as market size and distance. We expect 𝛽1 to be negative and
17
statistically significant. We also expect the sign on destination country GDP to be positive,
while distance is expected to have a negative sign. We estimate equation (18) using the
Pseudo Poisson Maximum Likelihood (PPML) proposed by Silva and Tenreyro (2006).
3.3 Data and Data sources
Transaction level dataset
The transaction level dataset is obtained from the Customs Services Department of the Kenya
Revenue Authority (KRA) for the period 2004 to 2013. Each transaction contains information
on the product being exported at the 8- digit HS product level, the destination of shipment,
the free on board (FOB) value in Kenya shillings, and the exporter identifier. The transaction
dataset allows us to obtain proxies for firm trade characteristics such as the firm level exports
to a destination; the size measured in terms of the initial value of a firm’s export in the first
year of entry to exporting, the number of firms exporting to a destination, and the number of
firms exporting similar products to the same destination. We utilise the information on
destination countries to examine the choice of destinations for exporters and how this choice
is affected by destination country fragility.
Destination country fragility
Destination country fragility is multi-dimensional concept and complex to measure with
precision. As pointed out in the introduction, we follow the World Bank (2011) to define
fragility as periods when states or institutions lack the capacity, accountability or legitimacy
to mediate relations between citizen groups and between citizens and the state, making them
vulnerable to violence. From this definition, we require objective measure of, “periods when
states lack the capacity, accountability or legitimacy to mediate elations”. One of the popular
set of indicators that closely reflect this definition is the Worldwide Governance Indicators
(WGI), compiled by Kaufmann et al (2011).
Kaufmann et al (2011, p.222) defines governance as “the traditions and institutions by which
authority in a country is exercised. This includes (a) the process by which governments are
selected, monitored and replaced; (b) the capacity of the government to effectively formulate
and implement sound policies; and (c) the respect of citizens and the state for the institutions
that govern economic and social interactions among them.” Within each component, the
authors develop two indicators that measure that aspect. For example, the process by which
18
governments are selected, monitored and replaced is captured by voice and accountability
and political stability and absence of violence/terrorism. Voice and accountability captures
perceptions of the extent to which a country’s citizens are able to participate in selecting their
government, freedom of expression and free media, while political stability and absence of
violence/terrorism captures perceptions of the likelihood that the government will be
destabilized or overthrown by violent means.
The WGI database provide annual country ranking on each of the six indicators. Each
indicator is measured on a scale of -2.5 to 2.5 (worst to best performers) and they are highly
correlated. The advantage with the WGI dataset is that it provides us with a wide coverage
for all African countries. The available time series is also able to match the sample period in
the transaction data, allowing us to make use of the full panel estimations. The indicators also
broadly capture the definition of fragility situations espoused in the World Bank (2011).
We follow Blomberg and Hess (2006) and use the Principal Component Analysis (PCA)
method to obtain an average governance indicator from the underlying six indicators of
governance. Of the six indicators, we needed judgement on the indicators that most closely
reflect the multi-dimensional concept of fragility defined in this paper and the WDR (2011).
Consequently we created three average proxy measures for destination fragility using PCA.
The first average index include voice and accountability, political stability and rule of law as
inputs into the PCA to generate an average political risk index. This index is our preferred
measure and is used in the estimated equations. We are of the view that these indictors
capture “periods when states lack the capacity, accountability or legitimacy to mediate
relations” (World Bank, 2011). The second average index include regulatory quality,
government effectiveness and control of corruption. This index is viewed as closely reflecting
business environment risks in the destination country. Finally, the third average indicator is
obtained as an amalgam of all the WGI six indicators. This index is used largely in the
descriptive, presented in the paper. All the indicators perform well as proxies for destination
fragility in the estimation of our empirical results.
How do we interpret this average governance indexes obtained through PCA? An
improvement in the average score in the governance index implies a decline in fragility for a
given country. As such the index enters in the equation with a negative for ease of
19
interpretation. Table 10 in the appendix 1A contains the average fragility index for all
African countries in the sample. It also provides the average share of exports from Kenya to
each of these countries. We provide brief details in appendix 1B on how the PCA method is
used to obtain the average governance index. The measurement of other key variables is
contained in the appendix. Table 1 contains the summary statistics for the variables of interest.
Table 1: Summary statistics for the key variables
Source: Customs dataset, WDI and CEPII database
3.4 Stylized facts
In this section we present observable features in our dataset on the relationship between
exporters’ trade activity and destination country fragility. We start off by exploring the
variable N mean sd min max
Exp_ijt>0 dummy 4,345,836 0.01 0.12 0.00 1.00
Fragility Index 4,345,836 0.67 0.62 -0.91 1.83
Log(Initial fobvalue) 4,345,836 9.60 2.09 -0.03 18.36
Log(avg.nrfirms_sameHS6 to j) 4,345,836 0.03 0.24 0.00 3.20
Log(avg.nrfirms_same_sector to j) 4,345,836 0.07 0.55 0.00 5.48
Log(avg.nrfirm-product to j) 4,345,836 0.11 0.94 0.00 9.65
Log(gdp_cons) 4,345,836 22.70 1.55 18.74 26.50
Log( real exchange rate) 4,345,836 -0.17 2.30 -5.43 4.62
Log(days_to_import in j) 4,345,836 3.56 0.48 2.20 4.62
Log( Distance) 4,345,836 7.97 0.63 5.66 8.84
Common border dummy 4,345,836 0.08 0.28 0.00 1.00
Common language dummy 4,345,836 0.43 0.50 0.00 1.00
Common colonial history 4,345,836 0.33 0.47 0.00 1.00
20
relative differences in fragility across countries and changes in fragility within countries over
time. Figure 1 shows the ranking of the top 15 most fragile countries, together with their
fragility index in 2005 and 2012.
Figure 1: Ranking of African countries by their fragility index, 2005 and 2012
Notes: ZAR: DR Congo, ZWE: Zimbabwe, SDN: Sudan, CIV: Cote d’Ivoire, GNQ: Equatorial Guinea, TCD:
Chad, CAF: Central African, AGO: Angola, LBR: Liberia, TGO: Togo, BDI: Burundi, GIN: Guinea, ERI:
Eritrea, NGA: Nigeria, ETH: Ethiopia, CMR: Cameroon and COM: Comoros. Fragility index is computed using
PCA methods from the WGI compiled by Kaufmann et al (2011) and ranges between -0.9 (less fragile) to 1.8
(most fragile).
We observe that the top seven fragile states in 2005[DR Congo (ZAR), Zimbabwe (ZWE),
Sudan (SDN), Cote d’Ivoire (CIV), Equatorial Guinea (GNQ), Chad (TCD) and Central
African Republic (CAR)] account for the top 5 slots in the ranking of fragile states in 2012 [
DR Congo, Sudan, Zimbabwe, and CAR]. This implies that changes in the relative ranking of
countries is small among the top 15 fragile states. Within countries, however, there is more
variation in fragility score over time. For example in Zimbabwe the fragility index decreased
from 1.70 in 2005 to 1.48 in 2012 (12.9%), while it rose in Sudan from 1.69 in 2005 to 1.72
in 2012 (2%). In Cote d'Ivoire the index dropped significantly from 1.53 in 2005 to 1.05 in
2012 (31.4%). Rwanda’s fragility index dropped by 80% from 1.0 in 2005 to 0.20 in
2012.Table 2 presents Kenya’s top export destinations in Africa and each destination
country’s fragility index, in 2005 and 2012.
21
Table 2: Top 15 export destinations and the respective fragility index, 2005 and 2012
Notes: The share of exports is calculated as a ratio of Kenya’s exports to destination j relative to its total exports
to Africa in 2005 and 2012. Fragility index is computed using PCA methods from the WGI compiled by
Kaufmann et al (2011) and ranges between -0.9 (less fragile) to 1.8 (most fragile). The value of exports to
Africa is US$ 1,130 million and US$ 2,400 million, respectively for 2005 and 2012.
It can be observed that the top 15 destinations account for over 97% of Kenya’s exports to
Africa in 2005 and 2012. Uganda, Tanzania, Rwanda, and Burundi accounts for
approximately 59% of Kenya’s exports to Africa, which is an indicator of the importance of
the EAC trade block. Among top 5 export destination in 2005, DR Congo and Sudan had
high fragility index of 1.77 and 1.70, respectively. These positions are maintained in 2012.
A direct way to observe how the trade margins change with the changes in the fragility index
of countries is to explore the time dimension. This can be done using both pooled average
exports and fragility over time and country specific time plots. We start off with a plot of the
average exports relative to average fragility over time. Figure 2 shows a time plot of the
average export value across all exporters over time relative to the mean destination fragility
over time. It can be observed that in years when average fragility drops, the average exports
across countries increased. Similarly, in years when the average fragility rose, the average
exports across countries dropped. Thus it is clear to observe a negative relationship between
exports and levels of fragility along the time dimension.
Rank Country share of export Fragility index Country share of export Fragility index
1 Uganda 0.31 0.71 Uganda 0.28 0.62
2 Tanzania 0.20 0.46 Tanzania 0.19 0.48
3 Egypt 0.10 0.50 Sudan 0.11 1.73
4 DR Congo 0.08 1.77 Egypt 0.10 0.79
5 Sudan 0.06 1.70 DR Congo 0.08 1.74
6 Rwanda 0.05 1.00 Rwanda 0.07 0.20
7 Zambia 0.03 0.62 Zambia 0.03 0.24
8 Ethiopia 0.03 1.15 Burundi 0.02 1.33
9 Burundi 0.03 1.23 Malawi 0.02 0.39
10 South Africa 0.03 -0.46 Ethiopia 0.02 0.98
11 Malawi 0.02 0.46 Nigeria 0.01 1.21
12 Nigeria 0.01 1.19 South Africa 0.01 -0.22
13 Eritrea 0.01 1.20 Zimbabwe 0.01 1.48
14 Mozambique 0.01 0.41 Mauritius 0.01 -0.91
15 Mauritius 0.01 -0.78 Djibout 0.01 0.72
0.97 0.97
2005 2012
22
Figure 2: Average exports and mean destination fragility over time
Notes: Left hand y-axis if the mean (log export value) across all destinations and over time. On the right hand
side is the mean destination fragility index over time.
We also plotted the share of exports from Kenya to Rwanda and Madagascar that have
experienced opposite shock in their fragility index. In Rwanda fragility index has dropped
while in Madagascar, the index has increased over time. The negative relationship between
destination fragility and exports is also apparent.
Figure 3: Share of Kenya’s total exports to Rwanda and Madagascar, 2004-2013
Notes: Y-axis if the share of Kenya’s exports to Rwanda and Madagascar. On the right hand side is the fragility
index for the respective countries. Fragility index is computed using PCA methods from the WGI compiled by
Kaufmann et al (2011) and ranges between -0.9 (less fragile) to 1.8 (most fragile).
In summary, fragility has a negative effect on country level exports. This effect can be
identified more precisely along the time dimension, rather than across countries.
Identification across countries is further complicated by potential omitted variables. In the
next section we present econometrics results for our empirical models.
23
4.0 Empirical Results
4.1 Destination fragility and firm’s export status
We estimate the effects of destination country fragility on the probability of a firm serving a
given destination over the sample period 2004-2013. Table 3 shows the results.
Table 3: Firm’s export status and destination country fragility
Notes: The dependent variable is a dummy equal to 1, if an exporter exports to destination j in period t. The
regressions are based on observations measured at the firm-destination-year for the years 2004-2013.The main
explanatory variable is the measure of destination country fragility. Odds ratio estimated after as an exponent of
the coefficient on logit. *Other gravity controls include distance, distance squared, common border, common
language and common colony. Asterisk denotes level of significance (***p<0.01, **p<0.05, and *p<0.1).
Robust standard errors in brackets.
Column (1-3) results are obtained using a linear probability model (LPM). Column (1)
controls for time dummy and firm fixed effects while column (2) controls for time,
destination country dummy and firm fixed effects. Column (3-4) controls for time and firm-
country fixed effects. We are aware that the LPM may lead to predicted probabilities that are
not bound within a unit interval (Soderborm and Teal, 2014 p. 230). This problem is not
relevant if all explanatory variables are discrete rather than continuous. However, our
specification includes variables that are continuous. The LPM estimation also implies that
heteroscedasticity exists (Horrace and Oaxaca, 2006). But this latter problem is usually
corrected by making use of heteroscedasticity robust standard errors.
Dependent variable: Firm’s export status to country j (exp_ijt>0)
(1) (2) (3) (4) (5)
LPM LPM LPM LOGIT Odds ratio
L.(Fragility) 0.0018*** -0.0024*** -0.0120** -0.0795** 0.9236**
(0.0001) (0.0003) (0.0058) (0.0404) (0.0373)
L.Log (gdp_cons) 0.0036*** 0.0042*** 0.0516*** 0.3394*** 1.404***
(0.0000) (0.0005) (0.0154) (0.1013) (0.1423)
L.Log(real exchange rate) 0.0001*** 0.0027*** 0.0075 0.0317 1.0322
(0.0000) (0.0008) (0.0098) (0.0640) (0.0661)
L.Log(days_import) 0.0031*** -0.0099*** -0.0288*** -0.1640*** 0.8488***
(0.0001) (0.0007) (0.0056) (0.0368) (0.0312)
Observations 3,645,474 3,645,474 203,980 192,813 166,232
Number of groups 8,979 8,979 22,750 21,441 21,441
R-squared 0.1148 0.1272 0.2598 -- --
Year dummy YES YES YES YES YES
Firm FE YES YES NO NO NO
Country dummy NO YES NO NO NO
Firm-Country FE NO NO YES YES YES
Other gravity controls* YES YES -- -- --
F-Statistics/LR Chi2(12) 3376 3376 88.70 1124 1124
24
Despite these well-known challenges with the LPM, the method remain common in the
literature especially because of its ease to interpret and when there are complications with
non-linear models, such as the use of panel with fixed effects (Klaasen and Magnus, 2001,
Horrace and Oaxaca, 2006). Horrace and Oaxaca show that the LPM results are consistent if
the proportion of predicted probabilities that lie outside the unit interval is small. In our
estimation, the proportion of predicted probabilities that lies outside the unit interval is in fact
zero. We also present for robustness the results from a panel logit with fixed effects in
column (4).While this method ensures that predicted probabilities falls within the unit
interval, the computation of the average marginal effects is not trivial because the firm-
country fixed effects are not estimated (Cameron and Trevedi, 2010, P.630). As such, column
(5) computes the odds ratio6 after panel logit with firm-country fixed effects (Barrera-Gomez
and Basagana, 2015) and we interpret our results using the LPM in column (3).
The key hypothesis in equation (16) is that an increase in the destination country fragility is
associated with a reduction in the probability that a firm has positive exports to that country.
The true coefficient on fragility is expected to be negative and statistically significant.
Column (1) results identify for the within a firm and across destinations, the effect of fragility
on a firm’s export status. The sign on fragility is positive, which suggests that there are time
invariant country characteristics that bias the estimated coefficient. Column (2) adds
destination country dummy to account for the omitted variables and the sign on fragility
changes to negative and is statistically significant at 1% level.
Column (3-4) results includes firm-country fixed effects, such that the effect of destination
country fragility on export status is identified entirely along the time dimension. Column (3)
shows that an increase in destination fragility by one unit is associated with a reduction in the
probability to serve a given destination by 1.2 percentage points. This result is also confirmed
in column (5) using the odds ratio (OR) after panel logit with fixed effects. The OR on
fragility shows that a one unit increase in destination fragility is on average associated with
the odds ratio being multiplied by a factor of 0.9236, implying the odds for exporting to a
fragile state drops by approximately 7.6%.
6 Odds ratio and logit coefficients are related. 𝐿𝑜𝑔𝑖𝑡 (
P
1−𝑃) = log(𝑜𝑑𝑑𝑠 𝑟𝑎𝑡𝑖𝑜) 𝑎𝑛𝑑 𝑂𝑅 = 𝑒𝑎+𝛽𝑋. This means the odds ratio
can be obtained by simply taking the exponent of the coefficient from a logit regression.
25
In reality, a one unit increase in fragility is very large given the scale of measurement of this
variable. The scale ranges between -0.9 (less fragile) to 1.8 (very fragile) with the mean
fragility index of 0.67 (see Table 1). This implies, the effect of fragility on firm’s decision to
serve a given market with exports is economically small for the firms in the sample. This
small magnitude is not unique to us. Comparing our results to that found by Crozet et al
(2007) using the within estimation for the French firms, a unit increase in the destination
country’s insecurity index was associated with a reduction in the probability that a French
firm maintained export to that destination by 5%.
4.2 Firm’s size and exports to fragile states
In line with the theoretical framework of this paper, we suspect that large firms are more
likely to afford additional costs to export into fragile states rather than smaller ones. Size is
expected to be positively correlated with export status (Bernard and Jensen, 1999, Chaney,
2008). Table 4 shows the results in which we only control for the firm-destination fixed
effects and time fixed effects.
Table 4: Firm’s size and export status to a given country
Notes: The dependent variable is a dummy equal to 1, if an exporter exports to destination j in period t. The
regressions are based on observations measured at the firm-destination-year for the years 2004-2013.The main
explanatory variable is the measure of destination country fragility and its interaction with size. Odds ratio
estimated after panel logit. Asterisk denotes level of significance (***p<0.01, **p<0.05, and *p<0.1). Robust
standard errors in brackets.
Dependent variable: Firm’s export status to country j (exp_ijt>0)
(1) (2) (3)
LPM LOGIT Odds ratio
L.(Fragility) -0.0823*** -0.6674*** 0.5130***
(0.0213) (0.1475) (0.0757)
(L.Fragility)(logfob_ini) 0.0066*** 0.0543*** 1.0557***
(0.0020) (0.0131) (0.1379)
L.Log (gdp_cons) 0.0525*** 0.3448*** 1.4117***
(0.0154) (0.1013) (0.1430)
L.Log(real exchange rate) 0.0092 0.0475 1.0486
(0.0098) (0.0641) (0.0672)
L.Log(days_import) -0.0269*** -0.1479*** 0.8626***
(0.0056) (0.0371) (0.0319)
Observations 203,980 192,813 192,813
Number of groups 22,750 21,441 21,441
R-squared 0.2598
Year dummy YES YES YES
Firm-Country FE YES YES YES
F-Statistics/LR Chi2(13) 83.48 1141 1141
26
The results in column (1) shows that the sign on destination fragility remains negative and
statistically significant. More importantly, the magnitude of the coefficient is now 8 times
larger relative to the baseline estimation in Table 3. An increase in destination country
fragility by one unit is associated with a reduction in the probability of exporting to that
destination by 8.2 percentage points, holding all other factors constant.
The coefficient on interaction term of size with fragility is positive and statistically
significant at 1%. An increase in size of a firm by 10% is associated with an increase in the
probability that a firm exports to a destination by 4.4% (=0.066x0.67) holding everything else
constant. This percentage points are calculated at an average fragility value of 0.67 across all
countries in the sample. Finally, gravity variables such as real GDP, real exchange rate and
number of days taken to import a container in the destination country have the expected signs
and most are statistically significant
4.3 Firm’s network and exporting to fragile states
Next, we examine the role of firm’s network in mediating the challenges associated with
exporting to fragile states. Exporting to fragile states is costly. We speculate in line with the
existing literature (Aiteken et al.,1997, Cadot et al., 2013, and Greenaway and Kneller,2008)
that firm’s network provide positive spillovers that enables new entrants to overcome
additional costs associated with serving fragile markets. A firm’s participation in export
market may reduce the costs of entry for the followers (Albornoz, et al., 2012, Eaton et al.,
2009). Table 5 shows the results.
The coefficient on destination fragile remains negative and statistically significant at 1%
level, across all estimations. Firm’s network is in general associated with an increase in the
probability to export to a given destination. Using the narrow measure of firm’s network, a
10% increase in the average number of firms exporting similar HS6 products to the same
destination, is associated with an increase in the probability of exporting to a given
destination by 0.7 percentage points, holding all factors constant. The interaction of this
variable with fragility weakens the effect, such that a 10% increase in the average number of
firms exporting similar products to the same destination is associated with 0.6 percentage
points (=0.0705-0.0065*0.67) increase in the probability of exporting to a fragile destination.
This contrary to what we hoped for, although we suspect this result is driven by the level at
which we measure firm’s network and export spillovers.
27
Table 5: Firm network and exports to a given destination
Notes: The dependent variable is a dummy equal to 1, if an exporter exports to destination j in period t. The
regressions are based on observations measured at the firm-destination-year for the years 2004-2013.The main
explanatory variable is the measure of destination country fragility, proxy for firm networks and the interaction
with fragility. Asterisk denotes level of significance (***p<0.01, **p<0.05, and *p<0.1). Robust standard errors
in brackets.
In column (2) firm’s network is measured as the number of firms, belonging to same sector
(HS2) and exporting to the same destination (see Requena-Silvente and Gimenez, 2007). A
10% increase in this variable is associated with an increase in the probability that a firm will
export to a given destination by 0.3 percentage points. Furthermore, this results show that
destination specific intra-industry spillovers helps in overcoming the costs associated with an
increase in the probability of exporting to fragile states (positive interaction term), although
this effect is not statistically significant.
Finally, column (3) considers the effect of firm’s broad network measure as the average
number of firm-product to the same destination. This variable is used as a proxy measure for
Dependent variable:
(1) (2) (3)
LPM LPM LPM
L.Fragility -0.0865*** -0.0808*** -0.0812***
(0.0210) (0.0209) (0.0209)
(L.Fragility)(Logfob_ini) 0.0068*** 0.0062*** 0.0064***
(0.0019) (0.0019) (0.0019)
L.Log(avgnrfirms_sameHS6) 0.0705***
(0.0015)
(L.Fragility)(L.Log(avgnrfirms_sameHS6) -0.0065***
(0.0012)
L.Log(avgnrfirms_same sector) 0.0303***
(0.0006)
(L.Fragility)(L.Log(avgnrfirms_same sector) 0.0008
(0.0006)
L.Log(avgnrfirm-product) 0.0179***
(0.0004)
(L.Fragility)L.Log(avgnrfirm-product) 0.00004
(0.0003)
L.Log (gdp_cons) 0.0284* 0.0324** 0.0329**
(0.0152) (0.0152) (0.0152)
L.Log(real exchange rate) -0.0029 -0.0008 0.0010
(0.0097) (0.0097) (0.0097)
L.Log(days_import) -0.0105* -0.0147*** -0.0151***
(0.0055) (0.0055) (0.0055)
Observations 203,980 203,980 203,980
Number of groups 22,751 22,751 22,751
R-squared 0.2727 0.2733 0.2735
Year FE YES YES YES
Firm-Country FE YES YES YES
Firm’s export status to country j (exp_ijt>0)
28
foreign market inter-industry spillovers (Choquette and Meinen, 2015). A 10% increase in the
number of firm-product lines to a given destination is associated with an increase in
probability for exporting to that destination 0.2 percentage points. The interaction term with
fragility is also positive but not statistically significant.
4.4 Destination fragility and Trade margins
In this section we examine the relationship between destination country fragility and Kenya’s
export trade margins. To decompose exports to a given destination into trade margins, we
follow a method proposed by Bernard et al (2007) and (2014). This method breaks down
exports into extensive margin (the number of firms and the number of products exported to a
market) and the intensive margin (the average exports per firm-product). We conducted this
decomposition (see appendix) and the results reveal a significant role of the extensive margin
in the variation of Kenya’s export across countries. For example, in 2012, the extensive
margin accounted for 53% of the variation in cross-country exports, while the intensive
margin accounted for approximately 47% of the variation.
We estimate Equation (18) using the Pseudo Poisson Maximum Likelihood (PPML)
proposed by Silva and Tenreyro (2006). All regressions include year and destination country
dummies. The PPML method is able to account for the zero trade flows that is frequent in the
trade matrix (Haveman and Hummels, 2004). Dropping zero observations, as is the case
when log-linear models are used, leads to sample selection bias. In addition, the PPML
provides a consistent estimate of non-linear gravity model and includes country fixed effects.
This is consistent with the recent theoretical gravity models that requires inclusion of fixed
effects by exporter and by importer (Anderson and Van Wincoop, 2003).
According to Silva and Tenreyro (2006), the PPML estimator is consistent and can do well in
a variety of circumstances. For instance, the estimator does not assume normality of the
residuals and it is valid with general forms of heteroscedasticity. A key concern in the
literature7 is on the likelihood that exports and fragility are endogenously related. That is,
trade may depend on fragility, but the occurrence of fragility may also depend on trade
7 A large body of literature in political science addresses the question of how the probability of fragility and conflict depends
on various measures of economic interdependence, including trade openness (see Barbieri, 2002, Mansfield and Pavehouse,
2000).
29
dependence between countries. Our estimation uses destination country fragility and per
capita GDP with a lag of one period. This strategy cannot eliminate endogeneity concerns
totally but by using key explanatory variables pre-determined in the previous period may
reduce this bias (Glick and Taylor, 2010). Table 6 shows the regression results.
Table 6: Destination fragility and trade margins to destination j
Notes: The dependent variables: Column (1-4) number of firms, number of products, average shipment and total
exports to a destination country j. All regressions are from PPML estimation as proposed by Silva and Tenreyro
(2006). We control for the year and destination country dummy. Asterisk denotes level of significance
(***p<0.01, **p<0.05, and *p<0.1), robust standard errors in brackets.
The PPML is an exponential model, which means we can interpret the effect of explanatory
variables in a straight forward way, just as in OLS. Firstly, all the dependent variables are in
levels rather than in logs, which means we retain the zeros in the estimation and secondly,
coefficients of explanatory variables that enter in levels are interpreted as semi-elasticities,
while coefficients of explanatory variables that are in logs are interpreted as simple
elasticities. The results in column (1) shows that the effect of destination fragility on the
number of firms is negative and statistically significant at 5% level. A 1 unit increase in
fragility is associated with a reduction in the number of firms exporting to that destination by
29%, which is in line with the theoretical prediction of our model, in which the effect of
Dependent variables: Firms Products Av.value Tot.value
(1) (2) (3) (4)
L.Fragility -0.2984** -0.3486*** 0.6561 -0.3136**
(0.1501) (0.0962) (0.4876) (0.1519)
L.Log(gdp_pc_cons) -0.2359 -0.4221** -0.3693 -0.2168
(0.2378) (0.2141) (0.4596) (0.3273)
Log(distance) -0.8288*** -0.7969*** -0.5226** -1.5403***
(0.1264) (0.1078) (0.2403) (0.1879)
Common language -1.3475*** -1.3260*** 2.6236*** -0.9606***
(0.1820) (0.2079) (0.5999) (0.3071)
Common colony 1.1164** 1.3692*** 1.4944* 1.6279**
(0.4383) (0.4507) (0.8732) (0.6863)
Common border 1.5393*** 1.3834*** -1.7692*** 1.4008***
(0.1343) (0.1523) (0.4704) (0.2231)
COMESA dummy 0.2150 0.3937 0.8277 0.4581
(0.2518) (0.2678) (0.5207) (0.4300)
Constant 12.7017*** 14.0837*** 14.3512*** 29.3487***
(0.4765) (0.5362) (1.3798) (0.9840)
Observations 458 458 458 458
R-squared 0.9886 0.9859 0.9082 0.9952
Year dummy YES YES YES YES
Destination Country Dummy YES YES YES YES
30
destination fragility on exports is shown to be largely through the reduction in the number of
exporters. Column (2), also finds a significant negative effect of fragility on the number of
products exported. A 1 unit increase in destination fragility is associated with a 35%
reduction in the number of products exported to that destination. Column (3) results shows
that there is no significant relationship between fragility and the average value of shipment
(intensive margin).This suggests that the main channel through which fragility affects
bilateral exports is in the reduction in the number of exporters and the number of products
(extensive margin).
Finally, the results on column (4) confirms in line with most gravity findings in the literature
that the overall effect of destination fragility on bilateral exports is negative and statistically
significant (see Pollins 1989, Mansfield and Bronson, 1997, Martin et al., 2008, and Glick
and Taylor, 2010). A 1 unit increase in the destination fragility is associated with a 31%
reduction in Kenya’s aggregate exports to a given destination. Comparing our results to the
findings in the literature, Crozet et al (2007) finds for France that a 1 unit increase in the
country insecurity index was associated with a reduction in the number of exporters by 48%
and no significant effect on the mean exports (intensive margin) per firm, just as our results
show. Blomberg and Hess (2006) find the effect of conflict on bilateral trade to be equivalent
to a 30% tariff on trade, while Martin et al (2008) found an effect of conflict on bilateral trade
of approximately 22%.
4.5 Robustness tests
This section considers sensitivity tests to our empirical results. Firstly, we suspect that our
result for selection of firms is driven by the proxy measure of destination fragility used, in
particular the aggregation using the Principal Component Analysis (PCA) method. We re-run
equation (16) where each of the six indicators from the WGI compiled by Kaufmann et al
(2011) is used as an explanatory variable in a separate regression. The results are shown in
Table 7:
31
Table 7: Firm’s export status to a given destination and other proxies for fragility
Notes: The dependent variable is the probability that a firm maintains an export status to destination j. Each of
the six indicators in the WGI from Kaufmann et al.(2011) is used as an explanatory variable in a separate LPM
regression with time and firm-country fixed effects. Asterisk denotes level of significance (*** p<0.01, **
p<0.05, * p<0.1). Robust standard errors are in brackets.
Column (1) contains results for our preferred measure, which is the average index obtained
through the PCA method. The other proxy measures includes voice and accountability,
regulatory quality, rule of law, political stability and absence of violence and terrorism,
government effectiveness and control of corruption. The results are obtained using the LPM
method with year and firm-country fixed effects. Across all estimations, the coefficient on
destination country fragility is negative and statistically significant, except for the voice and
accountability indicator that is positive and significant. However, the interaction term
between this variable and firm size is negative and statistically significant, suggesting the
relationship flips for large firms.
Secondly, there are concerns that the use of the WGI by Kaufmann et al (2011) as proxy for
destination country fragility may be driving the results. To test the robustness of our results to
alternative measure of fragility, we follow Crozet et al (2007) and use the International
Dependent variable
(1) (2) (3) (4) (5) (6) (7)
L.Fragility -0.0895***
(0.0209)
L.ln(fob_ini)(Fragility) 0.0065***
(0.0019)
L.(voice & account_vac) 0.0988***
(0.0329)
L.(lnfob_ini) (voice & account_vac) -0.0088***
(0.0030)
L.(regulatory quality_rq) -0.2084***
(0.0279)
L.(lnfob_ini)(regulatory quality_rq) 0.0138***
(0.0026)
L.(rule of law_rol) -0.1559***
(0.0307)
L.(lnfob_ini)(rule of law_rol) 0.0098***
(0.0029)
L.(political stability_pse) -0.0560***
(0.0138)
L.(lnfob_ini)(political stability_pse) 0.0051***
(0.0013)
L.(government effect_ge) -0.2056***
(0.0268)
L.(lnfob_ini)(government effect_ge) 0.0120***
(0.0026)
L.(control of corruption_coc) -0.1184***
(0.0214)
L.(lnfob_ini)(control of corruption_coc) 0.0065***
(0.0021)
Observations 208,912 208,912 208,912 208,912 208,912 208,912 208,912
Number of groups 23,323 23,323 23,323 23,323 23,323 23,323 23,323
R-squared 0.2584 0.2583 0.2587 0.2585 0.2584 0.2589 0.2587
Year FE YES YES YES YES YES YES YES
Firm-Country FE YES YES YES YES YES YES YES
F-Statistics 97.59 95.30 106.4 101.1 96.45 109.8 105.5
Firm export status to j (exp_ijt>0)
32
Country Risk Guide (ICRG) as proxy for fragility (PRS Group, 2011). Table 8 shows the
results.
Table 8: Robustness to alternative measurement of fragility using ICRG Index
Notes: The dependent variable is the probability that a firm export to destination j. The proxy for destination
country fragility is the ICRG obtained from the PRS Group website. Asterisk denotes level of significance (***
p<0.01, ** p<0.05, * p<0.1). Robust standard errors are in brackets.
International Country Risk Guide (ICRG) indices provides a total score on political stability
across countries and the availability of this variable matches the sample period of 2004-2013.
However, the data is only available for 32 countries in Sub-Sahara Africa, out of the
approximately 140 countries covered, which means we lose some observations. The indices
ranges from 0 (very unstable) to 100(most stable), which means to measure fragility we must
use an inverse of the index. An increase in the score translates to a decrease in fragility in a
destination country. This index measure political stability along dimensions such as
socioeconomic conditions, democracy, ethnic tensions and military conflicts.
Across all estimations (column 1-5) the sign on the coefficients on the destination country
fragility remains negative and statistically significant. Column (3-5) controls for the firm’s
Dependent variable:
(1) (2) (3) (4) (5)
LPM LPM LPM LPM LPM
L.Fragility -0.0724** -0.6778*** -0.6249*** -0.6282*** -0.6261***
(0.0286) (0.0968) (0.0961) (0.0960) (0.0960)
(L.Fragility)(Logfob_ini) 0.0579*** 0.0541*** 0.0537*** 0.0539***
(0.0089) (0.0088) (0.0088) (0.0088)
L.Log(avgnrfirms_sameHS6) 0.0095
(0.0145)
(L.Fragility)(L.Log(avgnrfirms_sameHS6) -0.0112***
(0.0041)
L.Log(avgnrfirms_same sector) 0.0281***
(0.0063)
(L.Fragility)(L.Log(avgnrfirms_same sector) 0.0014
(0.0017)
L.Log(avgnrfirm-product) 0.0135***
(0.0036)
(L.Fragility)L.Log(avgnrfirm-product) -0.0000
(0.0010)
L.Log (gdp_cons) 0.0631*** 0.0645*** 0.0414** 0.0467** 0.0465**
(0.0205) (0.0205) (0.0203) (0.0203) (0.0203)
L.Log(real exchange rate) -0.0489*** -0.0471*** -0.0522*** -0.0475*** -0.0468***
(0.0141) (0.0141) (0.0140) (0.0140) (0.0140)
L.Log(days_import) 0.0029 0.0028 0.0146* 0.0096 0.0094
(0.0087) (0.0087) (0.0086) (0.0086) (0.0086)
Observations 138,651 138,651 138,651 138,651 138,651
Number of groups 17380 17380 17380 17380 17380
R-squared 0.2555 0.2557 0.2639 0.2641 0.2642
Year FE YES YES YES YES YES
Firm-Country FE YES YES YES YES YES
Firm’s export status to country j (exp_ijt>0)
33
network and again the results are remarkably similar to those when using WGI index. This
suggests that our regression results are robust to alternative choice of proxy used to measure
destination country fragility. More importantly, all these results confirms the predictions from
our model as well as the important role of firm size and networks in overcoming the costs to
export to fragile states.
Table 9 show that fragility reduces bilateral trade largely through a reduction in the number of
exporters and on the number of products exported (extensive margin). Unlike the baseline
estimation in Table 6 the effect on the average firm-product export value (intensive margin)
is positive and statistically significant (column 3). This is not totally unexpected but in fact an
explanation of the likely unambiguous effect of destination fragility on firm’s export activity
once firm enters the given destination. Furthermore, because fragility causes small firms to
exit, this raises the average export value for the remaining large firms.
Table 9: Alternative measurement of Fragility and trade margins
Notes: The dependent variables: Column (1-4) number of firms, number of products, average shipment and total
exports to a destination country j. All regressions are from PPML estimation as proposed by Silva and Tenreyro
(2006). We control for the year and destination country dummy. Asterisk denotes level of significance
(***p<0.01, **p<0.05, and *p<0.1), robust standard errors in brackets.
Dependent variables: Firms Products Av.value Tot.value
(1) (2) (3) (4)
L.Fragility -0.3396* -0.6624*** 1.8141** -0.3898**
(0.1968) (0.2482) (0.9086) (0.1882)
L.Log(gdp_pc_cons) -0.1195 -0.0544 -0.6494* -0.1903
(0.2090) (0.2386) (0.3913) (0.2979)
L.Log(days to import) 0.0607 0.2714*** 0.2854 -0.2014***
(0.0864) (0.1045) (0.4089) (0.0672)
Log(distance) -1.7670*** -1.8352*** -0.5718* -3.2528***
(0.0849) (0.1110) (0.3228) (0.2167)
Common language -2.5710*** -2.8125*** 1.9363* -3.4546***
(0.2267) (0.2790) (1.0073) (0.2584)
Common colony 0.1812 -0.0258 0.5451 0.0137
(0.1713) (0.2075) (0.4501) (0.2334)
Common border 0.3897** 0.0627 0.0567 -0.6017***
(0.1692) (0.2028) (0.5270) (0.2184)
COMESA dummy -0.2843*** -0.3627*** -0.0976 -0.3073**
(0.0738) (0.0876) (0.3025) (0.1421)
Constant 19.5665*** 18.7519*** 22.5127*** 45.0213***
(1.1023) (1.4448) (2.7932) (2.0442)
Observations 315 315 315 315
R-squared 0.9893 0.9890 0.9231 0.9957
Year dummy YES YES YES YES
Destination Country Dummy YES YES YES YES
34
5.0 Conclusion
This paper examines the effect of destination country fragility on a firm’s decision to serve a
given destination with exports and firm attributes that mediates the effect of fragility. While it
is has been known that fragility reduces aggregate trade between countries, we show that this
effect occurs mainly through the extensive margin. In particular, we find a large and
significant negative effect of fragility on the probability of a firm serving a fragile
destination. A 1 unit increase in the destination country fragility index is associated with a
decrease in the probability that a firm serves that destination by 8.2 percentage points.
An increase in the firm size in exports is found to be a critical attribute in overcoming the
negative effects of destination country fragility. The interaction term between destination
fragility and proxy for firm size is positive and significant. A 10% increase in the size of the
firm is on average, associated with an increase in the probability that a firm exports to a given
destination by 4.4 percentage points.
Firm’s network is in general associated with a high probability that a firm exports to a given
destination. A 10% increase in the average number of firms that export similar products
(HS6) to the same destination is associated with a 0.7 percentage points increase in the
probability that a firm belonging to this network exports to the destination. However the
interaction of firm network and destination fragility is mostly positive but not statistically
significant, which suggests that the gains from positive spillovers weakens if a destination
market is fragile.
On the effect of destination fragility on Kenya’s trade margins, we find that the main channel
through which fragility affects bilateral trade is in the reduction of the number of exporters
and the number of products to a fragile state (extensive margin). There is no significant
relationship between destination country fragility and the average firm-product export value
(intensive margin) of trade. Overall, destination fragility is associated with significant
reduction in aggregate exports to a given destination. These findings are in line with our
theoretical formulation and are robust to a number of consistency checks.
35
References
Abadie, A. & Gardeazabal, J. 2003. The Economic Costs of Conflict: A Case Study of the Basque Country. The American Economic Review. 93(1):113-132.
Acemoglu, D., Johnson, S. & Robinson, J.A. 2005. Institutions as a fundamental cause of long-run growth. Handbook of Economic Growth. 1:385-472.
Aitken, B., Hanson, G.H. & Harrison, A.E. 1997. Spillovers, foreign investment, and export behavior. Journal of International Economics. 43(1):103-132.
Albornoz, F., Pardo, H.F.C., Corcos, G. & Ornelas, E. 2012. Sequential exporting. Journal of International Economics. 88(1):17-31.
Alesina, A. & Perotti, R. 1996. Income distribution, political instability, and investment. European Economic Review. 40(6):1203-1228.
Anderson, J.E. & Marcouiller, D. 2002. Insecurity and the pattern of trade: An empirical investigation. Review of Economics and Statistics. 84(2):342-352.
Anderson, J.E. & Van Wincoop, E. 2003. Gravity with gravitas: a solution to the border puzzle. The American Economic Review. 93(1):170-192.
Araujo, L.F. & Ornelas, E. 2007. Trust-based trade. SSRN Working Papers no.642803. 1(02):1-36.
Barbieri, K. 2002. The liberal illusion: Does trade promote peace? Michigan: University of Michigan Press.
Barrera-Gomez, J. & Basagana, X. 2015. Models with transformed variables: interpretation and software. Epidemiology. 26(2):e17.
Bernard, A.B., Eaton, J., Jensen, J.B. & Kortum, S. 2003. Plants and productivity in international trade. The American Economic Review. 93(4):1268-1290.
Bernard, A.B. & Jensen, J.B. 1999. Exceptional exporter performance: cause, effect, or both? Journal of International Economics. 47(1):1-25.
Bernard, A.B., Jensen, J.B. & Lawrence, R.Z. 1995. Exporters, jobs, and wages in US manufacturing: 1976-1987. Brookings Papers on Economic Activity.Microeconomics. 1995:67-119.
Bernard, A.B., Jensen, J.B., Redding, S.J. & Schott, P.K. 2007. Firms in international trade. The Journal of Economic Perspectives. 21(3):105-130.
Bernard, A.B., Jensen, J.B., Redding, S.J. & Schott, P.K. 2009. The margins of US trade. The American Economic Review. 99(2):487-493.
Bernard, A.B., Redding, S.J. & Schott, P.K. 2011. Multiproduct firms and trade liberalization. The Quarterly Journal of Economics. 126(3):1271-1318.
Bernard, A.B., Van Beveren, I. & Vandenbussche, H. 2014. Multi‐Product Exporters and the Margins of Trade. Japanese Economic Review. 65(2):142-157.
Besedeš, T. & Prusa, T.J. 2011. The role of extensive and intensive margins and export growth. Journal of Development Economics. 96(2):371-379.
Blomberg, S.B. & Hess, G.D. 2006. How much does violence tax trade? The Review of Economics and Statistics. 88(4):599-612.
Blomberg, S.B., Hess, G.D. & Weerapana, A. 2004. Economic conditions and terrorism. European Journal of Political Economy. 20(2):463-478.
Brenton, P., Cadot, O. & Pierola, M.D. 2012. Pathways to African export sustainability. Washington, DC: World Bank Publications.
Bussmann, M., Schneider, G. & Wiesehomeier, N. 2005. Foreign economic liberalization and peace: The case of sub-Saharan Africa. European Journal of International Relations. 11(4):551-579.
Cadot, O., Iacovone, L., Pierola, M.D. & Rauch, F. 2013a. Success and failure of African exporters. Journal of Development Economics. 101:284-296.
Carrre, C. 2004. African regional agreements: impact on trade with or without currency unions. Journal of African Economies. 13(2):199-239.
Chaney, T. 2008. Distorted gravity: the intensive and extensive margins of international trade. The American Economic Review. 98(4):1707-1721.
Chauvet, L. & Collier, P. 2008. What are the preconditions for turnarounds in failing states? Conflict Management and Peace Science. 25(4):332-348.
36
Chauvet, L., Collier, P. & Hoeffler, A. 2007. The cost of failing states and the limits to sovereignty. Fragile States: Causes, Costs and Responses. :91-110.
Choquette, E. & Meinen, P. 2015. Export spillovers: opening the black box. The World Economy. 38(12):1912-1946.
Colin Cameron & Pravin Trivedi. 2010. Microeconometrics using Stata: Revised Edition. College Station, Texas: Stata Corp LP.
Collier, P. & Gunning, J.W. 1999. Explaining African economic performance. Journal of Economic Literature. 37(1):64-111.
Collier, P. & Hoeffler, A. 2004. Greed and grievance in civil war. Oxford Economic Papers. 56(4):563-595.
Crozet, M., Koenig, P. & Rebeyrol, V. 2007. Exporting to risky markets: a firm-level analysis. MimeoUniversity of Reims. 01(2007-01):1-31.
Easterly, W. 2005. What did structural adjustment adjust?: The association of policies and growth with repeated IMF and World Bank adjustment loans. Journal of Development Economics. 76(1):1-22.
Eaton, J., Eslava, M., Krizan, C.J., Kugler, M. & Tybout, J. 2009. A search and learning model of export dynamics. Penn State University Mimeo.
Fearon, J.D. & Laitin, D.D. 2003. Ethnicity, insurgency, and civil war. American Political Science Review. 97(01):75-90.
Franc J. G. M. Klaassen & Jan R. Magnus 2001. Are Points in Tennis Independent and Identically Distributed? Evidence from a Dynamic Binary Panel Data Model. Journal of the American Statistical Association. 96(454):500-509.
Glick, R. & Taylor, A.M. 2010. Collateral damage: Trade disruption and the economic impact of war. The Review of Economics and Statistics. 92(1):102-127.
Greenaway, D. & Kneller, R. 2008. Exporting, productivity and agglomeration. European Economic Review. 52(5):919-939.
Haveman, J. & Hummels, D. 2004. Alternative hypotheses and the volume of trade: the gravity equation and the extent of specialization. Canadian Journal of Economics/Revue Canadienne D'Conomique. 37(1):199-218.
Horrace, W.C. & Oaxaca, R.L. 2006. Results on the bias and inconsistency of ordinary least squares for the linear probability model. Economics Letters. 90(3):321-327.
Iqbal, Z. & Starr, H. 2008. Bad neighbors: Failed states and their consequences. Conflict Management and Peace Science. 25(4):315-331.
Kaufmann, D., Kraay, A. & Mastruzzi, M. 2011. The worldwide governance indicators: methodology and analytical issues. Hague Journal on the Rule of Law. 3(2):220-246.
Klaassen, F.J. & Magnus, J.R. 2001. Are points in tennis independent and identically distributed? Evidence from a dynamic binary panel data model. Journal of the American Statistical Association. 96(454):500-509.
Macchiavello, R. & Morjaria, A. 2015. The value of relationships: evidence from a supply shock to Kenyan rose exports. The American Economic Review. 105(9):2911-2945.
Mansfield, E.D. & Bronson, R. 1997. Alliances, preferential trading arrangements, and international trade. American Political Science Review. 91(01):94-107.
Mansfield, E.D. & Pevehouse, J.C. 2000. Trade blocs, trade flows, and international conflict. International Organization. 54(04):775-808.
Martin, P., Mayer, T. & Thoenig, M. 2008. Make trade not war? The Review of Economic Studies. 75(3):865-900.
McMillan, J. & Woodruff, C. 1999. Interfirm relationships and informal credit in Vietnam. The Quarterly Journal of Economics. 114(4):1285-1320.
Melitz, M.J. 2003. The impact of trade on intra‐industry reallocations and aggregate industry productivity. Econometrica. 71(6):1695-1725.
Messner, J.J. 2005. Fragile States Index... Morrow, J.D., Siverson, R.M. & Tabares, T.E. 1998b. The political determinants of international
trade: the major powers, 1907–1990. American Political Science Review. 92(03):649-661. Murdoch, J.C. & Sandler, T. 2004. Civil Wars and Economic Growth: Spatial Dispersion. American
Journal of Political Science. 48(1):138-151. Penubarti, M. & Ward, M.D. 2000. Commerce and democracy. Center for Statistics and the Social
Sciences Working Paper. 2(6):1-36.
37
Pollins, B.M. 1989. Conflict, cooperation, and commerce: The effect of international political interactions on bilateral trade flows. American Journal of Political Science. :737-761.
PRS Group 2011. ICRG methodology. The PRS Group. Qureshi, M.S. 2013. Trade and thy neighbor's war. Journal of Development Economics. 105:178-
195. Rauch, J.E. & Watson, J. 2003. Starting small in an unfamiliar environment. International Journal
of Industrial Organization. 21(7):1021-1042. Sambanis, N. 2001. Do ethnic and nonethnic civil wars have the same causes? A theoretical and
empirical inquiry (Part 1). Journal of Conflict Resolution. 45(3):259-282. Silva, J.S. & Tenreyro, S. 2006. The log of gravity. The Review of Economics and
Statistics. 88(4):641-658. Silvente, F.R. & Gimnez, J.C. 2007. Information spillovers and the choice of export destination: a
multinomial logit analysis of Spanish young SMEs. Small Business Economics. 28(1):69-86. Soderbom, M., Teal, F., Eberhardt, M., Quinn, S. & Zeitlin, A. 2014. Empirical development
economics. New York: Routledge. Stewart, F. & Brown, G. 2009. Fragile states. Oxford: University of Oxford. Centre for research on
inequality, human security and ethnicity (CRISE). World Bank. 2011. World development report 2011: conflict, security, and
development. Washington, DC: World Bank. Yeats, A.J. 1998. What can be expected from african regional trade arrangements?: Some
empirical evidence. SSRN Working Papers. 7(1):1-30.
38
Appendix 1A: Country rank by average share of export and fragility
Table 10: Export share and designation fragility index (mean values 2004-2013) Rank ISO3 Country Share of export Fragility
index
FCSdum
1 UGA Uganda 0.29 0.62 0
2 TZA Tanzania 0.19 0.42 0
4 EGY Egypt 0.10 0.63 0
4 SDN Sudan 0.09 1.65 1
5 ZAR DR Congo 0.07 1.73 1
6 RWA Rwanda 0.06 0.47 0
7 ZMB Zambia 0.04 0.40 0
9 ETH Ethiopia 0.03 1.00 1
9 BDI Burundi 0.03 1.25 1
10 ZAF South Africa 0.02 -0.35 0
10 MWI Malawi 0.02 0.38 0
12 NGA Nigeria 0.01 1.21 1
14 DJI Djibouti 0.01 0.71 1
14 MUS Mauritius 0.01 -0.84 0
15 MOZ Mozambique 0.01 0.38 0
18 MDG Madagascar 0.004 0.52 0
18 GHA Ghana 0.003 -0.05 0
19 COM Comoros 0.003 1.07 1
19 ZWE Zimbabwe 0.004 1.62 1
19 ERI Eritrea 0.004 1.42 1
21 SYC Seychelles 0.002 -0.14 0
22 AGO Angola 0.002 1.20 1
24 MAR Morocco 0.001 0.34 0
27 CIV Cote d'Ivoire 0.001 1.29 1
30 MLI Mali 0.001 0.43 0
31 SEN Senegal 0.001 0.29 0
31 CMR Cameroon 0.001 0.97 1
31 SLE Sierra Leone 0.0004 0.83 1
32 GIN Guinea 0.001 1.35 1
33 LBR Liberia 0.001 1.02 1
33 BWA Botswana 0.0004 -0.75 0
34 BEN Benin 0.0003 0.33 0
34 TCD Chad 0.0003 1.46 1
34 NER Niger 0.0004 0.68 1
35 NAM Namibia 0.0003 -0.34 0
36 STP Sao Tome- Princ. 0.0001 0.39 0
37 BFA Burkina Fasso 0.0003 0.37 0
37 COG Congo, Brazz. 0.0003 1.17 1
37 TUN Tunisia 0.0004 0.13 0
37 CAF Central Africa 0.0002 1.44 1
39
Rank ISO3 Country Share of export Fragility
index
FCSdum
37 LBY Libya 0.0002 1.17 1
38 MRT Mauritania 0.0002 0.77 1
38 GMB Gambia 0.0002 0.54 0
39 TGO Togo 0.0002 1.05 1
39 GAB Gabon 0.0002 0.60 0
40 GNQ Equatorial Guinea 0.0001 1.42 1
41 DZA Algeria 0.0004 0.84 1
42 LSO Lesotho 0.0001 0.19 0
42 SWZ Swaziland 0.0001 0.67 1
46 CPV Cape Vade 0.00004 -0.51 0
47 GNB Guinea-Bissau 0.00002 1.16 1
1.00 26.00
Notes: FCSdum equal 1, if the country average fragility index is greater than the median value of 0.66.
The data is restricted to exports to Africa from Kenya over the sample period 2004-2013. The rank is based on
export share ascribed to a country, over the sample period.
Appendix 1B: Measurement of Key variables
Exporter entry to destination 𝑗 (𝑒𝑥𝑝_𝑖𝑗𝑡)
This is a binary variable taking 1, if a firm exports to destination 𝑗 in period 𝑡 and zero
otherwise over the period 2004 to 2013. The choice to maintain or terminate an export
relationship in a given destination can be observed for all exporters. To give an example, if a
firm is observed in 2004 and exports to three destination countries, it is assigned a 1 for each
of these destinations in 2004. In 2005, if the firm exports to same three countries, it again gets
a 1 for each of the destination in 2005 and so on. If in addition in 2005, the firm adds a fourth
destination that gets a 1 as well. Thus for every new and maintained export status to a given
country, the firm gets 1. The firm gets a zero if it either drops one of its destinations or for
each of the potential destinations where the firm has not ventured.
Firm’s export size at entry (𝑓𝑜𝑏_𝑖𝑛𝑖)
The literature that uses transaction data has established a positive correlation between the
initial value of exports at entry and subsequent export performance. We follow the literature
in using the initial value of exports at entry in export status as proxy for the firm size. Using
the initial value at entry to classify firms has the advantage of avoiding unobserved events
that may cause firms to substantially ramp up exports after entry.
Accounting for firm’s network
We constructed three different proxy measures for firm network. The first measure is the
average number of firms exporting similar products (HS6) to the same destination. This
variable constitute a narrow measure of firm’s network. Firm’s exporting similar products to
the same destination can be considered competitors. At the same time, there is a lot of
signalling effect and learning from the fortunes of other exporters, including potential for new
profit (Cadot et al., 2013). A second proxy is the average number of firms in a sector (HS2),
exporting to the same destination. This variable is expected to capture foreign market intra-
industry spillovers (Choquette and Meinen, 2015). A third proxy is the average number of
firm-product to a given destination. This is used in the literature to capture foreign market
40
specific inter industry spillovers (Cadot et al., 2013, Choquette and Meinen, 2015). By the
nature of construction, these variables are highly correlated and log transformation of each is
included in a separate regression.
Kenya’s trade margins to a given destination
Kenya’s export to each destination j in period t is decomposed into the unique number of
firms that trade with the country, the unique number of products traded with the country, and
the average value of exports per firm-product (Bernard et al 2007, 2014). The number of
firms and the number of products constitutes the extensive margins, while the average value
of exports per firm-product constitute the intensive margin. These margins are used as
dependent variables in equation (19).
Market size and other gravity variables
We use destination country GDP at 2005 constant prices as proxy for market size. This is
obtained from the world development indicators (WDI) of the World Bank over the sample
period 2004-2013. We include common border, common language, common colonial history,
population, and distance between Nairobi and the capital city of the partner country as
controls. These variables are obtained from CEPII database. Distance is used as proxy for
variable costs between Kenya and the partner country. We also include COMESA dummy to
control for the regional trade agreement.
Appendix 1C: Decomposition of Kenya’s exports into intensive and extensive margins
To illustrate, let Kenya’s total exports with a partner country 𝑗 be Xj . This value is then
decomposed into the unique number of firms that trade with the country (Fj), the unique
number of products traded with the country (Pj), and the average value of exports per firm-
product, (Xj/FjPj ). Since firms generally are active in a small subset of the overall number of
products exported by the country, Bernard et al (2013) defines an additional term in the
decomposition to account for the density of trade (Dj ) (i.e. the fraction of all possible firm-
product combinations for country j for which trade is positive).
𝐿𝑜𝑔(𝑋𝑗) = 𝐿𝑜𝑔(𝐹𝑗) + 𝐿𝑜𝑔(𝑃𝑗) + 𝐿𝑜𝑔(𝐷𝑗) + 𝐿𝑜𝑔(��𝑗)
𝐷𝑗 =𝑂𝑗
𝐹𝑗𝑃𝑗
Oj = is the number of firm-product observations for which trade with country 𝑗 is above zero
and Xj = Xj/Oj is the intensive margin, which is the average value per observation with
positive trade. The density ranges from min {1/Fj, 1/Pj} to one unit as the number of
observations approaches Fj ∗ Pj. Since firms generally are active in only a small subset of the
overall number of products traded, density is typically negatively correlated with the numbers
of trading firms and the number of traded products.
The above decomposition equation provides the basis for a regression decomposing Kenya’s
export trade across countries for a particular year. In this example we use OLS to regress the
logarithm of each margin of trade on the logarithm of total exports trade to a country j for the
year, 2012.Table 11 shows the results.
41
Table 11: decomposition of Kenya’s export trade to destination j, 2012
Notes: Each cell contains the result of a separate regression of the log dependent variable on
the log of total exports to a given destination, with the coefficient and robust standard errors
in the brackets. The asterisks denotes the level of significance *** p<0.01, ** p<0.05, *
p<0.1. The sum of the four coefficients adds up to 1, fully accounting for the exports to a
given country.
The results shows the significant role of the extensive margin in the variation of Kenya’s
trade across countries. In 2012, approximately 53% of the variation in cross-country exports
was as a result of the extensive margin (i.e. sum of number of firms, number of products, and
density). The intensive margin accounted for approximately 47.1% of the variation. This
shows that the extensive margin plays a dominant role in the cross country variation in export
shipment from Kenya.
Appendix 2: Mathematical derivations
Equation (3)
𝐸(𝑃) = 𝜇2𝑌𝑓𝑃𝑓 (19.)
𝜕𝐸(𝑃)
𝜕𝑝𝑖𝑓
=𝜕𝐸(𝑃)
𝜕𝑃𝑓
𝜕𝑃𝑓
𝜕𝑝𝑖𝑓
𝜕𝐸(𝑃)
𝜕𝑃𝑓
= 𝜇2𝑌𝑓
𝜕𝑃𝑓
𝜕𝑝𝑖𝑓
= [∫ 𝑝𝑖𝑓1−𝜎
𝑛
𝑖
𝑑𝑖]
11−𝜎
−1
𝑝𝑖𝑓−𝜎
𝜕𝐸(𝑃)
𝜕𝑝𝑖𝑓
= 𝜇2𝑌𝑓 [∫ 𝑝𝑖𝑓1−𝜎
𝑛
𝑖
𝑑𝑖]
σ1−𝜎
𝑝𝑖𝑓−𝜎 = 𝑃𝑓𝑞𝑖𝑓
𝜕𝐸(𝑃)
𝜕𝑝𝑖𝑓
= 𝜇2𝑌𝑓𝑃𝑓𝜎 𝑝𝑖𝑓
−𝜎 = 𝑃𝑓𝑞𝑖𝑓
𝑞𝑖𝑓 = 𝜇2𝑌𝑓𝑃𝑓𝜎−1 𝑝𝑖𝑓
−𝜎
Equation (7)
𝑀𝑅 = 𝑀𝐶 (20.) 𝑀𝑅
𝑝𝑖𝑓
= 1 −1
𝜎= (
𝜎 − 1
𝜎)
𝑀𝐶 =𝜕𝑇𝐶
𝜕𝑞𝑖𝑓
=𝜏
𝜑𝑖
Dependent variables: Log(Firms) Log(Products) Log(Density) Log(Av.value)
(1) (2) (3) (4)
Log(ctotexport_TX_us) 0.458*** 0.456*** -0.385*** 0.471***
(0.0167) (0.0214) (0.0159) (0.0221)
Constant -3.413*** -3.220*** 2.798*** 3.835***
(0.238) (0.304) (0.227) (0.315)
Observations 166 166 166 166
R-squared 0.821 0.734 0.780 0.734
42
𝑝𝑖𝑓 = τ (𝜎
𝜎 − 1)
𝜏
𝜑𝑖
Equation (12)
(1 − 𝛼)𝜋𝑋𝐻(𝜑𝑖) + 𝛼𝜋𝑋𝐻(𝜑𝑖) ≥ 𝜋𝐷𝐻(𝜑𝑖) ⇔ (1 − 𝛼)(𝜋𝑋𝐻(𝜑𝑖) − 𝛽𝐹𝐻𝑋) + 𝛼𝜋𝑋𝐻(𝜑𝑖)
≥ 𝜋𝐷𝐻(𝜑𝑖)
(𝜋𝑋𝐻(𝜑𝑖) − 𝛽𝐹𝐻𝑋) − 𝛼 ((𝜋𝑋𝐻(𝜑𝑖) − 𝛽𝐹𝐻
𝑋)) + 𝛼𝜋𝑋𝐻(𝜑𝑖) = 𝜋𝐷𝐻(𝜑𝑖)
(𝜋𝑋𝐻 − 𝜋𝐷𝐻) = (1 − 𝛼)𝛽𝐹𝐻𝑋
LHS
(𝜋𝑋𝐻 − 𝜋𝐷𝐻) =𝜇2
𝜎(
𝜎
𝜎 − 1)
1−𝜎
𝜏1−𝜎𝑌𝐹𝑃𝐹𝜎−1 (
1
𝜑𝑖
)1−𝜎
− 𝐹𝐻𝑋
(𝜋𝑋𝐻 − 𝜋𝐷𝐻)1
1−𝜎 = (𝜇2
𝜎)
11−𝜎
(𝜎
𝜎 − 1) 𝜏 𝑌𝐹
11−𝜎 𝑃𝐹
−1 (1
𝜑𝑖
) − 𝐹𝐻𝑋
11−𝜎
((1 − 𝛼)𝛽𝐹𝐻𝑋)
11−𝜎
= (𝜇2
𝜎)
11−𝜎
(𝜎
𝜎 − 1) 𝜏 𝑌𝐹
11−𝜎 𝑃𝐹
−1 (1
𝜑𝑖
) − 𝐹𝐻𝑋
11−𝜎
Let: 𝜆1 = [𝜎
𝜇2]
1
𝜎−1 (
𝜎
𝜎−1) and collect like terms and solve for 𝜑𝑖
(21.)
𝜑𝑖 ≥ 𝜑𝑋𝐻 = 𝜆1 [𝐹𝐻
𝑋 + (1 − 𝛼)𝛽𝐹𝐻𝑋
𝑌𝐹
]
1𝜎−1 𝜏
𝑃𝐹
Equation (13)
𝜑𝑋𝐻 = 𝜆1 [𝐹𝐻
𝑋 + (1 − 𝛼)𝛽𝐹𝐻𝑋
𝑌𝐹
]
1𝜎−1 𝜏
𝑃𝐹
(22.)
𝜕(𝜑𝑋𝐻
)
𝜕𝛽=
𝜕(𝜑𝑋𝐻
)
𝜕𝑈 𝜕𝑈
𝜕𝛽
𝜕(𝜑𝑋𝐻
)
𝜕 ∝=
𝜕(𝜑𝑋𝐻
)
𝜕𝑈 𝜕𝑈
𝜕 ∝
𝑈 =𝐹𝐻
𝑋 + (1 − 𝛼)𝛽𝐹𝐻𝑋
𝑌𝐹
𝜕(𝜑𝑋𝐻
)
𝜕𝑈=
𝜆1
𝜎 − 1𝑈
1𝜎−1
−1 𝜏
𝑃𝐹
> 0
𝜕(𝑈)
𝜕𝛽=
(1−∝)𝐹𝐻𝑋
𝑌𝐹
> 0
𝜕(𝑈)
𝜕 ∝=
−𝛽𝐹𝐻𝑋
𝑌𝐹
< 0
𝜕(𝜑𝑋𝐻
)
𝜕𝛽=
𝜕(𝜑𝑋𝐻
)
𝜕𝑈 𝜕𝑈
𝜕𝛽> 0
𝜕(𝜑𝑋𝐻
)
𝜕 ∝=
𝜕(𝜑𝑋𝐻
)
𝜕𝑈 𝜕𝑈
𝜕 ∝< 0
The effect of destination fragility raises the productivity threshold to maintain an export status in the fragile
destination.
43
Equation (15)
𝑋𝐻𝐹 = 𝑁𝐻𝐹 𝑝𝑖𝐻 𝑞𝑖𝐻
𝑋𝐻𝐹 = 𝜇2𝑌𝐹𝑃𝐹𝜎−1 (
𝜎
𝜎 − 1)
1−𝜎
𝜏1−𝜎 𝑁𝐻𝐹
(23.)
𝜕(𝑋𝐻𝐹)
𝜕𝛽=
𝜕(𝑋𝐻𝐹)
𝜕𝑁𝐻𝐹
𝜕𝑁𝐻𝐹
𝜕𝛽
𝜕(𝑋𝐻𝐹)
𝜕 ∝=
𝜕(𝑋𝐻𝐹)
𝜕𝑁𝐻𝐹
𝜕𝑁𝐻𝐹
𝜕 ∝
𝜕(𝑋𝐻𝐹)
𝜕𝑁𝐻𝐹
= 𝜇2𝑌𝐹𝑃𝐹𝜎−1 (
𝜎
𝜎 − 1)
1−𝜎
𝜏1−𝜎 > 0
𝑁𝐻𝐹 = 𝐿𝐻 ∫ (1
𝜑)
1−𝜎
𝑑𝐺(𝜑)
𝜑
𝜑𝑋𝐻
𝑑𝜑 + 𝛼𝐿𝐻 ∫ (1
𝜑)
1−𝜎
𝑑
𝜑𝑋𝐻
0
𝐺(𝜑)𝑑𝜑
Where: 0 < 𝜑 < 1, and 𝐺(𝜑) = (1
𝜑)
𝜌
is Pareto distributed and 𝜌 = 𝜎 − 1 integrating and solving:
𝑁𝐻𝐹 = 𝐿𝐻[𝜑𝑋𝐻−𝜌 − 𝜑𝑋𝐻 −𝜌+∝ 𝜑𝑋𝐻 −𝜌]
𝑁𝐻𝐹 = 𝐿𝐻[𝜑𝑋𝐻𝜎−1 − (1−∝) 𝜑𝑋𝐻 𝜎−1]
From Eq.(24)
𝜑𝑋𝐻 = 𝜆1 [𝐹𝐻
𝑋 + (1 − 𝛼)𝛽𝐹𝐻𝑋
𝑌𝐹
]
1𝜎−1 𝜏
𝑃𝐹
= 𝑈
𝑁𝐻𝐹 = 𝐿𝐻[𝜑𝑋𝐻𝜎−1 − (1−∝)𝑈𝜎−1]
𝜕𝑁𝐻𝐹
𝜕𝛽=
𝜕𝑁𝐻𝐹
𝜕𝑈 𝜕𝑈
𝜕𝛽;
𝜕𝑁𝐻𝐹
𝜕 ∝=
𝜕𝑁𝐻𝐹
𝜕𝑈
𝜕𝑈
𝜕 ∝
𝜕𝑁𝐻𝐹
𝜕𝑈= 𝐿𝐻[−(1−∝)(𝜎 − 1)𝑈(𝜎−1)−1] < 0
𝜕𝑈
𝜕𝛽= 𝜆1
𝜏
𝑃𝐹
(1
𝜎 − 1) [
𝐹𝐻𝑋 + (1 − 𝛼)𝛽𝐹𝐻
𝑋
𝑌𝐹
]
1𝜎−1
−1(1−∝)𝐹𝐻
𝑋
𝑌𝐹
> 0
𝜕𝑁𝐻𝐹
𝜕𝛽=
𝜕𝑁𝐻𝐹
𝜕𝑈 𝜕𝑈
𝜕𝛽< 0
𝜕𝑈
𝜕 ∝= 𝜆1
𝜏
𝑃𝐹
(1
𝜎 − 1) [
𝐹𝐻𝑋 + (1 − 𝛼)𝛽𝐹𝐻
𝑋
𝑌𝐹
]
1𝜎−1
−1−𝛽𝐹𝐻
𝑋
𝑌𝐹
< 0
𝜕𝑁𝐻𝐹
𝜕 ∝=
𝜕𝑁𝐻𝐹
𝜕𝑈
𝜕𝑈
𝜕 ∝> 0
𝜕(𝑋𝐻𝐹)
𝜕𝛽=
𝜕(𝑋𝐻𝐹)
𝜕𝑁𝐻𝐹
𝜕𝑁𝐻𝐹
𝜕𝛽< 0
𝜕(𝑋𝐻𝐹)
𝜕 ∝=
𝜕(𝑋𝐻𝐹)
𝜕𝑁𝐻𝐹
𝜕𝑁𝐻𝐹
𝜕 ∝> 0
The effect of fragility on bilateral exports is transmitted largely through the firm extensive margin.