A partnership with academia

Building knowledge for trade and development

Vi Digital Library - Text Preview

Survival Analysis of the Exports of Least Developed Countries: The Role of Comparative Advantage

Working paper by Alessandro Nicita, Miho Shirotori, Bolormaa Tumurchudur Klok, 2013

Download original document (English)

This paper investigates whether comparative advantage affects the duration of exports from least developed countries. The research finds that exports of products that are close to the country’s comparative advantage are likely to be more durable.

U n i t e d n at i o n s C o n f e r e n C e o n t r a d e a n d d e v e l o p m e n t










Alessandro Nicita

Miho Shirotori
Bolormaa Tumurchudur Klok

UNCTAD, Geneva

New York and Geneva, 2013



The purpose of this series of studies is to analyse policy issues and to stimulate discussions
in the area of international trade and development. The series includes studies by UNCTAD staff
and distinguished researchers from academia. This paper represents the personal views of the
authors only, and not the views of the UNCTAD secretariat or its member States.

The designations employed and the presentation of the material do not imply the
expression of any opinion on the part of the United Nations concerning the legal status of any
country, territory, city or area, or of authorities or concerning the delimitation of its frontiers or

This publication has not been formally edited.

Material in this publication may be freely quoted or reprinted, but acknowledgement is
requested, together with a copy of the publication containing the quotation or reprint to be sent to
the UNCTAD secretariat at the following address:

Chief, Trade Analysis Branch
Division on International Trade in Goods and Services, and Commodities

United Nations Conference on Trade and Development
Palais des Nations
CH-1211 Geneva

Series Editor:
Victor Ognivtsev

Officer-in-Charge, Trade Analysis Branch



ISSN 1607-8291

© Copyright United Nations 2013
All rights reserved



Motivated by the standard Heckscher-Ohlin theory, we investigate whether comparative
advantage affects the duration of exports from least developed countries (LDCs). To do so, we
first calculate each exported product’s distance from the country’s comparative advantage. Then
we estimate a semi-parametric Extended Cox model with time-independent/dependent explanatory
variables to measure export survival rates. We find evidence that a product's distance to
comparative advantage is a determinant of export survival for LDCs. Moreover, we find that the
influence of comparative advantage over LDC export survival increases with time. This implies
that exports of products that are close to the country’s comparative advantage are likely to be more
durable. In the long run, comparative advantage can evolve dynamically. Export diversification by
LDCs into non-traditional sectors calls for vigorous investment to improve the quantity and the
quality of their factor endowments.

Keywords: Export diversification, comparative advantage, LDCs

JEL Classification: F11



The authors wish to thank Ralf Peters for his valuable comments.



1 Introduction .......................................................................................................................... 1

2 Survival analysis findings .................................................................................................... 3

2.1 Estimated survival pattern of least developed countries’ exports
(Kaplan–Meier analysis) .............................................................................................. 3
2.2 Factors influencing the survival of least developed countries’ exports
(Extended Cox Proportional Hazards Model) .............................................................. 5

3 Conclusions ......................................................................................................................... 11

Appendix: Model specification ....................................................................................................... 12
1 Kaplan–Meier analysis (nonparametric approach) .................................................... 12
2 Extended Cox Proportional Hazard Model (semiparametric approach) .................... 12
3 Estimating a product’s distance from the exporting country’s
comparative advantage ............................................................................................... 13

References .................................................................................................................................... 14


List of figures

Figure 2.1 Kaplan–Meier survival estimates for least developed countries’ exports ....................... 4
Figure 2.2 Survival functions by export destination ......................................................................... 4
Figure 2.3 Time-varying effect of distance to comparative advantage on export survival ............... 9
Figure 2.4 Time-varying effect of distance to comparative advantage,
least developed countries and MERCOSUR ................................................................. 11

List of tables

Table 1.1 Export termination incidence rate ................................................................................... 2
Table 1.2 Survival duration of least developed countries’ exports, by country .............................. 3
Table 2.1 Results derived from semiparametric Cox model with gravity-type variables ............... 7
Table 2.2 Results derived from semiparametric Cox model ........................................................... 8
Table 2.3 Results derived from Cox model with time interactions ................................................. 8
Table 2.4 Results derived from Cox model with time interactions,
ASEAN and MERCOSUR ............................................................................................ 10


1. Introduction

Export diversification has long been instrumental to the successful development of
emerging economies, as it creates a broad base for export-led growth. For a large number of
least developed countries (LDCs), however, export diversification has been a chimera.
Exports of LDCs are still largely concentrated in very few sectors and are often limited to
primary commodities. Diversification, or the lack of it, is reflected in the statistics. The index
of export concentration (Herfindahl–Hirschmann) for LDCs has been increasing over the past
10 years. In 2009, the concentration of exports of LDCs was about four times as large as that
of developing economies as a whole.1

The lack of technological capacity to produce a wide variety of products and Dutch
disease symptoms may have been the major cause for export concentration in many LDCs.
Moreover, even after a country begins to export a non-traditional product, empirical evidence
indicates that sustaining the diversification in a way that ensures the continuing flow of new
exports is not an easy task. On average, an export spell – or the duration of a product being
exported to a particular market – is short: just one year for exports from LDCs and two years
for other developing countries and developed countries. Besedes and Prusa (2006a, 2006b)
showed that half of exports to the market of the United States of America did not survive
longer than two years. Similar tendency has been found in other studies such as Nitch (2009)
for German exports; Carballo and Volpe (2008) for Peruvian exports; Alvarez and Fuentes
(2009) for Chilean exports; Eaton, Eslava, Kugler and Tybout (2008) for Colombian exports;
and Gorg, Kneller and Murakozy (2008) for Hungarian exports. That is, to achieve successful
export diversification, we should focus on how to make new exports survive in the
international market.

In the economic literature, a number of studies looked into the patterns of export
duration and possible determinants of the survival of exports in their respective markets. Such
studies show that the likelihood of the survival of exports widely varies across sectors as well
as across different levels of development of exporters. Besedes and Prusa (2006a) find exports
from industrial countries last significantly longer than those from developing countries.

Our data suggest the same: Bilateral trade flows are characterized by lack of
continuity. An export spell – the number of continuous years during which a particular
product was exported from a LDC to an importing market – is of short duration, often lasting
only for one or two years.2 The short duration of export spells is common across all countries,
but particularly so for the exports of LDCs. The median survival time of exports from LDCs
is just one year, compared with two years of exports from developed and developing
countries. That is, at least half of the exports LDCs do not last more than one year. The
incidence rate (the rate of a product stops being exported during the studied period) of LDCs
is high at 41 per cent (table 1.1).


The export products are classified at the six-digit level of the Harmonized Commodity Description and Coding

System (HS).


Table 1.1. Export termination incidence rate




(per cent)

Median Survival Time



Developed countries
7 785

638 13 2 4

5 313
635 15 2 3

LDCs 245 840 41 1 1

Note: Authors' own calculations.

To further investigate the patterns of the survival of LDCs’ exports, we look into
export flows of 17 LDCs to 190 importing countries during a period of 14 years (1993–
2007).3 As is standard in survival analysis, we convert annual trade data into export spells.4
For example, Benin’s export of cotton to Austria from 1994 to 2002 without a break
constitutes one spell with the length of nine years. The total number of time periods for which
exports are observed is 464,251, which corresponds to 245,840 exporting spells of various

Table 1.2 provides the summary statistics of 17 LDCs. Forty-one per cent of the
observed export spells of all the studied LDCs ended during the sample period. The highest
incidence rate of exports observed is 67 per cent for the Gambia and the lowest is 29 per cent
for Bangladesh. The mean length of a spell for the whole sample is 1.88 year. Longer spells
are observed in two Asian countries (Bangladesh and Nepal), and the lowest is found in the
Gambia, Liberia and Rwanda. The median spell length is 1 across all LDCs in the sample,
which means at least half of exports from each country did not last longer than one year
during the studied period. The export duration at the 75th percentile varies between one year
for the Gambia and six years for Bangladesh.

So what determines the duration of exports of the export survival in the international
market? Brenton, Saborowski and von Uexkull (2009) show that learning-by-doing reduces
the risk of short life of exports from developing countries. Fugazza and Molina (2009) find
that high fixed costs linked to trade reduces the rate of survival of exports. Jaud, Kukenova
and Strieborny (2009) find that exports that do not bear a country’s comparative advantage
(e.g. labour-intensive products in a capital-abundant country) survive shorter in the United
States market.

This study investigates whether there is any interaction between the Heckscher–Ohlin
theorem and the chances of export survival. That is, we examine if a difference between a
product’s factor intensity and the exporting country’s factor endowments – or comparative
advantage – would have any impact upon the export duration of the product concerned.

For export statistics, we use mirror data from importing countries extracted from the UNCTAD South–South

Trade Information System. The 17 LDCs are selected on the basis of data availability.
Some of the available data required left- or right-censoring. Left-censoring implies that we exclude all the spells

that are observed in 1993 from the analysis, as we do not know how long these exports had been in existence prior
to 1993. Similarly, with regard to right-censoring, we do not know how long the exports that are observed in the
final year would continue to exist. The estimation technique of survival methods will take care of this right-
Whenever the interval between spells is just one year, there is a high probability of misreporting, i.e. trade is not

reported for that year. We correct this measurement error by merging spells with a one-year gap.


Table 1.2. Survival duration of least developed countries’ exports, by country


time at



(per cent)



Survival time

Mean* Median At 75th percentile
Benin 16 406 50 9 220 1.78 1 2
Bangladesh 113 636 29 46 770 2.43 1 6
Central African Republic 6 706 61 4 845 1.38 1 2
Gambia 8 989 67 6 607 1.36 1 1
Liberia 7 556 63 5 589 1.35 1 2
Mali 21 001 55 13 570 1.55 1 2
Mozambique 18 320 47 10 870 1.69 1 3
Malawi 16 309 46 9 314 1.75 1 3
Niger 14 749 61 10 601 1.39 1 2
Nepal 59 220 31 25 354 2.34 1 4
Rwanda 5 405 62 4 036 1.34 1 2
Sudan 15 601 54 10 173 1.53 1 2
Senegal 51 777 40 27 220 1.90 1 3
Sierra Leone 22 186 54 14 589 1.52 1 2
Togo 34 114 41 17 134 1.99 1 3
Uganda 26 764 42 15 136 1.77 1 3
Zambia 25 512 45 14 812 1.72 1 3

Total 464 251 41 245 840 1.88 1 3

Notes: * uncensored observations
Authors' own calculations.

The rest of this paper is organized as follows: Chapter 2 discusses the results derived
from our empirical analysis; chapter 3 concludes with some policy suggestions to LDCs.

2. Survival analysis findings

The object of interest in survival analysis is the length of spells and the identification
of the factors that make spells end. The survival analysis allows us to focus on the long-term
sustainability of trade relationships rather than year-to-year changes in the trade volumes and
to examine the relationship between the survival time and some covariates, i.e. factors that
may influence the survival.

In this study, we use two survival models. First, we use the Kaplan–Meier survival
function to examine general patterns of the survival of exports of LDCs. The survival function
gives the probability of a trade relationship (e.g. exports) surviving past a given time t.
Second, we use the Extended Cox Proportional Hazard model to examine the degree of
influence of the independent variables on the export survival. A detailed explanation of the
models we use is given in the annex.

2.1. Estimated survival pattern of least developed countries’ exports
(Kaplan–Meier analysis)

The Kaplan–Meier analysis on export spells of 17 LDCs suggest that more than 60
per cent of the sample is likely to disappear after the first year of exporting (figure 2.1). Only
8 per cent is estimated to survive to the end of the studied period.


Figure 2.1. Kaplan–Meier survival estimates
for least developed countries’ exports






0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Analysis time

However, the estimated length of export survival changes when the export spells are
grouped according to the market destination: (a) high-income countries; (b) middle-income
countries, and (c) low-income countries. As shown in figure 2.2, LDC exports to high-income
countries have higher survival rates than those to developing countries (i.e. middle- and low-
income countries). Over 40 per cent of the exports to high-income countries on average
survive more than one year, while only 37 per cent of the exports to low-income countries
survive after the first year. This outcome could be due to a combination of different reasons,
including a different composition of the LDC export basket to different markets, import
demand being more stable in higher-income countries than developing countries and
preferential market access conditions granted to LDCs’ exports in high-income countries.

Figure 2.2. Survival functions by export destination


2.2. Factors influencing the survival of least developed countries’ exports
(Extended Cox Proportional Hazards Model)

We turn to examining the factors that are likely to influence the likelihood of export
survival. We are particularly interested in the degree of influence, if any, of comparative
advantage of a country over a product’s export survival.

Our choice of covariates, i.e. factors that may determine the likelihood of the survival
of exports, are as follows: an exported product’s distance from the country’s comparative
advantage, the degree of market diversification, the degree of product diversification, initial
export values and the variables commonly included in the trade gravity model. These
covariates are explained below.

(a) Distance from comparative advantage

The key variable of interest is the distance from the exporting country’s comparative
advantage at the time when the product is exported. This is calculated as the Euclidian
distance between (a) the country’s factor endowments in human capital, physical capital and
natural resources of the exported year; and (b) the product’s revealed factor intensity (RFI),
using the RFI indices developed by Cadot, Shirotori and Tumurchudur (2010).6 A product is
“distanced” from a country’s comparative advantage when the product is not utilizing the
factor endowment that is relatively abundant in the country, or only utilizing the factor
endowment that are relatively scarce in the country. The methodology to calculate a product’s
distance from comparative advantage is given in the appendix.

(b) Market diversification and product diversification

Export survival would also depend on specific export capability of countries, which
can be measured by the degree of market diversification or product experience and product
diversification or market experience. Volpe and Carballo (2008), using disaggregated firm-
level export data of Peru for the period 2000–2006, find that both market diversification (i.e.
exporting the same products to many buyers) and product diversification (i.e. exporting many
products to the same buyer) increase the chances of surviving in foreign export markets.
Cadot, Iacovone, Rauch and Pierola (2010) also find that a firm’s product experience and
market experience matter for export survival. They used the number of destinations to which
the firm already exports that product as a proxy to product experience, and the number of
products the firm already exports to that market as market experience.

We control for market and product diversification in a similar way, but only at the
national level. Market diversification (or product experience) is measured by the number of
importing markets for a particular product exported by a country, and product diversification
(or market experience) is measured by the number of products that are exported by a country
to a particular market.

Based on Cadot, Shirotori and Tunurchudur (2008), UNCTAD constructed a database of revealed factor intensity

indices (RFII) of products at the HS 6-digit level of product classification. RFII approximates the revealed factor
content of a product as a weighted average of the national factor endowments of countries that export the product,
the weights being these exporters’ revealed comparative advantage indices. The database contains three indices of
RFI: Revealed physical capital intensity index, revealed human capital intensity index and revealed natural
resource intensity database. The database also includes the updated country endowment data (capital stocks per
worker, average years of schooling and arable land per worker) for a cross-section of 137 countries (countries that
have all three endowments) for the period 1961–2007. The database can be freely accessed online at


(c) Initial export values

Initial export values, i.e. the value of exports at the start of export spells, is a proxy to
the level of confidence between the trading partners. We expect that the higher the initial
trade value of the export, the higher the likelihood of the export to survive.

(d) Other gravity variables

We included the variables commonly used in a gravity model as the covariates in the
model. These are the following: exporting country’s gross domestic product (GDP) per capita;
importing country’s GDP per capita; geographical distance between the exporter and the
importer: a dummy for a common border; and a dummy for common language. The
coefficients on these variables are expected to be in line with those typically found in gravity
model regressions.

Table 2.1 shows the results derived from different semiparametric Cox models with
time-fixed effects to control for year-specific conditions that might influence export survival.
Before discussing the results, some discussion on the interpretation of coefficient is in order.
A coefficient above one implies that a positive correlation between the hazard rate and
changes in a covariate. A unit increase in the covariate would result in an increase in the
hazard rate, i.e. export survival becomes less likely. A coefficient below one implies a
negative correlation, i.e. a unit increase in the covariate results in a longer export survival.

As mentioned at the outset, this study is particularly interested in possible interaction
between the Heckscher–Ohlin theorem and the chances of export survival. That is, we
examine if an exported product based on the country’s comparative advantage (e.g.
intensively using input factors that are relatively abundant in the country) would have a better
chance to survive in the international market.

The estimation result finds a coefficient on the distance to comparative advantage
being greater than one. This implies that, as a product becomes farther away from the
exporting country’s comparative advantage, the hazard rate becomes higher so the export
survival period becomes shorter.

The survival rate also depends on supply capacity of the exporting country. As
discussed above, these are captured by two variables: market diversification and product
diversification. The coefficients on these variables are below one and statistically significant.
This implies that the hazard rate decreases as a country becomes more diversified in markets
and/or products.

In line with other previous studies, the gravity-model-motivated variables have the
expected signs. Exporting countries with larger GDP (exporter GDP), countries exporting to
larger economies (importer GDP), those having a common language (dummy for common
language) or sharing a border with a trading partner (dummy for border) all reduce the hazard
rate of exporting, i.e. increase the likelihood of export survival. As the geographical distance
to the trading partners (physical distance) increases, the hazard of exporting increases due to
the increase in the transport-related costs.


Table 2.1. Results derived from semiparametric Cox model
with gravity-type variables

Model 1 Model 2

Distance to comparative advantage (log) 1.035*** 1.019**
(0.008) (0.008)

Market diversification (log) 0.811***

Product diversification (log) 0.883***

Initial export (log) 0.959*** 0.959***
(0.001) (0.001)

Exporter GDP (log) 0.905*** 1.032***
(0.009) (0.008)

Importer GDP (log) 0.953*** 0.991***
(0.002) (0.002)

Physical distance (log) 1.032*** 0.978***
(0.005) (0.004)

Dummy for common language 0.945*** 0.950***
(0.006) (0.005)

Dummy for border 0.751*** 0.823***
(0.007) (0.008)

Multispell 1.101*** 1.194***
(0.008) (0.01)
Fixed effect by

year Yes Yes
Stratified by

HS 6-digit Yes Yes
Number of observations 140 385 140 385
Wald chi2 2 176 796 2 505 962
Log pseudolikelihood - 380 077 - 378 103

Notes: Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1

Besides gravity variables, the models also control for the starting level of confidence
between the trading partners. This is done by adding the initial value of export as a control
variable. The result indicates that the higher is the original confidence of the trading partners
the longer the export duration.

In the results of table 2.1 above, country-specific characteristics and trade costs are
controlled for by gravity- type variables. In order to check the robustness of the result, i.e. to
better control for a wide array of observable and unobservable characteristics of the exporting
and the importing countries that might affect the hazard rate, we estimate the model with the
inclusion of fixed effects by exporter and importer on the same sample. The results are
reported in table 2.2. An additional advantage of the use of fixed effects is that it allows to
considerably extending the dataset, as it does not require GDP data, which are not always
available for our set of LDCs. Sample selection does not appear to affect the overall results as
the coefficients of the variables of interest are very similar.

The results presented so far treat the covariates as time-independent. That is, the
results represent a kind of average effects over the observation period. However, the effect
arising from changes in the covariates may be time-variant. We now examine whether the
impact of covariates on export survival changes through time by splitting the data at every
observed failure time, i.e. at every export year and for each spell. This enables us to estimate
the Cox model with the changing effects of covariates through time.


Table 2.2. Results derived from semiparametric Cox model
Model 1 Model 2
Distance to comparative advantage (log) 1.026*** 1.012**

(0.006) -(0.005)
Market diversification (log) 0.819***

Product diversification (log) 0.899***

Initial export (log) 0.959***

Multispell 1.154***
Fixed effect by

exporter Yes Yes
importer Yes Yes
year Yes Yes

Stratified by
HS 6-digit Yes Yes

Observations 245 840 245 840
Log pseudolikelihood -780 549 -777 626
Notes: Robust standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%

The results derived from the Cox regression with time-varying covariates are
presented in table 2.3. The initial effect of the distance to comparative advantage on export
survival at the very start of an export spell (time t0) now loses its significance. However, the
interaction terms, i.e. the changing effects of covariates through time, are all significant at the
0.01 level, which confirms the violation of proportional hazard assumption for the covariates.
The coefficient of the distance to comparative advantage interacted with time is above one.
This implies that the impact of the distance from comparative advantage over the hazard rate
increases (or its impact upon export survival decreases) through time.

Table 2.3. Results derived from Cox model with time interactions
Model with time-varying covariates

exponential decay hazard rate
Distance to comparative advantage (log) 1.007 0.007

(0.005) (0.005)
Market diversification (log) 0.835*** -0.180***

(0.003) (0.004)
Product diversification (log) 0.906*** -0.099***

(0.002) (0.002)
Initial export (log) 0.959*** -0.042***

(0.001) (0.001)
Multispell 1.151*** 0.140***
(0.007) (0.006)
Distance to comparative advantage 1.032* 0.032*
(log)*time (log) (0.017) (0.017)
Market diversification 0.914*** -0.090***
(log)*time (log) (0.006) (0.007)
Product diversification 0.947*** -0.055***
(log)*time (log) (0.004) (0.004)
Observations 464 251 464 251
Log pseudolikelihood - 777 451 - 777 451
Notes: Robust standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%


The increasing impact of the distance from comparative advantage through time is
shown in figure 2.3. The hazard rate, i.e. the likelihood of an export being terminated, of a
product whose factor contents do not reflect the country’s comparative advantage increases
through time.

A negligible impact of comparative advantage on the hazard rate in the first few years
may suggest that when countries select new exportable products, they do not take into
consideration the underlying comparative advantage of the country. Also, for the initial period
of exports, the exporting firms may try to stay in the market to absorb fixed costs, to acquire
better knowledge of the market, or because of a presence of government support schemes.
Once these supporting factors end, however, lack of competitiveness of the product arising
from weak comparative advantage would make exports unsustainable.

Figure 2.3. Time-varying effect of distance to comparative advantage
on export survival








1 2 3 4 5 6 7 8 9 10 11 12 13 14

Time effect

We then check if the impact of comparative advantage on export survival is
significant only for LDCs, or the same applies to other groups of developing countries. In
order to assess this, we estimate the same Cox model for the member countries of the
Association of Southeast Asian Nations (ASEAN) and those of the Southern Common Market
(MERCOSUR). These countries are at a rather different level of development from LDCs,
and their economic structure, export capacity and technological know-how may be dynamic
enough to make distance to comparative advantage less critical.

Table 2.4 compares the results for ASEAN, MERCOSUR and LDCs. The signs and
the magnitude of the impacts arising from market or product diversification upon the export
hazard rate are similar across three groups, except the impact arising from comparative
advantage. In the case of ASEAN, the coefficient of the distance to comparative advantage is
negatively correlated with the hazard rate under the time-independent setup, then the
coefficient loses its statistical significance once the interaction term with time is added.


Table 2.4. Results derived from Cox model with time interactions,

Model with time-varying covariates
Distance to comparative advantage (log) 0.983* 0.934*** 1.007

(0.01) (0.014) (0.005)
Market diversification (log) 0.878*** 0.889*** 0.835***

(0.005) (0.007) (0.003)
Product diversification (log) 0.861*** 0.986** 0.906***

(0.004) (0.006) (0.002)
Initial export (log) 0.926*** 0.925*** 0.959***

(0.001) (0.001) (0.001)
Multispell 1.795*** 1.638*** 1.151***
-0.01 -0.009 (0.007)
Distance to comparative advantage (log) *time(log) 1.008 1.064*** 1.032*

(0.011) (0.019) (0.017)
Market diversification (log) *time(log) 0.805*** 0.849*** 0.914***

(0.005) (0.006) (0.006)
Product diversification (log) *time(log) 0.949*** 0.897*** 0.947***
(0.004) (0.005) (0.004)
Observations 2 882 511 1 151 673 464 251

Log pseudolikelihood -2 963 905 -1 000 507
- 777

Notes: Robust standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%

As for MERCOSUR, the relationship between distance to comparative advantage and
export hazard rate follows the same trend as that of LDCs. The time-varying effect of
comparative advantage over the hazard rate through time of MERCOSUR and of LDCs is
illustrated in figure 2.4. We find that the effect of the distance to comparative advantage on
export survival in the first years of export is greater for LDCs than for MERCOSUR. Only
after about 10 years, the distance to comparative advantage has a similar impact on export
survival across the two groups of countries.

Further research is required to examine why the effect of comparative advantage on
export survival is different across these country groups. Our educated guess at this stage is
that the impact of comparative advantage in a traditional sense, i.e. human capital, physical
capital and natural resources, is different across different product groups. There is a huge
divergence in the export composition between the studied LDCs and ASEAN. For instance,
the survival of exports of intermediary inputs (e.g. parts and components of electric devices),
which represent a large share of ASEAN’s exports, may depend less on the relative
abundance of human or physical capital and natural resources and more on the existence of
export-facilitating institutions or infrastructure in their countries that enhance them being a
part of a global supply chain.


Figure 2.4. Time-varying effect of distance to comparative advantage,
least developed countries and MERCOSUR






1 2 3 4 5 6 7 8 9 10 11 12 13 14

Time effect: LDC Time effect: MERCOSUR

3. Conclusions

Our study on LDCs’ export spells finds that exported products that do not closely
reflect the country’s comparative advantage at the time when the product is exported are
likely to survive only for a short duration in international market. Moreover, the impact of the
country’s comparative advantage on LDCs’ export survival increases through time. Products
outside the comparative advantage frontier are most likely to lack competitiveness in the
international market. Thus sustaining export flows of such products becomes increasingly
more difficult with time.

These findings lead to two implications to export diversification strategy of LDCs.
First, in the short run, the government needs to pay close attention to whether a particular
sector (or product) of export interest reflects the country’s comparative advantage. Second, in
order for LDCs to successfully diversify into non-traditional sectors that bear higher value
added or more sophisticated factor contents in the long-run, their policy priority should be to
constantly improve their comparative advantage. Comparative advantage of a country,
particularly human capital, evolves dynamically through the years. Policies designed to
increase both the quantity and the quality of human capital endowment (e.g. managerial skills
and technological know-how), as well as to improve trade-facilitating institutions and
infrastructure, would help increase the likelihood of successful export diversification into
non-traditional product sectors.

Although suggestive, the analysis calls for further research. In particular, the effect of
comparative advantage on export survival appears to differ across countries, and this call for a
more thorough analysis. The importance of comparative advantage in explaining export
survival may be related to differences in the export basket (i.e. the impact of comparative
advantage upon export survival varies across product sectors) or, more generally, comparative
advantage matters less and less as the overall economic capacity of a country increases (e.g.
the availability of well-functioning trade-related institutions). A deeper investigation into
these issues should help in the development of more tailored policy options for LDCs.


Appendix: Model specification

1. Kaplan–Meier analysis (nonparametric approach)
Export survival patterns can be characterized by the Kaplan–Meier estimator, which

is a nonparametric estimate of the survivor function )(ts or the probability of survival past
time t. It is calculated as


ttj j


j n



Where jn is the number of spells at risk at jt , and jd is the number of completed
spells of duration jt . The Kaplan–Meier survival estimator gives the estimated probability of
completing a spell at duration jt , given that the spell has reached duration jt .

2. Extended Cox Proportional Hazards Model (semiparametric approach)
To do so, we adopt a classic extended Cox Proportional Hazards Model. In this

model, the distribution of durations can be characterized in terms of the hazard function. Let ( )tTttTtP ≥∆+≤ be the probability of an export flow to be terminated in the interval
[ ]t)t(t, ∆+ , given that it has lasted until time t. The hazard function is obtained by taking the
limit of this probability for t∆ .

( ) )()(lim)(




The hazard rate )(th is the probability that an export flow disappears at the duration t,
given that they lasted until t. The independent variables can affect the distribution of the
durations in several ways, depending on the specification used. The Cox Proportional Hazards
Model is a semiparametric approach where no parametric form of the baseline function is
specified, yet the effects of the independent variables are parameterized to alter the baseline
survival function in a way that the independent variables multiplicatively shift the baseline
hazard function.

The hazard function is as the following: )exp()()|( '0 βii XthXth =
where the baseline hazard rate )(0 th , which is not influenced by any independent variable,
and each sample's hazard rate )(thi that is a product of a function of the independent variables
and the baseline hazard rate.

The advantage of this approach is that it helps us avoid potential misspecification of
the hazard function.

The above function can be rewritten as: )(



XthX ii =β

This hazard ratio gives the degree of effect created by one unit change in the
independent variables on export survival. A hazard ratio that is smaller than one means that


the independent variables, or covariates, have a reducing impact on the hazard rate, leading to
a longer export survival.

One issue with using Cox regression is that it is based on the Proportional Hazard
Assumption, whose implication is that the explanatory variables have the same effects at all
points in time, meaning there is no interaction with time. There are, in general, two ways to
deal with the violation of Proportional Hazard Assumption: (a) by stratifying or (b) by
including the time interactions. We have used both methods.

First, the Cox model is extended to allow for non-proportional hazards for different
products at the six-digit level of the Harmonized System using the method of stratification to
account for different strata such that the stratified Cox Proportional hazards model for any
country i and product j belonging to a stratum k is then given by the following model:

)exp()()|( ' βijkijij XthXth =
This extended model allows the baseline hazards to be arbitrary and unrelated for the

different strata to control for unobserved differences across products. The large number of
observations in our data set allows for a fine stratification. Unobserved differences across
countries and years were dealt with by a fixed effects model.

Second, the Cox model is extended further to include time-interaction terms for time-
dependent explanatory variables (covariates). A Cox model with time-varying covariates
allows us to test whether the effect of covariates on the hazard rates changes through time. To
incorporate time-varying covariates, the hazard rate is modelled as a step function, with
different values of the covariates through intervals between t = 0 and t = T. T is the terminal
value for the observation. A simplified Cox Extended Model with two covariates, one time-
independent and one time-varying may be written:

))(exp()()( 2211 txxthth k ββ +=
where the hazard rate at time t depends on the value of 2x .

Lastly, some exports may have several periods of continuous exporting that may be
interrelated. For example, an exporter exports a certain product to a certain importer for a
while, exit and re-exports again. As such consecutive exports may be interrelated, we use a
dummy for multiple spells for an exporter-importer product. About 29 per cent of the spells of
the study sample have multiple spells.

3. Estimating a product’s distance from the exporting country’s comparative

The formula of the Euclidean distance of a product k to the comparative advantage of
country c for any given year, omitting the time index, is the following:

222 )ˆ()ˆ()ˆ( kckckcck llstdkkstdhhstdD −+−+−=

where ccc lkh ,, are the country’s human-capital, capital and land endowments and

kkk lkh ˆ,ˆ,ˆ are the revealed factor intensities of product k. We use ckD as the measure of the
closeness of each export product to the country’s comparative advantage. We followed the
methodology applied by Jaud, Kukenova and Strieborny (2009) to assess the role of
comparative advantages in the survival of the exports of LDCs.



Alvarez R and Fuentes R (2009). Entry into Export Markets and Product Quality Differences.
Documentos de Trabajo 367. Instituto de Economía. Pontificia Universidad Católica
de Chile.

Besedes T and Prusa T (2006a). Ins, outs, and the duration of trade, Canadian Journal of
Economics. No. 104.

Besedes T and Prusa T (2006b). Product differentiation and duration of U.S. imports. Journal
of International Economics. 70:339–358.

Brenton P, Saborowski C and von Uexkull E (2009). What Explains the Low Survival Rate of
Developing Country Export Flows? Policy Research Working Paper 4951. The World

Cadot O, Shirotori M and Tumurchudur B (2010). Revealed Factor Intensity Indices at the
Product Level. UNCTAD Policy Issues in International Trade. Study Series No. 44.

Cadot O, Iacovone L, Rauch F and Pierola D (2010). Success and Failure of African
Exporters. World Bank Policy Research Working Paper.

Carballo J and Volpe C (2008). Survival of New Exporters: Does it Matter How They
Diversify? IDB Working Paper 04. Inter-American Development Bank.

Eaton J, Eslava M, Kugler M and Tybout J (2008). The margins of entry into export markets:
Evidence from Colombia. In: Helpman E, Martin D and Verdier T, Eds. The
Organization of Firms in a Global Economy. Harvard University Press. Cambridge,

Fugazza M and Molina A C (2009). On the Determinants of Exports Survival. HEI Working
Papers 05-2009. The Graduate Institute of International Studies, Economic Section.

Gorg H, Kneller R and Murakozy B (2008). What Makes a Successful Exporter? CEPR
Discussion Paper 6614.

Jaud M, Kukenova M and Strieborny M (2009). Financial Development and Intensive Margin
of Trade. Paris School of Economics Working Paper 2009-35.

Nitch V (2007). Die another day: Duration in German import trade. Review of World
Economics. 145(1):133–154.


UNCTAD study series on


No. 1 Erich Supper, Is there effectively a level playing field for developing country
exports?, 2001, 138 p. Sales No. E.00.II.D.22.

No. 2 Arvind Panagariya, E-commerce, WTO and developing countries, 2000, 24 p.
Sales No. E.00.II.D.23.

No. 3 Joseph Francois, Assessing the results of general equilibrium studies of
multilateral trade negotiations, 2000, 26 p. Sales No. E.00.II.D.24.

No. 4 John Whalley, What can the developing countries infer from the Uruguay
Round models for future negotiations?, 2000, 29 p. Sales No. E.00.II.D.25.

No. 5 Susan Teltscher, Tariffs, taxes and electronic commerce: Revenue implications
for developing countries, 2000, 57 p. Sales No. E.00.II.D.36.

No. 6 Bijit Bora, Peter J. Lloyd, Mari Pangestu, Industrial policy and the WTO, 2000,
47 p. Sales No. E.00.II.D.26.

No. 7 Emilio J. Medina-Smith, Is the export-led growth hypothesis valid for
developing countries? A case study of Costa Rica, 2001, 49 p. Sales No.

No. 8 Christopher Findlay, Service sector reform and development strategies: Issues
and research priorities, 2001, 24 p. Sales No. E.01.II.D.7.

No. 9 Inge Nora Neufeld, Anti-dumping and countervailing procedures – Use or
abuse? Implications for developing countries, 2001, 33 p. Sales No. E.01.II.D.6.

No. 10 Robert Scollay, Regional trade agreements and developing countries: The case
of the Pacific Islands’ proposed free trade agreement, 2001, 45 p. Sales No.

No. 11 Robert Scollay and John Gilbert, An integrated approach to agricultural trade
and development issues: Exploring the welfare and distribution issues, 2001,
43 p. Sales No. E.01.II.D.15.

No. 12 Marc Bacchetta and Bijit Bora, Post-Uruguay round market access barriers for
industrial products, 2001, 50 p. Sales No. E.01.II.D.23.

No. 13 Bijit Bora and Inge Nora Neufeld, Tariffs and the East Asian financial crisis,
2001, 30 p. Sales No. E.01.II.D.27.

No. 14 Bijit Bora, Lucian Cernat, Alessandro Turrini, Duty and quota-free access for
LDCs: Further evidence from CGE modelling, 2002, 130 p. Sales No.


No. 15 Bijit Bora, John Gilbert, Robert Scollay, Assessing regional trading
arrangements in the Asia-Pacific, 2001, 29 p. Sales No. E.01.II.D.21.

No. 16 Lucian Cernat, Assessing regional trade arrangements: Are South-South RTAs
more trade diverting?, 2001, 24 p. Sales No. E.01.II.D.32.

No. 17 Bijit Bora, Trade related investment measures and the WTO: 1995-2001, 2002.

No. 18 Bijit Bora, Aki Kuwahara, Sam Laird, Quantification of non-tariff measures,
2002, 42 p. Sales No. E.02.II.D.8.

No. 19 Greg McGuire, Trade in services – Market access opportunities and the
benefits of liberalization for developing economies, 2002, 45 p. Sales No.

No. 20 Alessandro Turrini, International trade and labour market performance: Major
findings and open questions, 2002, 30 p. Sales No. E.02.II.D.10.

No. 21 Lucian Cernat, Assessing south-south regional integration: Same issues, many
metrics, 2003, 32 p. Sales No. E.02.II.D.11.

No. 22 Kym Anderson, Agriculture, trade reform and poverty reduction: Implications
for Sub-Saharan Africa, 2004, 30 p. Sales No. E.04.II.D.5.

No. 23 Ralf Peters and David Vanzetti, Shifting sands: Searching for a compromise in
the WTO negotiations on agriculture, 2004, 46 p. Sales No. E.04.II.D.4.

No. 24 Ralf Peters and David Vanzetti, User manual and handbook on Agricultural
Trade Policy Simulation Model (ATPSM), 2004, 45 p. Sales No. E.04.II.D.3.

No. 25 Khalil Rahman, Crawling out of snake pit: Special and differential treatment
and post-Cancun imperatives, 2004.

No. 26 Marco Fugazza, Export performance and its determinants: Supply and demand
constraints, 2004, 57 p. Sales No. E.04.II.D.20.

No. 27 Luis Abugattas, Swimming in the spaghetti bowl: Challenges for developing
countries under the “New Regionalism”, 2004, 30 p. Sales No. E.04.II.D.38.

No. 28 David Vanzetti, Greg McGuire and Prabowo, Trade policy at the crossroads –
The Indonesian story, 2005, 40 p. Sales No. E.04.II.D.40.

No. 29 Simonetta Zarrilli, International trade in GMOs and GM products: National
and multilateral legal frameworks, 2005, 57 p. Sales No. E.04.II.D.41.

No. 30 Sam Laird, David Vanzetti and Santiago Fernández de Córdoba, Smoke and
mirrors: Making sense of the WTO industrial tariff negotiations, 2006, Sales
No. E.05.II.D.16.

No. 31 David Vanzetti, Santiago Fernandez de Córdoba and Veronica Chau, Banana split:
How EU policies divide global producers, 2005, 27 p. Sales No. E.05.II.D.17.

No. 32 Ralf Peters, Roadblock to reform: The persistence of agricultural export
subsidies, 2006, 43 p. Sales No. E.05.II.D.18.


No. 33 Marco Fugazza and David Vanzetti, A South-South survival strategy: The
potential for trade among developing countries, 2006, 25 p.

No. 34 Andrew Cornford, The global implementation of Basel II: Prospects and
outstanding problems, 2006, 30 p.

No. 35 Lakshmi Puri, IBSA: An emerging trinity in the new geography of international
trade, 2007, 50 p.

No. 36 Craig VanGrasstek, The challenges of trade policymaking: Analysis,
communication and representation, 2008, 45 p.

No. 37 Sudip Ranjan Basu, A new way to link development to institutions, policies and
geography, 2008, 50 p.

No. 38 Marco Fugazza and Jean-Christophe Maur, Non-tariff barriers in computable
general equilibrium modelling, 2008, 25 p.

No. 39 Alberto Portugal-Perez, The costs of rules of origin in apparel: African
preferential exports to the United States and the European Union, 2008, 35 p.

No. 40 Bailey Klinger, Is South-South trade a testing ground for structural
transformation?, 2009, 30 p.

No. 41 Sudip Ranjan Basu, Victor Ognivtsev and Miho Shirotori, Building trade-relating
institutions and WTO accession, 2009, 50 p.

No. 42 Sudip Ranjan Basu and Monica Das, Institution and development revisited: A
nonparametric approach, 2010, 26 p.

No. 43 Marco Fugazza and Norbert Fiess, Trade liberalization and informality: New
stylized facts, 2010, 45 p.

No. 44 Miho Shirotori, Bolormaa Tumurchudur and Olivier Cadot, Revealed factor
intensity indices at the product level, 2010, 55 p.

No. 45 Marco Fugazza and Patrick Conway, The impact of removal of ATC Quotas on
international trade in textiles and apparel, 2010, 50 p.

No. 46 Marco Fugazza and Ana Cristina Molina, On the determinants of exports
survival, 2011, 40 p.

No. 47 Alessandro Nicita, Measuring the relative strength of preferential market
access, 2011, 30 p.

No. 48 Sudip Ranjan Basu and Monica Das, Export structure and economic
performance in developing countries: Evidence from nonparametric
methodology, 2011, 58 p.

No. 49 Alessandro Nicita and Bolormaa Tumurchudur-Klok, New and traditional trade
flows and the economic crisis, 2011, 22 p.

No. 50 Marco Fugazza and Alessandro Nicita, On the importance of market access for
trade, 2011, 35 p.


No. 51 Marco Fugazza and Frédéric Robert-Nicoud, The 'Emulator Effect' of the
Uruguay Round on US regionalism, 2011, 45 p.

No. 52 Sudip Ranjan Basu, Hiroaki Kuwahara and Fabien Dumesnil, Evolution of non-
tariff measures: Emerging cases from selected developing countries, 2012, 38p.

No. 53 Alessandro Nicita and Julien Gourdon, A preliminary analysis on newly collected
data on non-tariff measures, 2013, 31 p.

No. 54 Alessandro Nicita, Miho Shirotori and Bolormaa Tumurchudur Klok, Survival
analysis of the exports of least developed countries: The role of comparative
advantage, 2013, 25 p.

Copies of the UNCTAD study series, Policy Issues in International Trade and Commodities,
may be obtained from the Publications Assistant, Trade Analysis Branch (TAB), Division on
International Trade in Goods and Services and Commodities (DITC), United Nations
Conference on Trade and Development, Palais des Nations, CH-1211 Geneva 10, Switzerland
(Tel: +41 22 917 4644). These studies are accessible on the website at

Since 1999, the Trade Analysis Branch of the Division on International Trade in Goods and
Services, and Commodities of UNCTAD has been carrying out policy-oriented analytical work
aimed at improving the understanding of current and emerging issues in international trade and
development. In order to improve the quality of the work of the Branch, it would be useful to
receive the views of readers on this and other similar publications. It would therefore be greatly
appreciated if you could complete the following questionnaire and return to:

Trade Analysis Branch, DITC
Rm. E-8065

United Nations Conference on Trade and Development
Palais des Nations, CH-1211 Geneva 10, Swi erland

(Fax: +41 22 917 0044; E-mail: tab@unctad.org)

1. Name and address of respondent (optional):

2. Which of the following describes your area of work?

Government Public enterprise

Private enterprise institution Academic or research

International organization Media

Not-for-profi t organization Other (specify) _________________

3. In which country do you work? _________________________________________

4. Did you fi nd this publication Very useful Of some use Li" le use
to your work?

5. What is your assessment of the contents of this publication?
Excellent Good Adequate Poor

6. Other comments:


UNCTAD Study series on



(Study series no. 54: Survival analysis of the exports of least developed countries:
The role of comparative advantage)

Readership Survey