Available via license: CC BY 4.0
Content may be subject to copyright.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
39
Firm-Level Determinants of Export
Performance in Kenya’s Manufacturing Sector
1*Dorothy Ngina Kimolo
Post Graduate Student, School of Business,
Economics and Tourism, Kenyatta University,
Kenya
2Dr. Jennifer Njaramba
Lecturer, School of Business, Economics and
Tourism, Kenyatta University, Kenya
3Dr. Laban Chesang’
Lecturer, School of Business and Economics,
Daystar University, Kenya
Article History
Received 17th March 2024
Received in Revised Form 15th April 2024
Accepted 3rd May 2024
How to cite in APA format:
Kimolo , D., Njaramba, J., & Chesang’, L. (2024). Firm-
Level Determinants of Export Performance in Kenya’s
Manufacturing Sector. International Journal of
Economics, 9(2), 39–64.
https://doi.org/10.47604/ijecon.2529
Abstract
Purpose: This study aimed at investigating the firm-
level determinants of export performance (export
propensity and export intensity) in Kenya’s
manufacturing sector using firm-level panel data
obtained from the World Bank Enterprise Surveys for
the periods 2007, 2013 and 2018.
Methodology: The study adopted a quantitative non-
experimental research design. The Heckman Two-
Stage estimation procedure was employed to jointly
establish the firm-level determinants of export
propensity and export intensity in Kenya’s
manufacturing sector.
Findings: Based on the estimation results, firm-level
total factor productivity, firm size, human capital, cost
of material, electricity cost and foreign ownership had
positive and significant effects on firms’ export
propensity while labor productivity negatively
influenced export propensity. Firm age, capital
intensity and research did not have significant effects
on export propensity. On the other hand, export
intensity was positively influenced by firm-level total
factor productivity, foreign ownership, firm size, firm
age, human capital and research. Labor productivity
had a negative effect on firms’ export intensity.
Whereas the effect of energy cost on export intensity
was weakly significant at 10 percent level of
significance, there was no significant effect of cost of
material on export intensity.
Unique Contribution to Theory, Practice and
Policy: Employing the new ‘new’ trade theory, the
study tested the self-selection hypothesis by analyzing
the determinants of export propensity and intensity.
According to the self-selection hypothesis, one of the
key positive determinants of export propensity and
export intensity is firm-level total factor productivity.
The study findings validated the self-selection
hypothesis since the results revealed firm-level total
factor productivity as a positive and significant
determinant of both export propensity and export
intensity for Kenya’s manufacturing firms. According
to the study's conclusions, the government and
enterprises must focus on policies that increase firm-
level total factor productivity, firm size, human capital,
and research in order to improve firms' export
performance.
Keywords: Export Propensity, Export Intensity, Self-
Selection Hypothesis, Total Factor Productivity,
Manufacturing Sector
JEL Classification: F14, D22, D24, L60.
©2024 by the Authors. This Article is an open access
article distributed under the terms and conditions of
the Creative Commons Attribution (CC BY) license
(http://creativecommons.org/licenses/by/4.0/
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
40
INTRODUCTION
Globally, international trade is vital to economic expansion since it allows international
participants to enhance their competitiveness and multiply their output and profits across their
domestic borders. Exposing firms to international trade improves their competitiveness,
productivity and innovation (Kasahara & Lapham, 2013). Enhancing a country’s trade
performance has become necessary to improving its economic performance and this requires
expanding manufacturing exports, since around 70 per cent of world exports are manufactured
(World Bank, 2023). The average percentage of worlds’ manufactured goods exported in total
merchandise exports over the period 2007-2022 was 67.72 (World Bank, 2023). This is an
indication that the manufacturing sector accounts for over two-thirds of the total world exports
hence its importance in achieving economic transformation via exports. Over the same period,
Sub-Saharan Africa (SSA) and the East African Community (EAC) had an average share of
manufactured exports in all exports of 24.54 per cent and 17.46 per cent, respectively. These
statistics imply that the share for SSA and EAC is way below the world average hence the need
for improvement if the countries in these regions are to industrialize.
Universally, the manufacturing sector contributes significantly to economic growth and
development through fostering and maintaining high productivity growth, expanding job
possibilities, and boosting national competitiveness via exports alongside other forms of
international commerce (KAM, 2022; KCCB, 2021). A vibrant manufacturing sector generates
interlinkages with other sectors, promotes industrial revolution and productivity gains hence
spurring economic development as evidenced by the Industrial Revolution and the East Asian
Miracle (KAM, 2022; Republic of Kenya, 2012). International trade is one way of boosting the
performance of the manufacturing sector (Bernard & Jensen, 1999). According to traditional
trade theories, international trade boosts specialization within sectors based on comparative
advantage leading to welfare gains whereas new trade theory argues that trade yields
productivity gains due to increased product variety and economies of scale (Bernard, Jensen,
Redding, & Schott, 2007). As such it is imperative to explore the determinants of export
performance by firms for proper policy formulation. The study explored two dimensions of
export performance, namely export propensity and export intensity. Export propensity refers
to whether or not a firm participates in exporting while export intensity represents the share of
a firm’s exports in its total sales.
The linkage between global commerce and economic performance both at the country and firm
level has been a popular subject (Charles & Richard, 2020). Mostly researchers have explored
the effect of international trade participation on the firm-level performance, mostly
productivity which were commenced by (Bernard, Jensen, & Lawrence, 1995) for US
manufacturing industries where exporters outperformed non-exporters in terms of productivity
growth. Two hypotheses have been put forth to explain the link between exporting and firm-
level productivity: The self-selection hypothesis and the learning-by-exporting hypothesis
(Bernard & Jensen, 1999; Bernard, Jensen, Redding, & Schott, 2007). According to the self-
selection hypothesis, since there exist additional costs of exportation, only more productive
firms participate in exporting activities. As such, firm-level total factor productivity is a key
determinant of export performance. On the other hand, the learning-by-exporting implies that
once firms start exporting, their performance is enhanced. There exists mixed and inconclusive
evidence on the two hypotheses especially for developing countries where the literature is
scanty. The current study focused on the self-selection hypothesis by exploring the firm-level
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
41
determinants of export performance by manufacturing firms in Kenya using the Heckman two-
stage sample selection model.
The Role and Performance of the Manufacturing Sector in Kenya
In Kenya and other developing countries, the manufacturing sector, due to its strong
interlinkages with other sectors, has a higher potential and stability for stimulating economic
growth and development compared to agriculture and service sectors (KAM, 2019; KAM,
2021). This implies that the development of the manufacturing sector is key to ensuring a stable
and sustainable economic growth as emphasized in the Kenya Vision 2030 whose goal is to
create a diverse, robust, and competitive manufacturing industry (Republic of Kenya, 2007).
According to the Kenya Vision 2030, the manufacturing sector is among the key areas in
obtaining an industrialized status as a nation especially through export promotion strategies.
Since gaining its independence, Kenya has pursued several policies aimed at promoting
international trade. This was witnessed in the 1980s and 1990s where the policy focus shifted
from a regime of import-substitution to outward-oriented strategies. The export promotion
policies included: Manufacture Under Bond (MUB); Export Compensation Scheme; Export
Processing Zones (EPZ); Export Promotion Programme Office (EPPO), Tax Remission for
Exports Office (TREO) and National Exports Development and Promotion Strategy (NEDPS)
among others (Republic of Kenya, 2012; Republic of Kenya, 2017). Kenya has also prioritized
trade promotion especially through national trade commitments at the ‘World Trade
Organization (WTO), East African Community (EAC), Common Market for East and Southern
Africa (COMESA), Tripartite Free Trade Area (TFTA), African Continental Free Trade Area
(AfCFTA), East African Community-European Union Economic Partnership Arrangement,
African Growth and Opportunity Act (AGOA) among others’ (Republic of Kenya, 2017).
Nevertheless, export performance of the manufacturing sector in Kenya, has been below
expectations and set targets. The average share of manufactured exports in all exports was
about 32.37 per cent for the period 2007-2022 (World Bank, 2023). This share falls short of
the targeted 60 per cent as per the National Exports Development and Promotion Strategy
(NEDPS) (Republic of Kenya, 2017). Moreover, according to the World Bank Enterprise
Survey (WBES) of 2018, the number of manufacturing firms engaging in exporting activities
has been decreasing as evidenced by a declining ratio of exporting firms to total firms surveyed
from 52 per cent in 2013 to 45 per cent 2018. KAM’s 2022-2027 manifesto also aims at
increasing exporting activities by manufacturing firms in Kenya so as to boost the performance
of the sector. It is therefore important to establish the firm-level determinants of export
performance for proper policy guidance.
Statement of the Problem
The Industrial Revolution and the East Asian Miracle are key success stories of how the
manufacturing industry contributes significantly to economic development and growth through
fostering and maintaining productive growth, expanding job opportunities, and improving
nations' competitiveness by trading abroad (KAM, 2021). By 2022, the National Export
Development and Promotion Strategy targeted manufactured exports to account for 60 per cent
of all exports (Republic of Kenya, 2017). More so, Kenya is committed to various regional and
international trade agreements so as to enhance her export performance. However, despite all
the government efforts, the set targets have not been achieved given that, Kenya’s
manufactured exports accounted for 33 per cent of all exports (below the target of 60 per cent)
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
42
from 2007 to 2022 on average. Furthermore, based on the WBES data, the share of exporting
firms declined from 52 per cent in 2013 to 45 per cent 2018.
Based on these statistics as well as the commendable efforts by the government towards export
promotion in the manufacturing sector, for further policy guidance, it is important to analyze
the firm level determinants of export performance which the current study pursued. There is
scanty literature on the same for Kenya with those available focusing on different contexts as
well as exhibiting methodological limitations (Okado, 2013; Bresnaham, Coxhead, Foltz, &
Mogues, 2016; Chebor, 2020; Esaku, 2020). This study therefore sought to add to the existing
corpus of literature by establishing the firm-level determinants of export propensity and export
intensity for Kenya’s manufacturing firms using the Heckman Sample Selection model.
LITERATURE REVIEW
Theoretical Review
New Trade and ‘New’ New Trade Theories
Paul Krugman pioneered a series of international trade models known as New Trade Theory
(NTT) in the late 1970s and early 1980s. It is based on the following assumptions: Imperfect
markets; economies of scale and product differentiation. It emphasizes the importance of
network effects and increasing returns to scale. Contrary to the arguments of the traditional
trade models, NTT suggests that international trade primarily occurs between nations that share
similar factor endowments, structural characteristics, and levels of development. To describe
international commerce, traditional trade models depended on variations in factor endowment
or productivity. NTT showed that trade flows between similar countries can be driven by
increasing returns, without differences in factor endowments and productivity (Krugman,
1979). Trade enables the nations to take advantage of greater economies of scale. NTT among
other contributions describes the possibility of the existence of intra-industry trade. Krugman
(1979) enhanced the traditional theories by incorporating imperfect markets, economies of
scale and product differentiation in his analysis of trade. As such, exporting firms are able to
produce a wide variety of goods for exports due to product differentiation and economies of
scale. Hence, according to NTT, regardless of homogenous tastes, technology and factor
abundance, countries can engage in trade and boost firm’s performance contrary to the opinion
of the traditional trade theories.
Melitz (2003) extended Krugman’s (1979) model and came up with the 'new' new trade theory
(NNTT). NNTT incorporated the aspect of firm level productivity differences and focused
more on the firms rather than sectors in understanding the relationship between global trade
and business productivity (Melitz, 2003). Since entry in to new export markets is very costly,
only efficient firms are able to enter these markets and reap the benefits there of. Industries
with a comparative advantage should grow while those with a comparative disadvantage should
contract as global trade becomes more liberalized. Some businesses in the same sector struggle
to compete internationally, while others succeed based on their attributes. Melitz (2003)
incorporated the concept of firm heterogeneity along with the suppositions of scale economies,
product differentiation and imperfect competition. Government policies towards promoting
free trade would result to shifting funds and market share from less productive to more
productive firms. As a result, firm’s productivity and performance in general would be boosted
through trade and the inefficient and non-productive firms would eventually exit the market.
With the reallocation of resources from less productive to more productive firms, there will be
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
43
self-selection into export markets by highly productive firms and productivity would increase
for exporting firms.
NNTT as well as NTT are improvements of the traditional trade theories since they incorporate
the concept of trade among homogeneous countries. More so, they relax the assumptions of the
traditional trade theories by incorporating firm heterogeneity, scale economies, product
differentiation and imperfect competition in the analysis for improved plausibility. Therefore,
the current study borrowed the arguments of the NNT and the NNTT to establish the firm-level
determinants of export performance whereby productivity is a key driver for exporting which
is in line with the self-selection hypothesis.
Empirical Review
Okado (2013) used firm-level panel data from the manufacturing sector in Kenya to analyze
the determinants of export propensity and intensity over the period 1992-2003. The focus of
the study was on effects of total factor productivity on export propensity and intensity. These
were estimated controlling for exogenous covariates such as location-specific and
characteristics of firms, notably firm age and sunk investment. The paper used the Heckman
sample selection model in the estimation. The main finding of the study was that export
propensity and intensity in Kenya were positively highly responsive to total factor productivity
and firm size thus validating the self-selection hypothesis. The study made commendable effort
in accounting for sample selection bias in the analysis. The current study borrowed the
methodology adopted by the reviewed study while employing a more recent data set to capture
current issues.
Fonchamnyo (2014) explored the determinants of export intensity and propensity of
manufacturing firms in Cameroon using data obtained from World Bank Investment Climate
Survey for the period. A logit model was employed to analyze the determinants of export
propensity whereas the determinants of export intensity were analyzed using a tobit model. The
explanatory variable of interest for both models were firm size, wage, human capital, firms’
turnover, firm age, experience, power outages, capital intensity, new vintage capital and
insecurity. The results for export propensity indicated that firm size, human capital, vintage
capital, turnover and age positively affected firms’ decision to export while capital intensity
negatively influenced export propensity. The results for the determinants of export intensity
indicated that firm size, human capital, turnover, firm age and experience positively affected
export intensity. This study provided evidence on the determinants of export participation for
manufacturing firms in Cameroon. Nevertheless, total factor productivity was not incorporated
in the analysis. In addition, to cater for sample selection bias, it is appropriate to jointly analyze
the determinants of export propensity and intensity using the Heckman sample selection model
which the current study did.
Reis and Forte (2016) analyzed the role of industry characteristics on export intensity for
Portuguese manufacturing firms. The study utilized panel data obtained from the firms’ balance
sheets for the period 2008-2010. The study estimated both a pooled OLS and a fixed effect
model where by export intensity was the dependent variable and the explanatory variables were
capital intensity, research, labor productivity, export orientation and concentration. A set of
control variables were used which included firm size, age and year dummies. The results
indicated that labor productivity positively affected export intensity while industry
concentration levels and export orientation negatively influenced export intensity. Firm size
also affected export intensity positively. The use of fixed effect model to analyze the
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
44
determinants of export intensity in this case suffered some drawbacks since it does not account
for sample selection bias in this subject. Hence sample selection models are more appropriate
in this regard which the current study utilized to analyze the determinants of export intensity
for Kenya’s manufacturing firms while incorporating total factor productivity instead of labor
productivity.
Vu et.al (2016) analyzed the link between exporting and productivity of Vietnamese
manufacturing firms using firm-level data obtained from the institute of Labor Science and
Social Affairs for the period 2005 and 2007. The self-selection hypothesis was tested using a
dynamic random effects probit model whereby the export dummy was regressed on its first lag
and total factor productivity while controlling for firm age, firm size, capital intensity, trade
relationship, average wage, innovation as well as urban and ownership dummies. Based on the
results, the coefficients of TFP and the lagged export dummy were positive and statistically
significant providing evidence of self-selection hypothesis. Among the control variables, firm
size, trade relationship and ownership dummies positively affected export propensity. The
reviewed study made commendable effort in addressing the endogeneity bias in this subject by
utilizing the dynamic random effects probit model. The current study followed a relatively
similar approach with little divergence by employing the Heckman Sample Selection model to
address sample selection bias in the Kenyan case.
Krammer et. al (2018) analyzed how firm attributes and institutional environments influence
the export performance of emerging economy firms. The study utilized WBES firm-level data
for Brazil, Russia, India and China for the period 2015. Export performance was measured
using export propensity and export intensity. Institutional environment was represented by:
political instability; competition from the informal sector and corruption. Firm attributes were
measured by: Skilled workers; managerial expertise and technological capabilities. A set of
control variables were incorporated in the analysis such as: firm size; firm age; foreign
ownership; public ownership; work force quality as well as country and industry dummies. To
account for selection bias, the analysis was conducted using the Heckman two-stage estimation
procedure. The results indicated that export propensity was positively influenced by: firm age;
firm size; foreign ownership; competition from the informal sector and political instability. On
the other hand, firm size, firm age negatively influenced export intensity while technological
capabilities and skilled workers positively influences export intensity. The study accounted for
the sample selection bias problem by utilizing the Heckman two-stage estimation procedure
which the current study adopted in the analysis of the drivers of export performance by firms
in Kenya’s manufacturing sector.
Chebor (2020) examined the firm-level determinants of growth of exports in Kenya’s
manufacturing sector using three waves of panel data (2007, 2013 and 2018) from the World
bank Enterprise Surveys. The analysis was conducted using the 2SLS technique to cater for
possible endogeneity and heterogeneity. The individual firm characteristics analyzed were age,
size, innovation, human capital and foreign ownership. The key findings showed that firm size,
foreign ownership, skilled human capital and innovation positively affect export intensity.
However, total factor productivity was not analyzed under the factors influencing exports yet
in literature, according to the self-selection hypothesis, business productivity influences
exporting behavior. The current study utilized the same data set and contributed to the existing
literature by incorporating total factor productivity in the analysis and utilizing the Heckman
Sample Selection model that corrects for sample selection bias.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
45
Kiendrebeogo (2020) tested the self-selection hypotheses using unbalanced panel data of
Egyptian manufacturing firms obtained from the World Bank Enterprise Surveys database for
the period 2003-2008. The self-selection hypothesis was tested through comparing productivity
between exporters and non-exporters in the current period, one year prior to commencement of
exporting incorporating control variables using matching techniques. The controls included
employment, wage, firm age, research and financial health not forgetting location and industry
dummies. The results did not support the self-selection hypothesis in the sense that exporters
did not experience total factor productivity improvements prior to entering the foreign markets.
The current study employed a different technique (Heckman sample selection model) to test
the self-selection hypothesis due to some limitations of matching techniques.
Dong and Zhou (2022) analyzed the moderating effect of firm ownership on the effect of
innovation on export performance for Chinese manufacturing firms for the period 2000-2007.
The dependent variable was export intensity while the explanatory variable of interest was
innovation outputs with foreign ownership and state ownership employed as moderators. A
pooled OLS model was estimated whereby export intensity was regressed on innovation, ratio
of state owned capital, ratio of foreign owned capital, interaction terms of innovation and the
two moderators while controlling for total factor productivity, firm size, firm age, financial
leverage, international openness, marketing capability, tangible resources, regional, industry
and time dummies. The results indicated that innovation and foreign ownership positively
affected export intensity while state ownership negatively affected export intensity. The
coefficient of the interaction term between innovation and state ownership was positive while
that of innovation and foreign ownership was negative. The other determinants of export
intensity were firm size, total factor productivity, international openness, firm age, financial
leverage, marketing capability and tangible resources. Nevertheless, analyzing the
determinants of export intensity independently using static panel data models does not account
for sample selection bias hence the need for sample selection models which the current study
utilized.
Camino-Mogro et.al (2023) tested the self-selection hypothesis for Ecuador’s manufacturing
firms using unbalanced panel data from firms’ financial statements and balance sheets for the
period 2007-2018. The variables of interest were: gross revenue, total factor productivity,
capital stock, foreign intermediates, domestic intermediates, total exports, wages, labor
productivity, capital productivity, size, age and export dummy. Region, state and location
dummies were also incorporated in the analysis. To test the self-selection hypothesis, the
lagged values of total factor productivity were regressed on the current export status while
controlling for the aforementioned set of control variables using OLS. From the results it was
evident that exporters outperformed non-exporters in all dimensions: total factor productivity;
gross revenue; employment; capital stock; total intermediates; wages; labor productivity;
capital productivity and age thus supporting the self-selection hypothesis. The study accounted
for selection bias using matching techniques. Nevertheless, due to the limitations of matching
techniques, the current study used the Heckman sample selection model to establish the
determinants of export performance for manufacturing firms in Kenya.
Research Gaps
The reviewed empirical literature reveals that there is limited empirical evidence on the
determinants of export performance for manufacturing firms in Kenya yet Kenya is on a
manufacturing export-led industrialization path. More so, some studies such as Fonchamnyo
(2014), Reis and Forte (2016), Krammer et. al (2018) and Chebor (2020) do not employ total
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
46
factor productivity in the analysis yet in literature, total factor productivity is very key when
testing the self-selection hypothesis. In addition, total factor productivity is a more suitable
indicator of technological progress as opposed to labor productivity. Some of the reviewed
studies also suffer from methodological issues whereby the methodologies applied do not cater
for sample selection bias in this subject which may lead to unreliable results. The current study
therefore explored the determinants of export performance by manufacturing firms in Kenya
while incorporating total factor productivity as well as utilizing the Heckman sample selection
model to account for sample selection bias as contribution to the literature for Kenya.
METHODOLOGY
Research Design
The study employed a quantitative non-experimental research design to achieve the research
objectives.
Theoretical Framework
The study aimed at identifying the firm-level determinants of export propensity and export
intensity by Kenya’s manufacturing firms. According to New Trade Theory (NNT) and ‘New’
New Trade Theory (NNTT) alongside empirical evidence, the choice to export is made in light
of level of profits derived from export markets (Krugman, 1979; Bernard & Jensen, 1999;
Bernard & Wagner, 2001; Melitz, 2003). A firm that seeks to maximize profits bases its
decision to export on the degree of anticipated current and future income from exporting
(Bernard & Jensen, 1999; Bernard & Wagner, 2001). Let
denote the profit maximizing
output level by the firm. Under the one period case with zero entry (sunk) costs, the firm’s
profits are given as follows:
Where the is the price of exports; is the variable production cost of
; denotes
exogenous factors affecting firm’s profits; represents firm-specific characteristics that
might influence export decision such as productivity, firm ownership, firm size, labor
composition and product mix. If predicted profits are positive, a firm will export as shown
below:
Where the firm’s export status in period is . Extending equation (3.1) to multiple periods
yields:
Where denotes the discount rate. The solution to the multiple period case is identical to the
one period case as shown in equation (3.2). With the introduction of sunk costs , the firm’s
profits under the single period case are:
Where is the prior period’s firm’s export status. If the firm was an exporter during the
prior period, ( it will not incur sunk costs in the current period. Thus, in period ,
the firm will optimize from exporting if Due to sunk costs, the decision to export today
by a firm will affect the probability of exporting in the succeeding periods. The firm therefore
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
47
chooses a chain of output levels,
that will optimize the present and discounted future
profits:
The value function is expressed as follows:
Where is the firm's expected value function from exporting in the succeeding
periodIn period a firm will find it optimal to export if the current and expected payoffs from
exporting outweigh the costs incurred as shown in equation (3.7).
Let
Thus:
Equation (3.9) implies that a firm will decide to export if it expects positive profits. Based on
equation (3.9), and following empirical evidence the decision to export can be presented as
follows (Bernard & Jensen, 1999; Bernard & Wagner, 2001):
Where is a vector of business traits that may influence the decision to export such as
productivity, firm ownership, firm size, human capital, capital per employee, research and
development, firm age and management quality and is the residual. Equation (3.10) can be
used to estimate the determinants of export propensity using binary choice models. It can also
be modelled to separately estimate the determinants of the export intensity based on the
specified firm attributes. Hence the export performance function was generally expressed as:
Where is firms’ export performance in period and is a vector of firm traits
that may influence export behavior such as total factor productivity, firm age, firm size, foreign
ownership, capital intensity, research and management quality as used in vast literature
(Bernard & Jensen, 1999; Bernard & Wagner, 2001; Bernard, Jensen, Redding, & Schott, 2007;
Bigsten & Gebreeyesus, 2009; Camino-Mogro, Ordeñana-Rodríguez, & Vera-Gilces, 2023).
Empirical Model Specification
To explore the determinants of export performance, equation (3.11) can be modelled for a
binary dependent variable (export propensity) or a continuous dependent variable (export
intensity) and estimating the two models separately. However, since the decision to export and
the amount exported by a firm are dependent due to self-selection in to export markets, they
cannot be modelled separately (Heckman, 1979; Okado, 2013). More so, export intensity was
measured as the share of firms’ exports in total sales hence it takes a value between zero and
one. Share variables, such as the share of exports in total sales, are common fractional response
variables (Wagner, 2001). Employing the Ordinary Least Squares (OLS) technique to analyze
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
48
a fractional response variable leads to inconsistent results since the predictions from the model
may not lie within the (0,1) interval as well as biased marginal effects (Papke & Wooldridge,
1996; Schwiebert, 2018). This limitation can be overcome using a tobit model or a fractional
probit or logit model introduced by Papke and Wooldridge (1996). Papke and Wooldridge
(1996) proposed specifying only the conditional mean of the fractional response variable rather
than the entire conditional distribution. Let y represent the fractional response variable (export
intensity) with x representing the collection of explanatory variables that have a conformable
parameter vector β. The conditional mean is thus expressed as:
Where is a bounded function either a logistic or normal cumulative distribution function.
Papke and Wooldridge (2008) put forth a panel data specification of equation (3.12) as follows:
Where represents the individual specific effect, is a logistic or normal cumulative
distribution and the rest of the variables are as defined in equation (3.11).
More so, due to the existence of sunk costs, firms self-select themselves into exporting based
on their attributes and this leads to self-selection bias which is not accounted for by the tobit,
fractional probit or logit models (Faria, Rebelo, & Gouveia, 2020). More so, since export
intensity can only be observed for exporting firms, sample selection bias arises. This implies
that, when analyzing the export intensity model, there is need to cater for the fractional nature
of the variable as well as selection bias. In his landmark study, Heckman (1979) noted that
sample selectivity happens when the choice of participants into the sample under study is non-
random. Thus, eliminating non-exporters and assessing export intensity independently using
just exporters may result in selectivity bias.
Consider the following model proposed by Heckman (1979) to rectify this sample selection
bias:
Where
denotes a latent dependent variable, is an observed binary variable (selection
equation) that in this case indicates the export status of the firm i.e. exporters ( and
non-exporters ( and is the observed dependent variable (export intensity in this
study) and its observed when . The observed explanatory variables are presented by
vectors and with corresponding parameters and . The error terms are denoted by
and which are assumed to follow a conditional bivariate normal distribution.
Based on equations (3.14) and (3.15), is observed when and this happens when:
The probability that is observed is:
Equation (3.18) holds by symmetry of the standard normal distribution. Where Pr denotes
probability and is the cumulative density function of the standard normal distribution.
Given the conditional mean of the observed dependent variable as:
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
49
Substituting equation (3.17) yields:
Substituting equation (3.14) produces:
Where is the correlation between the errors of equations (3.14) and (3.15), is the variance
of the error term in the main equation (3.14), is the probability density function of the
standard normal distribution and
denotes the inverse Mills ratio (IMR).
The Heckman sample selection model jointly estimates the export participation and export
intensity models (3.15) and (3.14), respectively by estimating and incorporating the inverse
Mills ratio obtained from equation (3.22) into the regression equation to eliminate bias,
resulting in unbiased findings. The Heckman model relies on distributional assumptions of the
residuals or imposition of appropriate exclusion restrictions. Satisfaction of either of the two
conditions and implementation of the two-step procedure for the Heckman model leads to
reliable estimates within the limited interval (0,1) even in the case of fractional response
outcome variables (Schwiebert, 2018). The exclusion restriction involves having an additional
explanatory variable (instrument) on the selection equation (3.15) which is excluded from the
main equation (3.14).
Therefore, the study adopted the Heckman (1979) two-step sample selection model in
establishing the determinants of export propensity and export intensity by estimating equations
(3.14), (3.15) and (3.16) while incorporating explanatory variables such as total factor
productivity, firm age, firm size, foreign ownership, capital intensity, human capital, research
and management quality as defined on equation (3.11). In addition, dummy variables for year,
industry and region were incorporated in the models to obtain the following empirical models:
Where is a vector of year, industry and region dummies, is the inverse mills ratio
and the rest of the variables and parameters are as defined in equations (3.14), (3.15) and (3.16).
The Heckman two-step sample selection procedure adopted by the study involved estimating
equation (3.24) first using a probit model to establish the determinants of export propensity
after which the inverse mills ratio (IMR) was computed. The second stage involved estimating
equation (3.23) while incorporating the IMR as an explanatory variable to account for the
selection bias when establishing the determinants of export intensity.
Data Type, Source and Analysis
The research utilized panel dataset obtained from the World Bank Enterprise Surveys (WBES)
for manufacturing firms in Kenya covering the periods 2007, 2013 and 2018. The study
performed descriptive analysis so as to understand the characteristics of the study data.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
50
Regression analysis using STATA was conducted to establish the determinants of export
performance by manufacturing firms in Kenya.
Diagnostic Tests
Normality Test
Since the study employed the probit model, the relevant variables were tested for normality by
checking the distribution of the variables as well as their descriptive statistics such as the
standard deviation, skewness and kurtosis. Logarithmic transformation of the variables helped
in achieving the normality assumption.
Multicollinearity Test
The multicollinearity test was conducted using the VIF and 1/VIF statistics. VIF values
exceeding 10 and 1/VIF values below 0.1 indicate high levels of multicollinearity that needs to
be addressed (Kutner, Nachtsheim, & Neter, 2004). Mostly this is addressed by dropping one
of each of the highly collinear variables until the problem is solved.
Regression Specification Error Test
To ensure that all the model was correctly specified, the study employed Ramsey regression
specification error test (RESET) under the null hypothesis of a correctly specified model
(Ramsey, 1969). The model is correctly specified if the probability value of the F-statistic is
greater than 0.05 (Ramsey, 1969).
Heteroscedasticity Test
The modified Wald test for group wise heteroscedasticity was employed to check the variance
of the residuals under the null hypothesis of homoscedasticity (Greene, 2012). A probability
value of the Chi-square statistic greater than 0.05 indicates constant variance (Greene, 2012).
In the presence of heteroscedasticity, the robust option can be applied to obtained robust
standard errors.
FINDINGS
Descriptive Statistics
This section provides the descriptive statistics for the study variables. The analysis is
categorized into two sections. The first section captures the summary statistics for the
continuous variables while the tabulation of the discrete variable is presented in the second
section.
Summary Statistics for the Continuous Variables
The summary statistics for the continuous variables are presented on Table 1.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
51
Table 1: Summary Statistics for Continuous Variables
Variables
N
Mean
S.D.
Min
Max
Skewness
Kurtosis
Export Intensity
482
0.19
0.30
0
1
1.72
4.74
Total Factor productivity
482
8.40
3.30
0
17.47
-.74
4.96
Material
482
344.84
2146.01
0
36000
12.05
176.11
Energy Cost
482
22.97
171.10
0
3500
17.95
357.68
Firm Size
482
204.17
572.47
0
8000
8.25
93.06
Firm Age
482
32.37
18.7
0
103
.82
3.64
Human Capital
482
23.45
35.67
0
100
1.28
3.03
Labor Productivity
482
5.05
28.51
0
600
19.09
395.67
Capital Intensity
482
12.09
229.00
0
5000
21.53
468.96
Material, Energy Cost, Labor Productivity, Capital Intensity are in Million KShs.
N = Number of Observations; S.D. = Standard deviation; Min = Minimum value and Max =
Maximum value.
Source: Author’s Computations from WBES Data (2007, 2013, 2018).
Export intensity, measured as the share of a firm’s exports in total sales, had a mean value of
0.19 and a standard deviation of 0.3 implying high dispersion from the mean. The mean value
of 0.19 implies that the sampled firms in the manufacturing sector exported an average of 19
per cent of their total sales within the study period. This was an indication of low export share
for the sampled firms in the sector. Export intensity had a maximum value of 1 (for firms that
exported all their sales) and a minimum value of 0 for non-exporters. With reference to a
skewness of zero and kurtosis of 3 for a standard normal distribution, based on the skewness
and kurtosis of 1.72 and 4.74, respectively, export intensity was positively skewed and mildly
leptokurtic.
The average total factor productivity (TFP) for the sampled firms in the sector within the study
period was 8.40 with a standard deviation of 3.30 implying low variation from the mean. Total
factor productivity was slightly negatively skewed with a value of -0.74 compared to a zero
skewness value of a normal distribution. Based on the kurtosis of 4.96, with reference to a
value of 3 for a standard normal distribution, TFP was mildly leptokurtic implying that it had
a slightly peaked curve. Material and Energy cost had mean values of 344.84 and 22.97 Million
Kenya Shillings, respectively. Based on their standard deviations of 2146.01 and 171.10,
respectively, they were highly dispersed from their mean values. Their skewness and kurtosis
values indicated that they were all leptokurtic and positively skewed.
Firm size, represented by the total number of workers employed by a firm, had a mean value
of 204 implying that, on average, the sampled firms employed 204 workers within the study
period. Firm size was highly volatile as indicated by a standard deviation of 572.47. For the
sampled firms, the largest firm employed 8,000 workers within the study period. With a
skewness of 8.25 and a kurtosis of 93.06, firm employment was positively skewed and
leptokurtic. The average age of the sampled firms in the sector was 32 years with a standard
deviation of 18.70 implying less variability. The oldest firm(s) was 103 years old. The
skewness of 0.82 and kurtosis of 3.64 were very close to a normal distribution.
Human capital representing the percentage of full time workers who received formal training
had a mean value of 23.45 indicating that, for the sampled firms, only 23.45 per cent of the
workers received formal training within the study period, on average. With a standard deviation
of 35.61 the variability from the mean was high implying that the level of formal training of
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
52
workers differed greatly across firms. Human capital was also moderately skewed (1.28) with
a normally peaked curved based on a kurtosis of 3.03. Labor productivity, was on average 5.05
Million Kenya Shillings with a high dispersion form the mean. Based on the skewness (19.09)
and kurtosis (395.67) labor productivity curve was positively skewed and leptokurtic. The
average capital intensity for the sampled firms in Kenya’s manufacturing sector was 12.09
Million Kenya Shillings for the study period with a very high dispersion from the mean. It also
exhibited positive skewness and had a highly peaked curve.
Tabulation of the Discrete Variable
Export propensity for the sampled firms in Kenya’s manufacturing sector, was presented as a
dummy variable = 1 for exporters and 0 for non-exporters. Table 2 presents the statistics for
export propensity for the sampled firms within the study period.
Table 2: Export Propensity Statistics
Export Propensity
Frequency
Percent
Cumulative
Non-Exporters
247
51.24
51.24
Exporters
235
48.76
100.00
Total
482
100.00
Source: Author’s Computations from WBES Data (2007, 2013, 2018)
Table 2 indicates that exporters accounted for 48.76 per cent of the total sampled firms in
Kenya’s manufacturing sector, on average.
Diagnostic Test Results
Normality Test Results
Apart from total factor productivity, all the variables presented on Table 1 had not been
transformed into natural logarithmic form. In order to achieve a reasonably normal distribution,
all the variables were transformed into natural logarithmic form.
Multicollinearity Test Results
The results indicated that the model did not suffer from multicollinearity since the VIF values
were less than 10 and 1/VIF values were greater than 0.1 for each variable as presented on
Table 3.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
53
Table 3: Multicollinearity Test Results on the Determinants of Export Intensity
Variables
VIF
1/VIF
Total Factor Productivity
5.959
.168
Firm Employment
1.659
.603
Firm Age
1.233
.811
Human Capital
1.245
.803
Labor Productivity
7.978
.125
Material
2.165
.462
Energy Cost
2.78
.36
Foreign Ownership
1.113
.898
Mean VIF
2.144
VIF: Variance Inflation Factor
Source: Author’s Computations from Study Data
Ramsey Regression Specification Error Test (RESET) Results
The results indicated that the model was correctly specified since the probability value of the
F-statistic was greater than 0.05 as shown on Table 4.
Table 4: Results of the Ramsey Regression Specification Error Test (RESET)
Model
F-statistic
P-value
Determinants of Export Intensity
1.35
0.2601
Source: Author’s Computations from Study Data
Heteroscedasticity Test Results
The results indicated the presence of heteroscedasticity given that the probability value of the
Chi-square statistic was less than 0.05 as presented on Table 5. This was corrected by
computing robust standard errors.
Table 5: Results for Modified Wald Test for Group-Wise Heteroscedasticity
Model
Chi-square statistic
P-value
Determinants of Export Intensity
1.2e+34
0.000
Source: Author’s Computations from Study Data
Empirical Results
Results on the Firm-Level Determinants of Export Propensity by Manufacturing Firms
in Kenya
The results from estimating the probit model (3.24) are presented on Table 6.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
54
Table 6: Regression Results on the Firm-Level Determinants of Export Propensity in
Kenya’s Manufacturing Sector
Dependent Variable: Export Propensity
Variables
Estimated Coefficient
P-Value
Marginal Effect
P-Value
Total Factor Productivity (TFP)
0.1790**
0.018
0.0412**
0.014
Firm Size
0.3735***
0.000
0.0861***
0.000
Capital Intensity
-0.0081
0.631
-0.0019
0.630
Firm Age
0.1273
0.300
0.0293
0.296
Human Capital
0.1361***
0.005
0.0314***
0.003
Labor Productivity
-0.1249**
0.025
-0.0288**
0.019
Material
0.0322*
0.066
0.0074*
0.062
Energy Cost
0.0784**
0.012
0.0181***
0.008
Foreign Ownership Dummy (FO)
Base: Domestic
Foreign
0.7902**
0.013
0.1822***
0.008
Year Dummy
Base: 2007
Year=2013
Year=2018
0.4473*
0.0714
0.051
0.780
0.1033**
0.0163
0.046
0.779
Research Dummy
Base: Non-Researchers
Research
0.2843
0.133
0.066
0.130
Industry Dummy
Base: Other Manufacturing
Food
Textiles and Garments
Chemical, Pharmaceutical, and Plastic
-0.1250
0.2844
0.5932*
0.597
0.317
0.052
-0.0290
0.0665
0.1378**
0.596
0.311
0.047
Constant
-3.994***
0.000
-
No. of Observations
471
-
Wald: Chi2
64.47***
-
P-Values in parentheses: *** p<0.01, ** p<0.05, * p<0.1
Source: Authors Computations
Based on the results presented on Table 6, the coefficient for total factor productivity was
positive and statistically significant at 5 percent. Therefore, an increase in firm-level total factor
productivity increased the probability of a firm becoming an exporter, ceteris paribus. This is
because highly productive firms are able to overcome the sunk costs involved in entering
foreign markets. The study results support the self-selection hypothesis are in line with vast
empirical evidence such as (Okado, 2013; Vu, Holmes, Tran, & Lim, 2016; Camino-Mogro,
Ordeñana-Rodríguez, & Vera-Gilces, 2023).
The coefficient of firm size was positive and statistically significant at 1 percent. This implies
that an increase in the firm size increased the probability of a firm becoming an exporter, all
else being equal. This is because larger firms are able to enjoy economies of scale and have
more resources to access better technologies for accessing foreign markets compared to small
firms. The study findings support existing empirical evidence including (Fonchamnyo, 2014;
Vu, Holmes, Tran, & Lim, 2016; Krammer, Strange, & Lashitew, 2018; Chebor, 2020;
Camino-Mogro, Ordeñana-Rodríguez, & Vera-Gilces, 2023).
Human capital, measured as the share of trained workers in a firm, had a positive effect on
export propensity as indicated by a positive and statistically significant coefficient. The results
implied that an increase in the share of trained workers in a firm led to increased probability of
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
55
the firm engaging in exporting, all else being equal. Training of workers improves their skills
and capabilities which gives the firm a competitive edge since the workers are able to work
and adopt to the modern technologies needed to produce high quality products and enter foreign
markets. The results were similar to (Fonchamnyo, 2014).
The coefficient for labor productivity was negative and statistically significant at 5 percent.
This meant that an increase in labor productivity reduced the probability of a firm becoming
an exporter, all other factors held constant. The literature on this is mixed (Guner, Lee, &
Lucius, 2010; Pham, 2015; Reis & Forte, 2016; Jakšić, Erjavec, & Cota, 2019). Economic
theory implies that workers are compensated based on their marginal productivity, indicating
that salaries are positively connected to labor productivity. An increase in labor marginal
productivity raises salary demands, and because this costs the firm money, the firm may end
up keeping only a few productive employees thus negatively affecting export propensity.
The cost of material was positively related to export propensity given that the coefficient was
positive and statistically significant. This implied that, an increase in the expenditure of
materials increased the probability of a firm becoming an exporter, ceteris paribus. This could
be attributed to high quality materials required for production of quality exports. Hence, for a
firm to become a successful exporter, they have to incur high costs on materials. The results
are supported by existing literature (Bas & Strauss-Kahn, 2014; Feng & Swenson, 2016).
Energy cost, measured as the cost of electricity, had a positive and statistically significant
coefficient implying that energy cost positively affected export propensity. The results meant
that an increase in cost of electricity by the firm, increased the probability of the firm becoming
an exporter. This can be explained by the fact that in Kenya, electricity does not have close
substitutes and those available may be costly to install. As such, most of the firms may not
have a choice rather than bear with the high costs of electricity and find a way to transfer the
burden to the consumers. For a firm to produce high quality goods for exports, energy cost is
inevitable just as with the case of materials, hence the positive relationship.
Foreign ownership, expressed as a dummy variable had a positive and statistically significant
coefficient. This meant that foreign firms had a higher probability of exporting compared to
domestic firms. This can be attributed to foreign enterprises' advantages in terms of direct
exposure to information and marketing networks regarding international markets, managerial
competence, access to more advanced technology, and financial resources in general. These
results are consistent with Okado (2013) for Kenya; Chebor (2020) for Kenya and Dong and
Zhou (2022) for China. The export propensity of firms was higher in 2013 compared to 2018
as indicated by the positive and statistically significant year dummy. In addition, firms in the
Chemical, Pharmaceutical and Plastic had a higher probability of becoming exporters
compared to those from other manufacturing based on the industry dummy results.
Results on the Firm-Level Determinants of Export Intensity by Manufacturing Firms in
Kenya
The results of the firm-level determinants of export intensity based on the Heckman sample
selection model presented on equations (3.23), (3.24) and (3.25) are presented on Table 7.
Results from the Tobit and fractional probit models are also presented for comparison purposes.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
56
Table 7: Regression Results on Firm-Level Determinants of Export Intensity in Kenya’s
Manufacturing Sector
Variables
Dependent Variable:
Export Intensity
Model
Tobit
P-
Value
Fractional
Probit
P-
Value
Heckman Sample
Selection Model
P-
Value
Total Factor
Productivity (TFP)
0.0488***
0.002
0.0904**
0.029
0.0351***
0.005
Foreign Ownership
Dummy (FO)
Base: Domestic
Foreign
0.7811***
0.000
1.3153**
0.027
0.4281**
0.013
FO*TFP
-0.0702***
0.002
-0.1180*
0.059
-0.0342**
0.048
Firm Size
0.0849***
0.000
0.1472***
0.001
0.0632***
0.002
Firm Age
0.0498
0.114
0.0804
0.263
0.0361**
0.040
Human Capital
0.0347***
0.003
0.0778***
0.004
0.0298***
0.000
Labor Productivity
-0.0252**
0.024
-0.0455
0.114
-0.0197**
0.022
Material
0.0027
0.515
0.0053
0.607
0.0025
0.288
Energy Cost
0.0108
0.140
0.0187
0.268
0.0096*
0.060
Year Dummy
Base: 2007
Year=2013
Year=2018
0.1145**
-0.0250
0.049
0.703
0.3048**
-0.0313
0.017
0.835
0.0996***
-0.0004
0.000
0.990
Research Dummy
Base: Non-Researchers
Research
0.0594
0.224
0.088
0.474
0.0464*
0.091
Industry Dummy
Base: Other
Manufacturing
Food
Textiles and Garments
Chemical,
Pharmaceutical, and
Plastic
-0.0315
0.0334
0.1001
0.605
0.650
0.180
-0.0056
0.0531
0.1191
0.964
0.704
0.432
-0.0103
0.0491
0.0756*
0.726
0.204
0.070
Inverse Mills Ratio
-
-
0.1445**
0.025
Constant
-1.0058***
0.000
-2.9362
0.000
-0.7188***
0.002
No. of Observations
Left-
Censored=247
Uncensored=235
482
Left-Censored=247
Uncensored=235
Wald: Chi2
148.83***
235.91***
876.54***
Pseudo R2
-
0.1863
-
P-Values in parentheses: *** p<0.01, ** p<0.05, * p<0.1
Source: Authors Computations from Study Data
The dependent variable was not transformed into logarithmic form while the independent
continuous variables were transformed in natural logarithmic form yielding a level-log model.
As such the estimated coefficients were interpreted based on the semi-elasticity model. The
results presented on Table 7 were consistent across all the three models in terms of the
coefficient signs and to some extend the statistical significance of the coefficients. However,
the coefficients had different magnitudes across the models. Above all, the coefficient of the
inverse mills ratio in the Heckman model was positive and statistically significant at 5 per cent
level of significance indicating presence of sample selection bias thus validating the Heckman
sample selection model. Therefore, the discussion of the study findings was based on the results
obtained from the two-step Heckman sample selection model.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
57
The coefficient of total factor productivity (TFP) was positive and statistically significant at 1
per cent in the Heckman sample selection model. The coefficient had a value of 0.0351 which
implied that a percentage increase in firms’ TFP resulted to 0.000351 units increase in firms’
export intensity on average, holding all other factors constant. This meant that firms with higher
levels of TFP exported a larger share of their total sales hence the need to improve firm-level
TFP. The positive effect of TFP on firms’ export intensity can be explained by the concept of
sunk costs. Since there exist huge entry costs (sunk costs) in to export markets, only more
productive firm can overcome these cost and enter international markets (Roberts & Tybout,
1997). Once they enter these markets, if they maintain or improve their productivity levels, the
highly productive firms have the capacity to produce and export more compared to the less
productive firms. These results are consistent with existing vast empirical evidence on the
effect of TFP on export behavior of manufacturing firms such as (Bigsten & Gebreeyesus,
2009; Haidar, 2012; Okado, 2013; Dong, Kokko, & Zhou, 2022; Camino-Mogro, Ordeñana-
Rodríguez, & Vera-Gilces, 2023).
The coefficient of foreign ownership was positive and statistically significant. For the Heckman
model, the coefficient had a value of 0.4281 which implied that the share of exports in total
sales was on average 0.4281 more for foreign owned firms compared to domestic firms all else
being equal. This can be associated to superiority of foreign firms in terms of direct access to
information and marketing networks regarding foreign markets, managerial expertise, access
to superior technology and financial resources in general than enhance their export
performance (Krammer, Strange, & Lashitew, 2018; Dong, Kokko, & Zhou, 2022). These
results are consistent with Okado (2013) for Kenya; Chebor (2020) for Kenya and Dong and
Zhou (2022) for China.
In addition to analyzing the independent effect of foreign ownership and TFP on export
intensity, the study incorporated an interaction term between TFP and foreign ownership to
capture the moderating effect of foreign ownership on the effect of TFP on export intensity.
The results indicated that the coefficient was negative and statistically significant at 5 per cent
with a value of -0.0342. This implied that although highly productive and foreign firms
independently had a higher share of exports in total sales, the effect of TFP on export intensity
was lower for foreign firms compared to domestic firms. As such domestic firms have more
room for improvement in terms of enhancing their TFP and export intensity compared to
foreign firms. These results are in line with Dong and Zhou (2022) for Chinese manufacturing
firms.
In line with existing theoretical and empirical evidence, the coefficient for firm size was
positive and statistically significant at 1 per cent. With a value of 0.0632 for the Heckman
model, a percentage increase in the number of workers by a firm increased export intensity by
0.000632 units on average, ceteris paribus. This can be associated to the economies of scale
advantages that large firms enjoy and are thus able to produce and export a larger share of their
sales. More so, large firms are highly capital intensive, can afford advanced technology,
possess intangible assets such as patents and goodwill thus enjoy competitive advantage
compared to small firms in international markets. These results support existing empirical
evidence on the same such as (Fonchamnyo, 2014; Reis & Forte, 2016; Chebor, 2020; Dong,
Kokko, & Zhou, 2022; Camino-Mogro, Ordeñana-Rodríguez, & Vera-Gilces, 2023).
The coefficient for firm age was positive and statistically significant at 5 per cent for the
Heckman model with a value of 0.0361. This implied that, holding other factors constant, a
percentage increase in firm age led to an increase in export intensity by 0.000361 units on
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
58
average. This could be associated with the increased experience in the international markets
which is directly proportional to the firms’ age. Older firms are more experienced hence they
have more knowledge and connections regarding the markets thus they may enjoy higher
international market shares compared to inexperienced younger firms. As a result, they produce
more for exports. The positive relationship between firm age and export intensity corroborates
with existing empirical literature (Bernard & Jensen, 1999; Bernard, Jensen, Redding, &
Schott, 2007; Bigsten & Gebreeyesus, 2009; Kiendrebeogo, 2020).
The coefficient for human capital was positive and statistically significant at 1 percent. A value
of 0.0298 implied that a percentage increase in the share of trained workers in a firm increased
export intensity by 0.000298 units, all else being equal. The positive relationship between
human capital and export intensity implies that firms possessing exceptional human capital can
gain some competitive advantages, which are vital in boosting export performance. The results
are in line with (Fonchamnyo, 2014; Mulliqi, Adnett, & Hisarciklilar, 2019; López Rodríguez
& Serrano Orellana, 2020; Mubarik, Devadason, & Govindaraju, 2020).
The coefficient for labor productivity was negative and statistically significant at 5 percent
significance level with a value of -0.0197. This implied that a percentage increase in labor
productivity resulted to 0.000197 units decrease in export intensity. This implied that, in this
context, labor productivity was inversely related to the firm’s export intensity. The existing
literature in this regard is mixed in the sense that the relationship could be either positive or
negative and in some cases insignificant (Guner, Lee, & Lucius, 2010; Pham, 2015; Reis &
Forte, 2016; Jakšić, Erjavec, & Cota, 2019). The negative effect of labor productivity on export
intensity may arise indirectly through wages and employment levels. Economic theory
suggests that workers get paid according to their marginal productivity, implying that wages
are positively related to labor productivity. An increase in the marginal productivity of labor
increases wage demands and since this is a cost to the firm, the firm may end up retaining few
productive workers. On the other hand, the study established a positive effect of the number of
workers and export intensity which implies that, a reduction in the number of productive
workers may in turn reduce export intensity.
Just as in the case of export propensity, energy cost had a positive effect on export intensity.
As expected, the coefficient for research was positive pointing towards high export intensity
for research oriented firms. However, it was weakly significant at 10 percent significance level.
The coefficient had a magnitude of 0.0464 implying that firms that engaged in research and
development had a 0.0464 higher share of exports in total sales on average compared to those
who did not engage in research, ceteris paribus. Engaging in research activities promotes
innovation and inventions which lead to introduction of new or high quality products in the
market which are more likely to enter international markets. As a result, research oriented firms
become more efficient and competitive thus export a higher share of their sales. These findings
support existing empirical evidence on the effect of research and development on export
intensity (Rialp-Criado & Komochkova, 2017; Benfratello, Bottasso, & Piccardo, 2022).
The 2013-year dummy coefficient was positive and statistically significant at 1 percent level
of significance. With a value of 0.0996 for the Heckman model, in 2013, the firms’ share of
exports in total sales was 0.0996 more compared to 2007, on average, holding other factors
constant. This could be attributed to the economy recovery strategies that were put in place to
spur the economy after the 2007-2008 post-election violence in Kenya.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
59
SUMMARY, CONCLUSIONS AND POLICY IMPLICATIONS
Summary
The study sought to establish the firm-level determinants of export propensity and export
intensity by manufacturing firm in Kenya. The World Bank Enterprise Survey panel data for
the period 2007, 2013 and 2018 was utilized. By employing the Heckman two-step sample
selection model, the study findings put forth, total factor productivity, firm size, human capital,
material, energy cost and foreign ownership as positive determinants of export propensity.
Export propensity was negatively influenced by labor productivity. On the other hand, export
intensity was positively affected by total factor productivity, foreign ownership, firm size, firm
age, human capital, energy cost and research. Labor productivity negatively influenced export
intensity.
Conclusions
The self-selection hypothesis argues that total factor productivity is one of the main
determinants of export performance by firms. Based on the study findings, total factor
productivity is a key determinant of export performance by manufacturing firms in Kenya.
Therefore, the study concludes that the self-selection hypothesis is validated by the study
findings. More so, other determinants of export performance include foreign ownership, firm
size, firm age, human capital, energy cost, material, research and labor productivity. Hence, to
achieve the set targets regarding export promotion in Kenya, the study concludes that it’s
imperative to keenly focus on these variables at the firm level.
Policy Implications
Based on the study findings, several policy implications can be drawn. First the study identified
total factor productivity as a key driver of export performance for Kenya’s manufacturing
firms. This implies that the government should support manufacturing firms in terms of TFP
enhancement. This can be achieved through government support on the invention and adoption
of new technologies and investment in human capital. Since new technology is very costly,
firms need government support to realize this. The Ministry of Cooperatives and Micro, Small
and Medium Enterprises (MSME), the Ministry of Trade, Investments and Industry and the
National Treasury and Economic Planning should work together and pool resources aimed at
supporting manufacturing firms to invent and adopt new technologies for productivity
enhancements. Firms also need skilled workers to work with these new technologies, hence the
need to invest in human capital. This can be achieved through training of workers.
Second, the study findings established a positive effect of human capital on firm’s export
performance in Kenya’s manufacturing sector. This meant that training workers was crucial in
terms of the firm’s performance. Therefore, the government, through the Ministry of
Cooperatives and Micro, Small and Medium Enterprises (MSME) Ministry of Education and
Ministry of Trade, Investments and Industry needs to ensure that the curriculum involves
training workers in the manufacturing sector as well as upgrading their skills to boost the
performance of the firms. Third, based on the study findings, foreign-owned firms exhibit
better export performance compared to domestic firms. The government should therefore
provide a conducive business environment to encourage foreign investors in order to reap these
benefits and provide positive spillovers to the domestic firms. More so, domestic firms need
more support from the government to overcome the sunk costs involved in successfully
penetrating and surviving in international markets.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
60
Fourth, based on the positive effect of firm size on export intensity established by the study,
the government, through the Ministry of Cooperatives and Micro, Small and Medium
Enterprises (MSME) and Ministry of Trade, Investments and Industry should support MSMEs
to grow and graduate into large enterprises so as to reap the benefits and enhance their export
performance. Fifth, firm age had a positive effect on export intensity. This implied that, as
firms stay longer in operation, they gain more experience and market access and are thus able
to export a larger share of their exports. These findings imply that it is important for the
government to consider supporting startups and young entrepreneurs to enable them realize
their potential over time and succeed. The Ministry of National Treasury and economic
Planning, the Ministry of Cooperatives and Micro, Small and Medium Enterprises (MSME) in
collaboration with the Ministry of Trade, Investments and Industry should support these young
enterprises through various channels like access to finance, adoption of new technology,
market access, product quality as well as offering them subsidies and tax exemptions where
possible.
Sixth, based on the positive effect of research on export intensity, the government should
consider supporting and encouraging firms to participate in research. The government should
ensure that the intellectual property rights of inventors and innovators are protected so as to
motivate research. Finally, the government should invest more on renewable energy and find
other means of cutting down on energy costs so as to provide a business friendly environment
for the firms and consumers in general. Firms should also adopt energy efficient technologies
so as to minimize their energy costs and remain competitive.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
61
REFERENCES
Bas, M., & Strauss-Kahn, V. (2014). Does importing more inputs raise exports? Firm-level
evidence from France. Review of World Economics, 150, 241-275.
doi:https://doi.org/10.1007/s10290-013-0175-0
Benfratello, L., Bottasso, A., & Piccardo, C. (2022). 'R&D and export performance:
exploring heterogeneity along the export intensity distribution. Journal of Industrial
and Business Economics, 49(2), 189-232. doi:https://doi.org/10.1007/s40812-022-
00209-1
Bernard , A. B., & Jensen, J. B. (1999). Exceptional exporter performance: cause, effect, or
both? Journal of international economics, 47(1), 1-25.
doi:https://doi.org/10.1016/s0022-1996(98)00027-0
Bernard, A. B., & Wagner, J. (2001). Export entry and exit by German firms. Review of
World Economics, 137(1), 105-123. doi:https://doi.org/10.1007/bf02707602
Bernard, A. B., Jensen, J. B., & Lawrence, R. Z. (1995). Exporters, jobs and wages in US
manufacturing. Brookings papers on economic activity, 67-119.
doi:https://doi.org/10.2307/2534772
Bernard, A. B., Jensen, J. B., Redding, S. J., & Schott, P. K. (2007). Firms in international
trade. Journal of Economic perspectives, 21(3), 105-130.
doi:https://doi.org/10.1257/jep.21.3.105
Bigsten, A., & Gebreeyesus, M. (2009). Firm productivity and exports: Evidence from
Ethiopian manufacturing. The Journal of Development Studies, 45(10), 1594-1614.
doi:https://doi.org/10.1080/00220380902953058
Bresnaham, L., Coxhead, I., Foltz, J., & Mogues, T. (2016). Does freer trade really lead to
productivity growth? Evidence from Africa. World Development, 1-29.
doi:https://doi.org/10.1016/j.worlddev.2016.05.007
Cameron , A. C., & Trivedi, P. K. (2005). Microeconometrics: Methods and Applications.
New York: Cambridge university press.
doi:https://doi.org/10.1017/cbo9780511811241
Camino-Mogro, S., Ordeñana-Rodríguez, X., & Vera-Gilces, P. (2023). Learning-by-
exporting vs. self-selection in Ecuadorian manufacturing firms: Evidence from
different industry classifications. The Journal of International Trade & Economic
Development, 32. doi:https://doi.org/10.1080/09638199.2022.2094451
Charles , S. W., & Richard, L. S. (2020). Applied International Economics. New York:
Routledge.
Chebor, T. K. (2020). Export Intensity and Manufacturing Firm Characteristics in Kenya.
Doctoral Dissertation, University of Nairobi.
Dong, G., Kokko, A., & Zhou, H. (2022). Innovation and Export Performance of Emerging
Market Enterprises. International Business Review.
doi:https://doi.org/10.1016/j.ibusrev.2022.102025
Esaku, S. (2020). Trade liberalization and productivity growth: A firm-level analysis from
Kenya. Review of Economic Analysis. doi:https://doi.org/10.15353/rea.v12i4.1791
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
62
Faria, S., Rebelo, J., & Gouveia, S. (2020). Firms’ export performance: a fractional
econometric approach. Journal of Business Economics and Management, 21(2), 521-
542. doi:https://doi.org/10.3846/jbem.2020.11934
Feng, L. L., & Swenson, D. L. (2016). The connection between imported intermediate inputs
and exports: Evidence from Chinese firms. Journal of International Economics, 101,
86-101. doi:https://doi.org/10.1016/j.jinteco.2016.03.004
Fonchamnyo, D. C. (2014). (2014). Determinants of export propensity and intensity of
manufacturing firms in Cameroon: an empirical assessment. Applied Economics and
Finance, 1(2), 30-36. doi:https://doi.org/10.11114/aef.v1i2.413
Greene, W. H. (2012). Econometric Analysis (Seventh international ed. ed.). Boston: Pearson.
Guner, B., Lee, J., & Lucius, H. W. (2010). The impact of industry characteristics on export
performance: a three country study. International Journal of Business and Economics
Perspectives, 5(2), 126-142.
Haidar, J. I. (2012). Trade and productivity: Self-selection or learning-by-exporting in India.
Economic Modelling, 29(5), 1766-1773.
doi:https://doi.org/10.1016/j.econmod.2012.05.005
Heckman, J. (1979). Sample Selection as a Specification Error. Econometrica, 47, 153-161.
doi:https://doi.org/10.2307/1912352
Jakšić, S., Erjavec, N., & Cota, B. (2019). The role of foreign direct investment and labor
productivity in explaining Croatian regional export dynamics. Central European
Journal of Operations Research, 3(27), 835-849. doi:https://doi.org/10.1007/s10100-
018-0583-2
KAM. (2018). Manufacturing in Kenya Under the ‘Big 4 Agenda’: A Sector Deep-dive
Report. Nairobi.: KAM.
KAM. (2019). Closing the manufacturing gap through the Big 4 Agenda for shared
prosperity. Nairobi: KAM.
KAM. (2021). From surviving COVID-19 to thriving: Manufacturing sector rebound for
sustained job and investment growth. Nairobi: KAM.
KAM. (2022). Manufacturing manifesto 2022-2027. Nairobi: KAM.
Kasahara, H., & Lapham, B. (2013). Productivity and the decision to import and export:
Theory and evidence. Journal of international Economics, 89(2), 297-316.
KCCB. (2021). Tackling Youth Unemployment through revamping the manufacturing Sector
in Kenya: Post COVID. Nairobi: KCCB. Retrieved from
http://www.cjpc.kccb.or.ke/wp-content/uploads/2021/03/REPORT-ON-
ADDRESSING-YOUTH-UNEMPLOYMENT-THROUGH-
MANUFACTURING.doc
Kiendrebeogo, Y. (2020). Learning by exporting or self-selection into exporting? Middle East
Development Journal, 12(2), 304-325.
doi:https://doi.org/10.1080/17938120.2020.1756105
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
63
Krammer, S. M., Strange, R., & Lashitew, A. (2018). The export performance of emerging
economy firms: The influence of firm capabilities and institutional environments.
International Business Review, 27(1), 218-230.
doi:https://doi.org/10.1016/j.ibusrev.2017.07.003
Krugman, P. (1979). A model of innovation, technology transfer, and the world distribution
of income. Journal of political economy, 87(2), 253-266.
doi:https://doi.org/10.1086/260755
Kutner, M. H., Nachtsheim, C. J., & Neter, J. (2004). Applied Linear Regression Models ( 4th
ed. ed.). New York: McGraw-Hill Irwin.
López Rodríguez, J., & Serrano Orellana, B. (2020). Human capital and export performance
in the Spanish manufacturing firms. Baltic Journal of Management, 1(15), 99-119.
doi:https://doi.org/10.1108/bjm-04-2019-0143
Melitz, M. (2003). The impact of trade on aggregate industry productivity and intra-industry
reallocations. Econometrica, 71(6), 1695-1725. doi:https://doi.org/10.1111/1468-
0262.00467
Mubarik, M. S., Devadason, E. S., & Govindaraju, C. (2020). Human capital and export
performance of small and medium enterprises in Pakistan. International Journal of
Social Economics, 5(47), 643-662. doi:https://doi.org/10.1108/ijse-03-2019-0198
Mulliqi, A., Adnett, N., & Hisarciklilar, M. (2019). Human capital and exports: A micro-level
analysis of transition countries. Journal of International Trade & Economic
Development, 7(28), 775-800. doi:https://doi.org/10.1080/09638199.2019.1603319
Okado, A. D. (2013). Export propensity and intensity of Kenyan manufacturing firms: An
empirical analysis. Journal of Emerging Issues in Economics, Finance and Banking
(JEIEFB), 2(2), 638-654.
Papke, L. E., & Wooldridge, J. M. (1996). Econometric methods for fractional response
variables with an application to 401 (k) plan participation rates. Journal of applied
econometrics, 11(6), 619-632. doi:https://doi.org/10.1002/(sici)1099-
1255(199611)11:6%3C619::aid-jae418%3E3.0.co;2-1
Pham, T. T. (2015). Does exporting spur firm productivity? Evidence from Vietnam. Journal
of Southeast Asian Economies, 84-105. doi:https://doi.org/10.1355/ae32-1e
Ramsey, J. B. (1969). Tests for Specification Errors in Classical Linear Least Squares
Regression Analysis. Journal of the Royal Statistical Society, 31(2), 350–371.
doi:https://doi.org/10.1111/j.2517-6161.1969.tb00796.x
Reis , J., & Forte, R. (2016). . The impact of industry characteristics on firms’ export
intensity. International Area Studies Review, 19(3), 266-281.
Reis, J., & Forte, R. (2016). The impact of industry characteristics on firms’ export intensity.
International Area Studies Review, 3(19), 266-281.
doi:https://doi.org/10.1177/2233865916646560
Republic of Kenya. (2007). Kenya Vision 2030: A Competitive and Prosperous Kenya.
Nairobi: Government printer.
International Journal of Economics
ISSN 2518-8437 (Online)
Vol.9, Issue 2, No.4. pp 39 - 64, 2024
www.iprjb.org
64
Republic of Kenya. (2012). National Industrialization Policy Framework for Kenya: 2012-
2030. Sessional Paper No. 9 of 2012. Nairobi: Government Printer. Retrieved from
http://www.industrialization.go.ke/images/downloads/policies/the-national-
industrialization-policy.pdf
Republic of Kenya. (2017). The National Export Development and Promotion Strategy for
Kenya: 2017-2022. Nairobi: Government Publisher. Retrieved from
https://www.trade.go.ke/sites/default/files/NEDPS_Integrated_Strategy_1.pdf
Rialp-Criado, A., & Komochkova, K. (2017). Innovation strategy and export intensity of
Chinese SMEs: The moderating role of the home-country business environment.
Asian Business & Management, 16, 158-186. doi:https://doi.org/10.1057/s41291-017-
0018-2
Roberts, M. J., & Tybout, J. R. (1997). The decision to export in Columbia: an empirical
model of entry with sunk costs. American Economic Review, 87, 545-564.
Schwiebert, J. (2018). A sample selection model for fractional response variables. Working
Paper Series in Economics(No. 382 ).
Vu, H. V., Holmes, M., Tran, T. Q., & Lim, S. (2016). Firm Exporting and Productivity:
What If Productivity Is No Longer a Black Box. Baltic Journal of Economics, 16(2),
95–113. doi:https://doi.org/10.1080/1406099x.2016.1187382
Wagner, J. (2001). A note on the firm size–export relationship. Small business
economics(17), 229-237. doi:https://doi.org/10.1023/a:1012202405889
World Bank. (2021). World development Indicators. Washington, D.C.: World Bank.
World Bank. (2023). World Development Indicators. Washington, D.C.: World Bank.
