Vol. 36, No.

3, September 2021, 491-517

Journal of Economic Integration https://doi.org/10.11130/jei.2021.36.3.491
ⓒ 2021-Center for Economic Integration, Sejong University, All Rights Reserved. pISSN: 1225-651X eISSN: 1976-5525

Global Value Chain Participation, Competition, and Markups:

Evidence from Ethiopian Manufacturing Firms

Jieun Choi1, Emiko Fukase1+, and Albert G. Zeufack1

1World Bank, USA

Abstract This study uses detailed manufacturing census panel data for 2000 to 2014 to explore the relationship
between Ethiopian firms’ global value chain (GVC) participation and markups. We find that GVC firms,
defined as firms involved in both exporting and importing intermediate inputs, tend to have lower markups
relative to non-trading firms and firms that are involved only in material imports. Moreover, the more
intensely a firm is integrated into a GVC (measured by the share of export value added and imported
inputs in total sales), the lower its markup is. Finally, we explore competition effects at the industry level
and find that firms operating in industries with a relatively high GVC presence and suppliers selling inputs
to such industries tend to have lower markups owing to horizontal competition and backward linkages,
respectively. All of these findings suggest that GVC participation is associated with greater competition
for Ethiopian firms.

Keywords: competition, Ethiopia, global value chain, markup

JEL Classifications: D22, F14, F61, L11, O14

Received 28 November 2020, Revised 18 June 2021, Accepted 23 July 2021

I. Introduction

Recent studies show that market power, as measured by markups (i.e., firms’ abilities to
charge prices above their marginal costs), tends to be rising globally.1) This change has been
mainly driven by advanced economies, however, and the trend in developing countries is mixed
(De Loecker & Eeckhout, 2018; Diez et al., 2019). The emergence of global value chains (GVCs)
is likely to have contributed to this phenomenon, as the fragmentation of production and changing
trade patterns have influenced the competition that countries, industries, and firms face. The
share of GVC2) trade (i.e., global exports that flow across at least two borders) in total world
trade rose rapidly from about 40 percent in 1990 to 50 percent prior to the financial crisis

+Corresponding Author: Emiko Fukase

Consultant, World Bank, 1818 H Street, NW, Washington DC, 20433, USA. Tel: +1-917-873-4053.
Email: [email protected]/[email protected]
Co-Author: Jieun Choi
Senior Economist, World Bank, 1818 H Street, NW, Washington DC, 20433, USA. Email: [email protected]
Co-Author: Albert G. Zeufack
Chief Economist, World Bank, 1818 H Street, NW, Washington DC, 20433, USA. Email: [email protected]
492 Journal of Economic Integration Vol. 36, No. 3

in 2008, at which point it slowed somewhat (World Bank, 2019a).

The markups that firms derive from GVC participation may depend on the stages of the
GVCs in which they participate. In turn, these stages are influenced by countries’ economic
development and comparative advantages. The consolidation and resilience of GVCs affect the
levels of competition that economic agents face on boths the supply and demand sides of global
production and distribution networks. Competition can change the distribution of the value added
and markups across various producers by upgrading firms to higher value-added stages within
GVCs (Gereffi, 2011; World Economic Forum, 2018). Within domestic value chains, markups
may be influenced by the presence of GVC firms in an industry, which may affect both
horizontal competition within industry and competition across industries (through backward and
forward linkages). 1)2)

In the United States and other advanced economies, the growth in markups may be driven
by a relatively small number of superstar firms that are large, productive, highly profitable,
and able to extract large markups (Autor et al., 2017; De Loecker & Eeckhout, 2017; Diez
et al., 2019). Global exports are also dominated by a small group of very large superstars
(Freund & Pierola, 2015). Fernandes et al. (2016) report that larger and richer countries have
both more and larger exporters on average and a greater concentration of exports among the
top five percent of exporting firms. GVC activities may have disproportionately benefited large
firms in developed countries, as these firms can reduce their production costs by outsourcing
labor-intensive components to low-wage countries and can benefit from the economies of scale
created by exports (World Bank, 2019a).
In developing countries, the effects of GVC participation on local supplier markups and
the prospects for upgrading depend on the type of GVC, GVC governance patterns, and
countries’ capabilities (Gereffi, 2011). On one hand, GVCs offer firms in developing countries
opportunities to participate in global production networks in segments in which they have
comparative advantages and to potentially upgrade to higher value activities. For instance, China
has been strengthening its position in GVCs, improving its production mix, relying less on
imported inputs, and increasing its competitiveness in domestic input sectors (Gereffi, 2011;
Kee & Tang, 2016). Fragmenting production across borders also allows for a finer division
of labor across countries (World Bank, 2019a). Thus, GVCs may magnify the effects of trade
on growth, employment, and functional specialization (Hummels et al., 2001; Linden et al.,
2009; Timmer et al., 2019).
However, the benefits from GVCs may be unevenly distributed across and within countries;

1) Using financial statement data for more than 70,000 firms in 134 countries, De Loecker and Eeckhout (2018)
estimate that the average global markup rose from nearly 1.1 in 1980 to around 1.6 in 2016.
2) Different studies define GVCs differently, but most authors consider both imports and exports in defining a GVC;
Taglioni and Winkler (2016) provide an overview.
 Global Value Chain Participation, Competition, and Markups: Evidence from Ethiopian Manufacturing Firms 493

whereas large and productive buyers in developed countries tend to earn greater profits and
markups owing to GVC participation, suppliers in developing countries may be squeezed (Lianos &
Lombardi, 2016; World Bank, 2019a; World Economic Forum, 2018).3) This outcome may
be driven by fierce competition from other low-wage countries or the disproportionately high
bargaining power of leading firms from advanced countries over their suppliers. However, little
research has been conducted on the relationship between GVC participation and markups in
low-income countries.
This study therefore investigates the relationship between GVC participation and markups
in a low-income country, Ethiopia, using manufacturing census panel data from 2000 to 2014.
We explore whether GVC firms, defined as firms that both import material inputs and export
some of their products, have different markups relative to non-trading firms (referred to as
“non-traders”) and firms that participate in exports or raw material imports but not both (referred
to as “one-way traders”). We estimate firm-level markups using De Loecker and Warzynski’s
method (2012).
Ethiopia provides an interesting setting for studying this topic. Since the turn of the millennium,
Ethiopia’s gross domestic product has grown quickly. Its per capita income has nearly tripled
from $197 in 2000 to $570 in 20184) (World Bank, 2019b). With the government’s efforts to
industrialize and open the economy, the country’s manufacturing GVC participation expanded
substantially. Specifically, the number of GVC firms nearly tripled from 2000 and 2001 to 2013
and 2014.5) Although GVC firms account for only about four percent of Ethiopia’s manufacturing
firms, they tend to be large in terms of both sales and employment, as they accounted for
21 percent and 23 percent of total manufacturing sales and employment, respectively, in 2013
and 2014. Moreover, GVC firms are more productive, more capital intensive, and pay higher
wages relative to non-traders and one-way traders when we control for year, regional, and
sector fixed effects (see Figure 1 below).
Although numerous previous studies investigate the relationships of export status and access
to imported inputs with firm productivity, fewer studies consider the relationship between firms’
participation in trade and markups. Some models of international trade with heterogenous producers
predict that exporters have a markup premium, as productive firms self-select into export markets
and charge higher markups (Melitz & Ottaviano, 2008). Several studies find empirical evidence
supporting this prediction in advanced countries. For example, Bellone et al. (2016), Gullstrand

3) For example, in the textile and apparel GVCs, the World Bank (2019a) finds that markups and exports tend
to be negatively related for supplying firms in selected developing countries, whereas buying firms in developed
countries earn higher profits. Lianos and Lombardi (2016) report that powerful retail chains and large multinational
food processing companies in agri-food value chains have been squeezing upstream suppliers with their superior
bargaining power.
4) Dollar values are given at 2010 constant prices.
5) To mitigate the effects of yearly fluctuations, the earliest two years (2000 and 2001) and the latest two years
(2013 and 2014) are averaged.
494 Journal of Economic Integration Vol. 36, No. 3

et al. (2014), Kato (2014), and Békés et al. (2016) find markup premiums in France, Sweden,
Japan, and four European Union countries, respectively.
For transition economies, however, the results appear to be mixed. De Loecker and Warzynski
(2012) analyze data from Slovenia during its transition from a planned to a market economy
following its independence from the former Yugoslavia. They find that markups were significantly
higher for exporters than for non-exporters and increased when firms entered export markets.
In contrast, considering firm-level data from Poland over the period from 2002 to 2016, Gradzewicz
and Mućk find that markups in Poland, a supplier in European GVCs, have been falling over
the past 15 years, particularly for exporters. Thus, the relationship between exports and markups
remains an open question.
A reduction in the cost of imported inputs will reduce firms’ marginal costs of production
and increase their markups if they do not pass the entire cost savings through to the final price,
as De Loecker et al. (2016) show for India and Brandt et al. (2017) report for China. Specifically,
De Loecker et al. (2016) show that when input tariffs were reduced in India, firms that imported
cheaper inputs did not reduce their sales prices to reflect the entire cost savings, and, thus, they
increased their markups. Using data on Hungarian manufacturing firms, Hornok and Muraközy
(2019) find that firms that import intermediate inputs charge higher markups because importing
helps these firms produce higher quality outputs. However, importing more intermediate inputs
is not always associated with lower costs. For example, in the context of Poland’s participation in
the European value chain, Gradzewicz and Mućk (2019) find that increasing reliance on imported
inputs in production, along with rising competition among domestic firms for access to export
markets, is a major factor in the observed reduction in markups.
Our study contributes to the literature on trade, GVCs, and development, particularly the
literature on the relationship between GVC participation and competition. This study makes three
important advancements. First, to the best of our knowledge, no previous study explores the
relationship between GVC participation and markups in low-income countries. We aim to fill this
gap by providing a case study of a low-income country, Ethiopia. Using the pooled ordinary
least squares (OLS) method to conduct a cross-sectional analysis, we find that GVC firms tend
to have lower markups relative to non-trading firms and firms that are involved only in importing
raw materials. Investigating within-firm variation using a model with fixed effects, we observe that
firms participating in GVCs tend to experience reductions in their markups relative to non-trading
firms and firms involved only in imports. Our results are robust to including a variety of control
variables, such as firm characteristics and sector, year, and regional fixed effects.
Second, this study investigates whether the depth of participation in a GVC (measured as
the share of export value added and imported inputs in total sales) affects firms’ markups by
interacting firms’ GVC participation with their trade intensity. We find that the degree of trade
participation matters, that is, the more GVC firms are involved in exports and raw material
 Global Value Chain Participation, Competition, and Markups: Evidence from Ethiopian Manufacturing Firms 495

imports, the lower their markups are.

Third, we investigate whether firms’ markups are influenced by the presence of GVC firms
at the industry level. GVC firms may introduce competitive pressure within their industries
(horizontal competition) and may even extend this pressure to their suppliers and buyers through
backward and forward linkages, respectively. We find that firms operating in industries with
more GVC presence have lower markups through horizontal competition and that suppliers
of inputs to industries with more GVC presence also have lower markups through backward
linkages. All of these findings suggest that GVC participation is associated with greater
competition in Ethiopia.
The remainder of this paper proceeds as follows. Section II describes our measures of GVC
participation, the methods for obtaining markups, and trends in GVC participation and markups
for Ethiopian manufacturing firms. Section III explains our estimation strategy and reports the
regression results. Section IV presents our conclusions.

II. Estimating GVC Participation and Markups

A. GVC participation - estimation and trends

Many previous studies document the level of GVC participation at the country and industry
level using global input-output tables, such as the World Input-Output Database (WIOD)
(Timmer, 2012). This information reflects that for GVCs, unlike in the case of traditional trade,
gross exports and imports are not accurate measures of the exported domestic value added
and the foreign value added that is consumed, respectively. Studies using global input-output
tables provide more accurate bilateral trade flows and bilateral value-added balances (Johnson
& Noguera, 2012; Koopman et al., 2014).
However, such global input-output tables as WIOD are unavailable for Ethiopia. Furthermore,
despite their widespread use, these tables do not measure firms’ participation in GVCs, as they
rely on aggregated input-output data. Their sectoral disaggregations of GVC flows tend to be
coarse, as they miss GVC activity within broadly defined sectors. Researchers are forced to impose
strong assumptions in constructing the tables to estimate some bilateral intermediate input trade
flows that cannot be readily determined from customs data or national input-output tables.
This study uses firm-level panel data on large and medium manufacturing industries (LMMI)
(hereinafter, Ethiopian Manufacturing Census Data) collected annually by the Ethiopian Central
Statistical Agency (CSA) from 2000 through 2014. The census covers all Ethiopian manufacturing
establishments that have at least 10 employees and use electricity for production.6) The dataset

6) In 2011, the CSA changed its survey questionnaire and firm identification numbers. We merge the data for 2000
496 Journal of Economic Integration Vol. 36, No. 3

provides detailed information on each firm’s location, number of employees, capital, sales,
production, exports and material imports, and wages and benefits. We classify manufacturing
establishments according to the International Standard Industrial Classification (ISIC) at the
four-digit level. We aggregate the data at the two-digit level to ensure that we have sufficient
observations to run production functions for each industry.7)
We consider two firm-level measures of GVC participation:

•     = 1 if firm i participates in both exporting and importing inputs at time t.
      
•      =  , where      is the trade intensity
  
of firm i at time t,   is the value-added8) portion of export sales,     is the
value of imported intermediate inputs, and    is the value of total sales.9)

Our first measure of GVC participation, GVC status, is a dummy variable equal to one for
firms that both export and import intermediate inputs (GVC = 1). We compare the performance
of GVC firms with those of one-way traders and non-traders, that is, firms that export but
do not import raw materials (OnlyX = 1), firms that do not export but import at least some
of their materials (OnlyM = 1), and firms that neither export products nor import raw materials
(NoXM = 1). We group firms into these four categories because participating in imports or exports
may require different and possibly higher capabilities than not participating in trade requires.
Although these dummy variables are suitable for classifying firms into four mutually exclusive
categories, however, they do not provide information about the extent of participation by firms
in each category. Thus, in the empirical section (Section III.A), we consider a second measure
of trade participation in which we interact trade status with trade intensity (regressions (5)
and (6) in Table 4).

to 2011 with the data for 2011 to 2014 data using electricity identification numbers. These identification numbers
are not available for some firms in the dataset, and this limitation of the merged panel data should be considered
when interpreting the results. In the econometric analysis, we estimate a regression using only data for the years
2000 to 2010 and find similar results (see regression (2) in Table 5).
7) Specifically, we estimate production functions for 14 ISIC categories at the two-digit level. We include some
ISIC industries with few observations under other ISIC codes. Tobacco products (ISIC 16) are included in food
and beverages (ISIC 15), and coke and petroleum products (ISIC 23) are included in chemical products (ISIC
24). We also combine machinery and equipment (ISIC 29); office, accounting, and computing machinery (ISIC
30); electrical machinery (ISIC 31); radio, TV, and communication equipment (ISIC 32); medical, precision, and
optical instruments (ISIC 33); and motor vehicles (ISIC 34) into one machinery category (ISIC 29-34).
8) Value-added is computed as follows: value-added = production - raw materials costs - other industrial costs -
other non-industrial expenses. Here, production ≈ sales + change in stock.
9) Adding (gross) export sales and imported material purchase values entails double counting because export sales
include the cost of imported raw materials. To avoid this issue, we sum the value-added component of exports
and imported materials and then divide this sum by total sales. We implicitly assume that the proportion of
value-added in total sales is the same for the domestic and export markets.
 Global Value Chain Participation, Competition, and Markups: Evidence from Ethiopian Manufacturing Firms 497

Our GVC measures are consistent with previous studies that define a GVC as a series of
stages in the production of a product or service for sale to consumers such that each stage adds
value and at least two stages are in different countries (World Bank, 2019a). Although the
LMMI data do not provide direct information on firm-to-firm transactions across countries, our
measures allow us to capture fine variation in GVC participation across firms within aggregated
industries and do not require us to assume that bilateral intermediate input trade flows are
the same within broadly defined sectors. Additionally, our measures allow us to compare levels
of GVC participation across different firms. However, our GVC measures neither specify the
form that the foreign value added in production takes, although it is often associated with
trade in raw materials, intermediate inputs, and tasks, nor do they specify the shape of a GVC
as either vertical or horizontal specializations crossing borders multiple times (Los et al., 2015;
Wang et al., 2013).

Table 1. Number of Establishments, Sales and Employment of Ethiopian Manufacturing Firms 2000-2014, Total vs.
GVC Firms
Number of Establishments Sales (million Birr)a Employment
Annual Annual Annual
2000/2001 2013/2014 2000/2001 2013/2014 2000/2001 2013/2014
Growth (%) Growth (%) Growth (%)
Total 708 2394 9.4 14476 73448 12.5 74987 199174 7.5
GVC firms 32 88 7.7 3565 15646 11.4 14828 45032 8.6
GVC/Total (%) 4.5 3.7 24.6 21.3 19.8 22.6
Notes: Only establishments which employed ten or more employees are included.
a
Sales deflated using manufacturing value added deflator available in FAOSTAT.
(Source) Ethiopian Manufacturing Census Data 2000-2014.

Table 1 shows trends in the number, sales, and employment of GVC firms relative to all
manufacturing firms in the dataset. The first three columns in the table show that the number
of Ethiopian establishments employing ten or more workers more than tripled from 708 to
2,394 establishments over the period from 2000 and 2001 to 2013 and 2014. During that same
period, the number of GVC firms rose from 32 (4.5 percent of firms in the dataset) to 88
(3.7 percent of firms in the dataset).
The middle three columns show that the total (deflated) sales generated by the firms in our
dataset rose from 14,476 million Ethiopian birr (ETB) in 2000 and 2001 to ETB 73,448 million
in 2013 and 2014, and the last three columns show that number of workers that these firms
employed increased from 74,987 to 199,174. Annual output growth (12.5 percent) was faster
than annual employment growth (7.5 percent) over this period, implying that sales per worker
rose substantially. During the same period, sales generated by GVC firms rose from ETB 3,565
million in 2000 and 2001 to ETB 15,646 million in 2013 and 2014, implying an annual growth
rate of 11.4 percent. By comparison, employment grew from 14,828 workers in 2000 and 2001
to 45,032 workers in 2013 and 2014, an average growth rate of 8.6 percent. These statistics
498 Journal of Economic Integration Vol. 36, No. 3

imply that although Ethiopia has only a small number of GVC firms, they play an important
role in the manufacturing sector, as they are responsible for substantial shares of sales (24.6
percent of total sales in 2000 and 2001 and 21.3 percent of total sales in 2013 and 2014) and
employment (19.8 percent of total employment in 2000 and 2001 and 22.6 percent of total
employment in 2013 and 2014).

Figure 1. Capital intensity, productivity and wage differences

between trading firms and non-traders (%)

Notes: Figure 1 shows the percentage differences in log of capital intensity, log of labor productivity
and log of wages between non-traders and trading firms. The results are obtained by regressing
the log of capital intensity (capital per worker), log of labor productivity (sales per worker)
and log of wages on dummy variables if the firm exports and imports inputs (GVC), exports
only (OnlyX), and imports intermediate inputs only (OnlyM), controlling for sector, year, and
regional fixed effects, as well as whether the firm is state-owned. Non-traders (NoXM) are
omitted as the reference category. The labor productivity regression also controls for capital
per worker. The percent differences reported in the figure are obtained as 100 multiplied by
the exponential of the coefficient estimates minus 1.
(Source) Ethiopian Manufacturing Census Data 2000-2014.

Figure 1 compares GVC firms with non-traders (NoXM) and one-way traders (OnlyX and
OnlyM) in terms of capital intensity, labor productivity (i.e., sales per worker), and wage levels.
The figure shows that GVC firms tend to be more capital-intensive and more productive (defined as
labor productivity controlling for capital intensity) and pay higher wages relative to one-way traders
or non-traders, controlling for regional, year, and industry fixed effects. The better performance
of GVC firms is consistent with the large body of previous literature focusing on firm heterogeneity
in international trade. Studies find that exporting firms, including those in sub-Saharan African
countries (e.g., Bigsten et al., 2004; Fatou & Choi, 2013; Mengistae & Pattillo, 2004; Van
Biesebroeck, 2005), tend to be larger and more productive and pay higher wages relative to
non-exporting firms.
The observed better performance of GVC firms is also in line with previous work reporting
the positive effects of access to imported inputs on productivity and product quality. For example,
 Global Value Chain Participation, Competition, and Markups: Evidence from Ethiopian Manufacturing Firms 499

Abreha (2019) and Bigsten et al. (2016) find this result for Ethiopia. Amiti and Konings (2007),
Halpern et al. (2015), and Topalova and Khandelwal (2011) consider Indonesia, Hungary, and
India, respectively. Moreover, some studies, such as those of Castellani et al. (2010) in Italy,
Smeets and Warzynski (2013) in Denmark, and Vogel and Wagner (2010) in Germany, find
that two-way traders are more productive on average than firms that only import or export.
This result is partly because the ability to export sometimes depends on the ability to import,
as access to better intermediate inputs or new input goods and varieties facilitates product
upgrades and creates new possibilities for exports. Studies by Bas and Strauss-Kahn (2015)
in China and Goldberg et al. (2009) in India support this finding.
Table 2 reports the average proportions of GVC firms, one-way traders (OnlyX and OnlyM),
and non-traders (NoXM) by industry over the period from 2000 to 2014. Column 1 shows that
GVC firms are concentrated in a few sectors. For example, they make up 25.5 percent of leather
and footwear (ISIC 19), 19.7 percent of textiles (ISIC 17), 10.7 percent of apparel (ISIC 18),
and 3.7 percent of food and beverage (ISIC 15) firms.
Overall, Ethiopian manufacturing firms are characterized by low export participation; on
average, just 4.7 percent of firms participated in exporting over the period from 2000 to 2014
(3.8 percent of all firms are GVC firms, and 0.9 percent are OnlyX firms). In contrast, trade
participation on the import side (i.e., GVC and OnlyM firms) is high, with nearly two thirds
(63.7 percent) of firms purchasing imported materials on average (3.8 percent of all firms are

Table 2. GVC and One-way Trade Participation by Industry (%)

Four Firm Categories GVC + GVC+
ISIC Description
GVC OnlyX OnlyM NoXM OnlyX OnlyM
15 Food and Beverages 3.7 1.3 53.1 41.8 5.1 56.8
17 Textiles 19.7 4.4 44.2 31.7 24.0 63.9
18 Apparel 10.7 5.0 54.2 30.1 15.7 64.9
19 Leather & Footwear 25.5 3.2 50.8 20.6 28.6 76.3
20 Wood products 0.1 0.2 44.2 55.5 0.3 44.3
21 Paper & Paper products 1.1 0.0 90.7 8.2 1.1 91.8
22 Publishing & Printing 0.2 0.1 81.9 17.7 0.4 82.1
24 Chemicals 2.6 0.4 86.1 10.9 3.0 88.7
25 Rubber & Plastics 0.8 0.1 90.8 8.2 1.0 91.6
26 Non-metallic mineral products 0.7 0.5 26.1 72.8 1.1 26.7
27 Basic metals 0.6 0.5 94.3 4.5 1.2 95.0
28 Fabricated Metal products 1.0 0.2 78.6 20.2 1.2 79.6
29-34 Machinery 29,30,31,32,33,34 1.0 0.4 81.2 17.4 1.4 82.2
36 Furniture 0.3 0.2 78.2 21.3 0.5 78.6
Total 3.8 0.9 59.9 35.3 4.7 63.7
(Source) Ethiopian Manufacturing Census Data 2000-2014.
500 Journal of Economic Integration Vol. 36, No. 3

GVC firms, and 59.9 percent are OnlyM firms). From 2000 to 2014, 35.3 percent of establishments
participated in neither exports nor material imports.

B. Estimating markups

Following De Loecker and Warzynski (2012), we use the firm-level markup, that is, the
percentage difference between a firm’s marginal cost and its selling price, as a measure of
competition. We assume that firms are cost-minimizing and that technology can be described
by a Cobb-Douglas production function. We estimate separate production functions for each
two-digit ISIC sector using Ackerberg et al.’s (2006) method.
A cost-minimizing firm i produces output at time t using the following production technology:

  =    (           )

where  ,   , and    denote labor, intermediate inputs, and capital assets, respectively, and
   is the firm-specific productivity.    ( ) is continuous and is twice differentiable with respect
to its arguments. The associated Lagrangian function is:

ℒ (     ,    ,  ) =      +  
           +    (   −    ( ⋅ )),

where    
  and   denote the labor, materials, and capital prices, respectively, for a firm.

The first-order condition for material inputs is

 ℒ     · 
•  =    −    = 0 , where the marginal cost of production at a given
     
 ℒ 
level of output is  , as  =  . Rearranging the terms and multiplying both sides
  

       ·         
by  , we obtain   =   .
        

The variable on the left-hand side is the output elasticity of the material input denoted as
ϴ
 . Thus, the equation implies that a cost-minimizing firm sets its optimal material input such
that the output elasticity of the material input equals the product of the material price and
the share of the material input in production divided by marginal cost of production at a given
level of output. Here, we use material inputs to estimate markups. However, this marginal
 Global Value Chain Participation, Competition, and Markups: Evidence from Ethiopian Manufacturing Firms 501

cost of production can be derived from the first-order condition for any non-dynamic input
variable (e.g., labor input).
We deflate sales and input costs as proxies for quantity terms.10) Thus, our estimated markups
may include information besides the level of competition, such as the perceived quality of products
and a firm’s ability to produce differentiated goods. However, De Loecker and Warzynski (2012)
show that although deflated sales may alter the levels of markups, they do not affect the correlation
between markups and firm-level characteristics or changes in markups over time.
 
Defining the markup   as the ratio of the price to the marginal cost,   ≡  , we
 
can estimate the markup using the material coefficient ϴ 
 .

 


   

  
ϴ
     and   = ϴ   
       

Because the sales term (    ) includes ɛ , an unobservable and unanticipated shock to

production (e.g., exogenous circumstances and unanticipated delays), and independent and identically
distributed shocks, including measurement error, we remove ɛ by estimating sales as   
 ɛ .   

Thus, the markup is estimated as

 

   
  = ϴ
  exp (
ɛ ). (1)
   

Table 3 shows the median markup by industry on average over the period from 2001 to
2014. The first column shows that the median markup in Ethiopian manufacturing firms is 1.55.
The markups vary across industries. Non-metallic mineral products (ISIC 26) has a relatively
high estimated markup of 2.24, whereas those of chemicals (ISIC 24), basic metals (ISIC 27),
food and beverages (ISIC 15), and leather and footwear (ISIC 19) are relatively low at 1.19,
1.22, 1.29, and 1.30, respectively. Columns 2-3 of Table 3 show that the median markup declined
from 1.62 in 2001 and 2002 to 1.52 in 2013 and 2014.

10) We define capital as the value of fixed assets at the beginning of the year deflated by the gross fixed capital
formation deflator provided by the Food and Agriculture Organization Statistical Database (FAOSTAT). The values
of sales and material inputs are deflated by the manufacturing value-added deflator obtained from FAOSTAT.
502 Journal of Economic Integration Vol. 36, No. 3

Table 3. Median Markups by Industry

ISIC Description 2001-2014 2001/2002 2013/2014
15 Food and Beverages 1.29 1.31 1.19
17 Textiles 1.40 1.61 1.33
18 Apparel 1.58 1.55 1.79
19 Leather & Footwear 1.30 1.21 1.36
20 Wood products 1.85 2.69 1.61
21 Paper & Paper products 1.83 1.59 1.58
22 Publishing & Printing 1.72 1.83 1.70
24 Chemicals 1.19 1.18 1.25
25 Rubber & Plastics 1.54 1.71 1.40
26 Non-metallic mineral products 2.24 2.25 2.40
27 Basic metals 1.22 1.11 1.20
28 Fabricated Metal products 1.52 1.80 1.61
29-34 Machinery 29,30,31,32,33,34 1.74 2.03 1.33
36 Furniture 1.84 2.10 1.76
Total 1.55 1.62 1.52
(Source) Ethiopian Manufacturing Census Data 2000-2014.

Figure 2. Evolution of median markup 2001-2014. Figure 3. Distribution of log of markups

by trade participation status.
1.5
3
2.5

1
2

.5
1.5
1

2000 2005 2010 2015

Year -1 0 1 2 3
Trend line Industry markup Median markup GVC OnlyX OnlyM NoXM

(Source) Ethiopian Manufacturing Census Data 2000-2014. (Source) Ethiopian Manufacturing Census Data 2000-2014.

Figure 2 shows the historic trends in estimated markups over the period from 2001 to 2014.
We show the median markups for the manufacturing industry as a whole and the median markups
for each industry. The estimated markups fluctuate from year to year, but the figure shows that
markups have slightly decreased over the estimation period.
Figure 3 shows the distribution of markups across the four trade participation categories.
The markup distributions for OnlyM and NoXM firms lie to the right of the markup distributions
for GVC and OnlyX firms, suggesting that GVC firms and firms that participate only in exports may
have lower markups relative to non-trading firms and firms that participate only in material imports.
 Global Value Chain Participation, Competition, and Markups: Evidence from Ethiopian Manufacturing Firms 503

III. Estimation Strategy and Regression Results

A. GVC participation and markups

Our regression model is

ln μ  = α    + β   + γ  + +  +  , (2)

where ln μ  is the logarithm of the markup for firm i in year t, defined as in equation (1) (De
Loecker & Warzynski, 2012).   is the set of GVC participation measures defined in Section
II.A, namely,     and     , and   is a set of firm characteristics,
including the logarithm of labor as a proxy for firm size and the logarithm of capital to capture
capital intensity.
State ownership may affect the level of market competition through various channels, such
as stricter regulations, barriers to entry for new and foreign firms, and government supports
and incentives that disproportionally benefit state-owned firms (Dewenter & Malatesta, 2001;
Gupta, 2005; Megginson & Netter, 2001). Consistent with the government’s privatization programs
(Oqubay, 2015), the share of state-owned firms in total firms in the Ethiopian Manufacturing
Census declined from 17.4 percent in 2000 to 3.7 percent in 2014. To control for the effect of
state ownership, we include a dummy variable indicating whether a firm is state-owned11) in our
regressions (this variable equals one for state-owned firms and zero for private firms).  is a
set of industry-specific variables, such as the upstream index (Antràs et al., 2012) and industry
dummies (i.e., 19 ISIC dummies), ρt is a set of year dummies (for 14 years), τl is a set of
location dummies (for 12 regions), and   is the error term. Appendix Table 1 reports summary
statistics for the variables included in the regressions.
Table 4 shows the results of estimating a series of pooled OLS regressions relating markups
with measures of trade participation. Controlling for firm characteristics (regression (1)), the
coefficients of the trade participation status dummies are significantly negative, suggesting that
firms’ trade participation is negatively associated with their markups. In particular, the coefficient of
GVC firms (GVC) is the largest in magnitude, followed by the coefficients of export only (OnlyX)
and import only firms (OnlyM). Adding regional and year dummies does not greatly alter the
results (regression (2)). Introducing ISIC dummies, as in regression (3), considerably reduces
the explanatory power of the GVC and one-way trade participation dummies, as the coefficients
of OnlyX and OnlyM lose statistical significance, and the coefficient of GVC firms is significant
only at the ten percent level. This result may arise because controlling for industry dummies may

11) The census asks if a firm is private, public, or public and private. The firms that responded “public” are defined
as state-owned. The census provides no information on whether a firm is foreign.
504 Journal of Economic Integration Vol. 36, No. 3

purge industry-level characteristics that are correlated with trade participation status.
Regression (4) includes the upstreamness index12), which measures the relative position of an
industry in the production chain (Antràs et al., 2012), in place of ISIC dummies. The upstreamness
index is always greater than or equal to one (≥ 1). Larger values are associated with relatively
higher levels of upstreamness of an industry’s output. The coefficient of the upstreamness index is
significantly positive, suggesting that firms operating in more upstream industries tend to have
higher markups. Notably, some of the industries in which Ethiopia’s GVC firms are concentrated,
namely, apparel and leather and footwear, appear to be downstream industries, with upstreamness
indexes of 1.01 and 1.04, respectively. The lack of competition in upstream sectors, that is, industries
that tend to supply raw materials or intermediate inputs, may partly explain why so many Ethiopian
firms rely on imported intermediary inputs for production rather than using domestic inputs
produced by firms with high markups. The coefficients of the GVC and one-way trade participation
dummies remain essentially unchanged when we adjust for the upstreamness of industries.
In all of the specifications, the coefficients of the dummies for GVC participation are significantly
negative, suggesting that GVC firms charge lower markups than non-trading firms (NoXM) do.
To test whether GVC firms charge different markups than one-way traders do, we conduct an
F-test for the statistical difference between the coefficient of GVC status and coefficients of
firms involved in only exports (OnlyX = GVC) and only importing materials (OnlyM = GVC).
Table 4 reports the results. In two of the four specifications, OnlyM = GVC can be rejected,
implying that GVC-participating firms have lower markups than firms participating only in
material imports do. OnlyX = GVC cannot be rejected in any of the specifications, and, thus,
we find no evidence that GVC-participating firms and firms involved in exports but not material
imports charge different markups.
The negative coefficient of the OnlyM dummy variable may seem counterintuitive and
inconsistent with previous studies finding that a decrease in the cost of imported inputs positively
affects markups (e.g., De Loecker et al., 2016) and previous studies suggesting that importing
intermediate inputs is associated with a markup premium because these imports lead to higher-
quality outputs (e.g., Hornok & Muraközy, 2019). A potential explanation is that importing materials
may be costlier than sourcing domestic inputs, as is the case in Poland (Gradzewicz & Mućk, 2019),
if, for instance, imported goods are more expensive owing to their higher quality13) or require
higher transportation or transaction costs. With incomplete passthrough of costs to consumers,
importing firms may need to lower their markups. Another potential explanation is that the coefficient

12) The upstreamness index is an industry measure of relative production line positions that is computed as the average
distance from final use in the production chain (Antràs et al., 2012). This index is calculated using the input-output
table for Ethiopia from the Global Trade Analysis Project (GTAP) database for 2011 (https://www.gtap.agecon.purd
ue.edu/). We map ISIC codes at the four-digit level for each firm to GTAP categories using the concordance
developed by Francois et al. (2014).
13) Abreha (2019) reports that most of Ethiopia’s manufacturing imports come from developed countries.
 Global Value Chain Participation, Competition, and Markups: Evidence from Ethiopian Manufacturing Firms 505

estimate for OnlyM is biased owing to the omission of variables that represent a competitive
business environment (relative to the environment faced by non-traders) and that are correlated
with both import participation and markups. Many variables, such as better infrastructure;14)
a supportive business and investment climate; and favorable geographical conditions, such as
access to seaports, may positively affect firms’ import participation and negatively affect their
markups. Not controlling for these variables may have caused us to overestimate the coefficient
of the import participation dummy. We address this potential problem by controlling for all
(time-invariant) firm- and location-specific characteristics using a fixed effects model, and we
find that the coefficient of OnlyM is insignificant (see regressions (1)-(3) in Table 5).
Finally, regression (5) estimates the relation between the depth of trade and markups by
interacting the trade participation variables with trade intensity. The negative slopes of the
interaction terms imply that for a given trade participation status, firms with higher trade
intensities tend to have lower markups. Based on the estimated coefficients for regression (5),
Figure 4 shows differences in the predicted means of markups depending on trade intensity.

Table 4. Pooled OLS Regression Results Relating Trade Participation and Markups
(1) (2) (3) (4) (5) (6) (7)
Participation dummiesa:
OnlyX -0.109** -0.103** 0.015 -0.094** -0.068 0.019
(0.048) (0.047) (0.043) (0.047) (0.070) (0.056)
OnlyM -0.084*** -0.095*** -0.007 -0.126*** 0.072*** 0.106***
(0.010) (0.010) (0.009) (0.010) (0.012) (0.010)
GVC -0.163*** -0.177*** -0.039* -0.160*** -0.004 0.064*
(0.023) (0.023) (0.022) (0.023) (0.037) (0.036)
Participation dummies*Trade intensity:
OnlyX*Trade intensity -0.112 -0.057
(0.106) (0.090)
OnlyM*Trade intensity -0.520*** -0.503***
(0.017) (0.019)
GVC*Trade intensity -0.460*** -0.514***
(0.100) (0.100)
Trade Intensity -0.379***
(0.016)
State ownership 0.152*** 0.130*** 0.092*** 0.118*** 0.059*** 0.060***
(= 1 state, = 0 private) (0.018) (0.018) (0.015) (0.018) (0.015) (0.015)
Firm size (log of labor) 0.036*** 0.044*** 0.058*** 0.044*** 0.056*** 0.060***
(0.005) (0.005) (0.005) (0.005) (0.005) (0.005)

14) For instance, Jones et al. (2018) report that firms that operate in markets with high transport costs appear to
be shielded from competition to some extent. Because high transport costs may be negatively correlated with
import participation and positively associated with markups, not controlling for transport costs may bias downwards
the coefficient of the dummy variable for firms that only import materials.
506 Journal of Economic Integration Vol. 36, No. 3

Table 4. Continued
(1) (2) (3) (4) (5) (6) (7)
Log of capital -0.022*** -0.021*** -0.001 -0.019*** 0.003 0.002
(0.002) (0.002) (0.002) (0.002) (0.002) (0.002)
Upstream index 0.122***
(0.007)
Constant 0.727*** 0.740*** -0.192** 0.470*** 0.564*** 0.462*** 0.224**
(0.024) (0.110) (0.090) (0.090) (0.008) (0.031) (0.096)
Year dummies X X X X X
Regional dummies X X X X X
ISIC dummies X X X
Number of observations 10324 10324 10324 10323 10227 10227 10227
R2 0.036 0.062 0.297 0.082 0.071 0.339 0.329
F-stat (OnlyX = GVC) 1.12 2.18 1.40 1.74
Prob F > 0 0.290 0.140 0.238 0.186
F-start (OnlyM = GVC) 13.91 15.21 2.38 2.50
Prob F > 0 0.000 0.000 0.123 0.114
F-stat (OnlyX = OnlyX*Trade intensity = 0) 3.530 0.01
Prob F > 0 0.029 0.812
F-stat (OnlyM = OnlyM*Trade intensity = 0) 602.71 367.27
Prob F > 0 0.000 0.000
F-stat (GVC = GVC*Trade intensity = 0) 33.79 19.65
Prob F > 0 0.000 0.000
Notes: The dependent variable is the log of markup, estimated using the ACF method.
a
Firms that have no trade (NoXM) are omitted as a reference category.
Standard errors are in parentheses. *, **, and *** indicate that the coefficients are significant at the 10, 5, and 1
percent level, respectively.

Figure 4. Trade intensity and predicted means of markups.

Notes: The predicted means of markups are calculated based on the coefficient
estimates reported in regression (5) in Table 4. To obtain predicted markup
levels, predicted log of markups are exponentiated.
(Source) Ethiopian Manufacturing Census Data 2000-2014.
 Global Value Chain Participation, Competition, and Markups: Evidence from Ethiopian Manufacturing Firms 507

As trade intensity increases, the markups of GVC firms and firms that only import materials
(OnlyM) decrease and approach one, whereas the slope for firms involved only in exports
(OnlyX) is less steep.
The joint significance tests reported in the last six rows of Table 4 reveal that the coefficients
of trade participation status and the interaction term are jointly significant for GVC and OnlyM
firms at the one percent level and for OnlyX firms at the five percent level. When we introduce
firm characteristics and region, year, and ISIC dummies as control variables in regression (6), the
coefficients of the interaction terms remain negative. However, the results of the joint significance
tests show that the two variables are no longer jointly significant for OnlyX firms, whereas they
remain jointly significant for GVC and OnlyM firms at the one percent level. The somewhat
imprecise results for OnlyX firms may partly reflect the small number of OnlyX firms, which
comprise about one percent of all firms. Evaluated at the sample means of trade intensity, the
implied combined impacts of trade participation and intensity for GVC and OnlyM firms relative to
non-trading firms (NoXM) are -0.056 and -0.040, respectively, in logarithm terms.15) Regression (7)
is an alternative specification that includes the trade intensity variable for each firm but excludes the
trade participation dummy variables. The coefficient of the trade intensity variable is significantly
negative at the one percent level, confirming that deeper trade participation is associated with lower
markups. In all of the regressions, the coefficient of state ownership is significantly positive at the
one percent level, suggesting that state-owned firms charge higher markups than private firms do.
Table 5 reports the results of regressions using the fixed effects model, which utilizes the variation
within firms. Regression (1) includes controls for firm characteristics and year dummies. Because
a firm’s location and ISIC industry do not typically change over time, we do not include regional
and ISIC dummies in this model. Although the magnitudes of the coefficients are smaller relative
to the estimates obtained by pooled OLS, the coefficients of the GVC and OnlyX dummies
remain significantly negative, suggesting that firms tend to lower their markups when they
participate in either exports or both exports and material imports. Because the fixed effects model
controls for any time-invariant unobservable firm- and location-specific effects, the coefficients of
the OnlyM dummy are not significant, implying that non-trading firms do not alter their markups
when they begin to import material inputs. The F-test results show that the hypothesis OnlyM =
GVC can be rejected, confirming that participating in GVCs and only importing material inputs
have different impacts on firms. One caveat for our merged panel dataset is that the firm identification
numbers changed around 2011, and merging the data before and after 2011 led to some firms
being excluded. As a robustness check, we run a regression using only the data from 2001 to
2010 and find similar results to those for the full sample (regression (2)).

15) The effect of GVC participation in logarithm terms is calculated at the mean of trade intensity (0.235): -0.056 ≈
0.064 -0.514*0.235.
508 Journal of Economic Integration Vol. 36, No. 3

Table 5. Fixed Effects Regression Results Relating Trade Participation and Markups
(1) (2) (3) (4)
Participation dummiesa:
OnlyX -0.085** -0.089* -0.033
(0.039) (0.046) (0.040)
OnlyM -0.011 -0.000 0.077***
(0.010) (0.013) (0.012)
GVC -0.083** -0.095** -0.039
(0.034) (0.038) (0.045)
Participation dummies*Trade intensity
OnlyX*Trade intensity -0.119
(0.119)
OnlyM*Trade intensity -0.423***
(0.030)
GVC*Trade intensity -0.183
(0.152)
Trade intensity -0.323***
(0.026)
State ownership -0.025 -0.018 -0.027 -0.027
(= 1 state, = 0 private) (0.029) (0.031) (0.029) (0.029)
Firm size (log of labor) 0.090*** 0.066*** 0.095*** 0.098***
(0.010) (0.012) (0.010) (0.010)
Log of capital 0.008* 0.013** 0.008* 0.008*
(0.004) (0.005) (0.004) (0.004)
Constant 0.100 0.101 0.120* 0.205
(0.069) (0.080) (0.067) (0.145)
Year dummies X X X
Number of observations 10324 6910 10227 10227
R2 0.065 0.023 0.119 0.117
F-stat (OnlyX = GVC) 0.00 0.01
Prob F > 0 0.955 0.907
F-start (OnlyM = GVC) 4.75 6.70
Prob F > 0 0.029 0.009
F-stat (OnlyX = OnlyX*Trade intensity = 0) 1.76
Prob F > 0 0.172
F-stat (OnlyM = OnlyM*Trade intensity = 0) 96.95
Prob F > 0 0.000
F-stat (GVC= GVC*Trade intensity = 0) 2.98
Prob F > 0 0.051
Notes: The dependent variable is the log of markup, estimated by the ACF method.
a
Firms that have no trade (NoXM) are omitted as a reference category.
Standard errors are in parentheses. *, **, and *** indicate that the coefficients are significant at the 10, 5, and 1
percent level, respectively.
 Global Value Chain Participation, Competition, and Markups: Evidence from Ethiopian Manufacturing Firms 509

Regression (3) includes the interactions between the trade participation dummies and trade
intensity, controlling for firm characteristics and year dummies. We find that the coefficients of the
interactions between participation status and intensity are negative for GVC, OnlyX, and OnlyM
firms, confirming the negative association between trade intensity and markups. The joint significance
tests confirm that the trade participation and interaction terms are jointly significant for GVC
firms and OnlyM firms at the five percent and one percent levels, respectively. The coefficients
of OnlyX and OnlyX*Trade intensity are not jointly significant. As in the pooled OLS model,
regression (4) includes only the trade intensity variable rather than both the trade participation
and intensity variables. The coefficient of the trade intensity variable is significantly negative at
the one percent level, confirming that an increase in trade intensity is associated with a reduction
in firms’ markups. The coefficient of state ownership is insignificant in all of the specifications,
most likely because firm ownership type does not frequently change.

B. Does the presence of GVC firms at the industry level influence firms’
markups through horizontal, backward, and forward linkages?

The presence of GVC firms within a domestic value chain may influence competition both
within and across industries. GVC firms may increase competition within an industry (horizontal
competition), potentially reducing all firms’ markups. They may also influence competition
vertically. Similar to arguments about spillovers from foreign direct investment (FDI) firms
(Javorcik, 2004)16), GVC firms in downstream industries (i.e., potential buyers) may influence
competition among upstream suppliers through backward linkages, and their presence in upstream
industries (i.e., potential suppliers of intermediate inputs) may affect the performance of and
competition experienced by downstream firms through forward linkages. GVC firms can affect
competition at different stages of production in many ways; for instance, an emerging cartel
may increase downstream or upstream prices, a horizontal or vertical merger may create market
power, and a dominant firm within a GVC may abuse its market position to extend its dominance
upstream or downstream (World Economic Forum, 2018).
This sub-section explores the influence of the presence of GVC firms, proxied by the share of
sales generated by such firms relative to total sales by industry and year, on firms’ markups through
horizontal, backward, and forward channels. Combining the Ethiopian Manufacturing Census
Data for 2000 to 2014 with an input-output table for 2011 taken from the GTAP database,17)
we compute horizontal, backward, and forward variables using a similar methodology to that

16) Javorcik (2004) investigates productivity spillovers from FDI firms to domestic firms through horizontal, backward,
and forward linkages. Borrowing her methodology, we investigate whether the presence of a GVC firm at the
industry level influences firm-level competition through horizontal, backward, and forward linkages.
17) We map ISIC codes at the four-digit level, which are available for each firm, to 19 GTAP categories using
the concordance developed by Francois et al. (2014).
510 Journal of Economic Integration Vol. 36, No. 3

developed by Javorcik (2004). Because the data rely on the manufacturing census, the backward
and forward linkages measured in this section reflect only linkages within manufacturing.
The Horizontal variable is used to proxy for the extent of the presence of GVC firms in
industry j at time t and is defined as

=    for   ∈           ̸   for   ∈    .

Here,     is the share of sales generated by GVC firm i at time t, and    is
firm i’s total sales at time t.
The Backward variable is designed to capture the presence of GVC firms in the industries
that are supplied by industry j at time t and is defined as

  = ∑  if  ≠     ,

where   is the share of industry j’s output supplied to sector k according to the 2011 input-
output table from the GTAP database. Intra-industry inputs are not included, as horizontal effects
are already captured by the Horizontal variable.
The Forward variable is defined as the weighted share of output in upstream (or supplying)
sectors produced by GVC firms.

  = ∑  if  ≠    [[∑  for   ∈     ⁕   ] / [∑  for   ∈    ]],

where   is the proportion of total inputs purchased by industry j that are sourced from industry
m. Appendix Table 2 shows summary statistics for the horizontal, backward, and forward
variables by GTAP category.
Regression (1) in Table 6 estimates the relationship between the logarithm of firm markups
and our measures of horizontal, backward, and forward linkages with GVC firms. Because our
linkage variables are industry-specific, whereas the other variables are based on firm-level data,
not adjusting the standard errors is likely to result in a spurious finding of statistical significance
for the linkage variables, which are aggregated at the industry level (Javorcik, 2004; Moulton,
1990). Thus, following Javorcik (2004), we correct the standard errors by clustering them for
each industry in each year.
The coefficient of the horizontal variable is significantly negative at the one percent level,
implying that firms operating in industries with more GVC firms have lower markups relative
to firms operating in industries with fewer GVC firms, perhaps because GVC firms bring
competition to the sectors in which they operate. Not only do GVC firms have lower markups
 Global Value Chain Participation, Competition, and Markups: Evidence from Ethiopian Manufacturing Firms 511

but other firms in the same sector may also need to reduce their markups to compete with
the GVC firms that have lower markups.
The coefficient of the backward linkage variable is significantly negative at the one percent
level, implying that firms whose industries supply industries with a stronger presence of GVC firms
also have lower markups. A potential explanation is that GVC firms are larger and therefore have
greater bargaining power vis-a-vis suppliers relative to non-GVC firms (Lianos & Lombardi, 2016;

Table 6. Exploring the Relationship between Markups and Horizontal, Backward and Forward Linkages
(1) (2)
a
Participation dummies :
OnlyX -0.065 -0.028
(0.053) (0.052)
OnlyM -0.079*** -0.055**
(0.027) (0.024)
GVC -0.117** -0.099**
(0.049) (0.046)
Horizontal (GVC firms) -0.002*** -0.002***
(0.001) (0.001)
Backward (GVC firms) -0.021*** -0.023***
(0.006) (0.007)
Forward (GVC firms) 0.016 0.037
(0.058) (0.056)
Horizontal (State-owned firms) 0.005***
(0.001)
Backward (State-owned firms) 0.000
(0.011)
Forward (State-owned firms) -0.005
(0.012)
Ownership (= 1 if state, = 0 private) 0.118*** 0.088***
(0.020) (0.019)
Firm size (log of labor) 0.047*** 0.048***
(0.010) (0.010)
Log of capital -0.016*** -0.016***
(0.004) (0.004)
Constant 0.694*** 0.477***
(0.127) (0.142)
Number of observations 10322 10322
R2 0.080 0.111
Notes: The dependent variable is the log of markup, estimated by the ACF method. a Firms that have no trade (NoXM)
are omitted as a reference category.
Standard errors in parentheses have been corrected for clustering by industry and by year. *, **, and *** indicate
that the coefficients are significant at the 10, 5, and 1 percent level respectively.
512 Journal of Economic Integration Vol. 36, No. 3

World Economic Forum, 2018). The coefficient of the forward linkage variable is insignificant,
meaning that we identify no competition effect stemming from the presence of GVC firms
in industries supplying intermediate inputs.
Regression (2) includes horizontal, backward, and forward linkages for state-owned firms18),
as their presence may influence an industry’s competitive environment. The coefficient of the
horizontal variable is significantly positive at the one percent level, suggesting that firms
operating in industries with a greater presence of state-owned firms have larger markups. When
we include linkage variables that proxy for the presence of state-owned firms, the coefficients
of the GVC linkage variables remain essentially unchanged.
In sum, the regression results at the industry level are consistent with competition effects
resulting from the presence of GVC firms within an industry and through backward linkages.
The results also show that the coefficient of the GVC dummy remains significantly negative
even after controlling for all of the linkage variables, confirming the robustness of the negative
relationship between firms’ GVC participation and markups.

IV. Conclusion

Using panel data from the Ethiopian manufacturing census from 2000 to 2014, this study
explored whether GVC firms, defined as firms that both export and import inputs, have different
markups relative to non-traders (NoXM) and firms that only export (OnlyX) or import inputs
(OnlyM). It also investigated whether the depth of participation in trade affects firms’ markups
by interacting firms’ trade participation status and trade intensity. Moreover, this study explored
how industry-level competition resulting from the presence of GVC firms may influences
markups through horizontal, backward, and forward linkages.
The Ethiopian Manufacturing Census Data reveal that GVC firms are large in terms of both
sales and employment, have more capital-intensive production, are more productive based on
labor productivity controlling for capital intensity, and pay higher wages than one-way trading
or non-trading firms do, controlling for regional, year, and industry fixed effects. The finding
that GVC firms perform better than other firms is in line with the large body of previous
work showing that exports and access to imported inputs have positive effects on productivity
and product quality and that two-way trading firms are more productive on average than firms
that only import or export are. However, few studies have investigated the relationship between
GVC participation and markups in low-income countries. This study is meant to help fill this
gap by focusing on Ethiopia.

18) The presence of state-owned firms is measured by the share of total sales generated by state-owned firms in
each industry and year.
 Global Value Chain Participation, Competition, and Markups: Evidence from Ethiopian Manufacturing Firms 513

Using a pooled OLS model, we found that in Ethiopia, GVC firms tend to have lower
markups relative to non-trading firms and firms that are involved raw material imports only.
Similarly, we used fixed effects models to investigate within-firm variation and found that firms
tend to experience reductions in their markups relative to non-trading firms and firms that are
involved in only imports when they participate in GVCs. Additionally, we found that the more
intensely a firm is integrated into a GVC (measured as the share of export value added and
imported inputs in total sales), the lower its markup is. Furthermore, we explored competition
effects at the industry level and found that firms operating in industries with a greater presence
of GVC firms and suppliers selling inputs to such industries tend to have lower markups through
horizontal competition and backward linkages, respectively. All of these findings suggest that
GVC participation is associated with greater competition in Ethiopia.
Participation in GVCs appears to create both opportunities and challenges for Ethiopia. We
suggest that Ethiopian firms seize any windows of opportunity to participate in GVCs, as doing
so may allow them to tap into new markets, become more efficient through competition, and
benefit from technology transfers and information by exporting and importing. This experience
may provide opportunities for firms to move into higher-valued segments of GVCs. However,
the intense competition and low markups earned by GVC participants alarmingly imply that
any constraint or barrier at the margin may affect these firms’ survival. For instance, manufacturing
wages appear to have started rising in Ethiopia (Abreha et al., 2019), which may undermine
Ethiopia’s competitiveness relative to other low-wage countries. Because Ethiopian GVCs are
highly concentrated in a few labor-intensive industries, such as apparel, leather, and footwear
products, a concern is that without investment in skills and education, Ethiopia may be locked
in (World Bank, 2019a) to relatively low value-added and low-skill segments of production. As
wages rise, increasing Ethiopia’s capabilities (Gereffi, 2011) through investments in human capital
and infrastructure while exercising good policies to ensure a conducive investment climate may
determine whether Ethiopian firms will be able to both increase their competitiveness and
upgrade themselves to higher value segments of GVCs.

Acknowledgements

We would like to thank Hibret Belete Maemir and the Central Statistical Authority of Ethiopia
for the Ethiopian Manufacturing Census Data. We are also grateful to Ana Margarida Fernandes,
Aaditya Mattoo, and anonymous referees of this journal for comments and suggestions. We
accept full responsibility for any remaining errors.
514 Journal of Economic Integration Vol. 36, No. 3

References

Abreha, K. G. (2019). Importing and firm productivity in Ethiopian manufacturing. The World Bank
Economic Review, 33(3), 772-792.
Abreha, K., Jones, P., Lartey E., Mengistae, T., & Zeufack, A. (2019). Manufacturing job growth in
Africa: What is driving it and what will sustain it? The cases of Côte d’Ivoire and Ethiopia
(Background paper for the World Bank Africa Chief Economist Office). Industrialization for Jobs
Report, World Bank, Washington, DC.
Ackerberg, D., Caves, K., & Frazer, G. (2006). Structural identification of production functions. MPRA
Paper No. 38349.
Amiti, M., & Konings, J. (2007). Trade liberalization, intermediate inputs, and productivity: Evidence
from Indonesia. American Economic Review, 97(5), 1611-1638.
Antràs, P., Chor, D., Fally, T., & Hillberry, R. (2012). Measuring the upstreamness of production and
trade flows. American Economic Review, 102(3), 412-416.
Autor, D., Dorn, D., Katz, L. F., Patterson, C., & Van Reenen, J. (2017). Concentrating on the fall
of the labor share. American Economic Review, 107(5), 180-185.
Bas, M., & Strauss-Kahn, V. (2015). Input-trade liberalization, export prices and quality upgrading. Journal
of International Economics, 95(2), 250-262.
Békés, G., Hornok, C., & Muraközy, B. (2016). Globalization and the markups of European firms
(Discussion Paper 2044). Kiel Institute for the World Economy.
Bellone, F., Musso, P., Nesta, L., & Warzynski, F. (2016). International trade and firm-level markups
when location and quality matter. Journal of Economic Geography, 16(1), 67-91.
Bigsten, A., Collier, P., Dercon, S., Fafchamps, M., Gauthier, B., Willem Gunning, J., Oduro, A.,
Oostendorp, R., Pattillo, C., Söderbom, M., & Teal, F. (2004). Do African manufacturing firms learn
from exporting? Journal of Development Studies, 40(3), 115-141.
Bigsten, A., Gebreeyesus, M., & Söderbom, M. (2016). Tariffs and firm performance in Ethiopia. The
Journal of Development Studies, 52(7), 986-1001.
Brandt, L., Van Biesebroeck, J., Wang, L., & Zhang, Y. (2017). WTO accession and performance of
Chinese manufacturing firms. American Economic Review, 107(9), 2784-2820.
Castellani, D., Serti, F., & Tomasi, C. (2010). Firms in international trade: Importers' and exporters'
heterogeneity in Italian manufacturing industry. World Economy, 33, 424-457.
De Loecker, J., & Eeckhout J. (2017). The rise of market power and the macroeconomic implications
(Working Paper Number 23687). National Bureau of Economic Research, Cambridge, MA.
De Loecker, J., & Eeckhout, J. (2018). Global market power (Working Paper Number 24768). National
Bureau of Economic Research, Cambridge, MA.
De Loecker, J., Goldberg, P. K., Khandelwal, A. K., & Pavcnik, N. (2016). Prices, markups, and trade
reform. Econometrica, 84(2), 445-510.
De Loecker, J., & Warzynski, F. (2012). Markups and firm-level export status. American Economic Review,
102(6), 2437-2471.
Dewenter, K. L., & Malatesta, P. H. (2001). State-owned and privately-owned firms: An empirical analysis
 Global Value Chain Participation, Competition, and Markups: Evidence from Ethiopian Manufacturing Firms 515

of profitability, leverage, and labor intensity. American Economic Review, 91(1), 320-334.
Diez, M. F. J., Fan, J., Villegas-Sánchez, C. (2019). Global declining competition (Working Paper
WP/19/82). International Monetary Fund, Washington, DC.
Fatou, C., & Choi, J. (2013). Do firms learn by exporting or learn to export: Evidence from Senegalese
manufacturers’ plants (Working Paper Series 191). African Development Bank, Tunis, Tunisia.
Fernandes, A. M., Freund, C., & Pierola, M. (2016). Exporter behavior, country size and stage of
development: Evidence from the Exporter Dynamics Database. Journal of Development Economics,
119, 121-137.
Francois, J., Manchin, M., & Tomberger, P. (2014). Explanatory note: Export of value-added database.
The export value added database documentation. World Bank, Washington, DC. Retrieved from
https://datacatalog.worldbank.org/dataset/export-value-added-database
Freund, C., & Pierola, M. (2015). Export superstars. Review of Economics and Statistics, 97(5), 1023-1032.
Gereffi, G. (2011). Global value chains and international competition. The Antitrust Bulletin, 56(1), 37-56.
Goldberg, P. K., Khandelwal, A. K., Pavcnik, N., & Topalova, P. (2009). Trade liberalization and new
imported inputs. American Economic Review, 99(2), 494-500.
Gradzewicz, M., & Mućk, J. (2019). Globalization and the fall of markups. Warszawa, Poland: Narodowy
Bank Polski Education & Publishing Department.
Gullstrand, J., Olofsdotter, K., & Thede, S. (2014). Markups and export-pricing strategies. Review of
World Economics, 150(2), 221-239.
Gupta, N. (2005). Partial privatization and firm performance. The Journal of Finance, 6(2), 987-1015.
Halpern, L., Koren, M., & Szeidl, A. (2015). Imported inputs and productivity. American Economic Review,
105(12), 3660-3703.
Hornok, C., & Muraközy, B. (2019). Markups of exporters and importers: Evidence from Hungary. The
Scandinavian. Journal of Economics, 121(3), 1303-1333.
Hummels, D., Ishii, J., & Yi, K. (2001). The nature and growth of vertical specialization in world trade.
Journal of International Economics, 54(1), 75-96.
Javorcik, B. (2004). Does foreign direct investment increase the productivity of domestic firms? In search
of spillovers through backward linkages. American Economic Review, 94(3), 605-627.
Johnson, R. C., & Noguera, G. (2012). Accounting for intermediates: Production sharing and trade in
value added. Journal of International Economics, 86(2), 224-236.
Jones, P., Mengistae, T., & Zeufack, A. (2018). Selection, firm turnover, and productivity growth: Do
emerging cities speed up the process? (Working Paper No. 8291). World Bank, Washington, DC.
Kato, A. (2014). Does export yield productivity and markup premiums? Evidence from the Japanese
manufacturing industry (Discussion Paper Series 14-E-037). Research Institute of Economy, Trade,
and Industry.
Kee, H. L., & Tang, H. (2016). Domestic value added in exports: Theory and firm evidence from China.
American Economic Review, 106(6), 1402-1436.
Koopman, R., Wang Z., & Wei, S. J. (2014). Tracing value-added and double counting in gross exports.
American Economic Review, 104(2), 459-494.
Lianos, I., & Lombardi, C. (2016). Superior bargaining power and the global food value chain: The
Wuthering Heights of holistic competition law? (Research Paper Series 1/2016). Centre for Law,
516 Journal of Economic Integration Vol. 36, No. 3

Economics and Society.

Linden, G., Kenneth, K. L., & Dedrick, J. (2009). Who captures value in a global innovation network?
The case of Apple's iPod. Communications of the ACM, 52(3), 140-144.
Los, B., Timmer, M. P., & de Varies G. J. (2015). How global are global value chains? A new approach
to measure international fragmentation. Journal of Regional Science, 55(1), 66-92.
Megginson, W. L., & Netter J. M. (2001). From state to market: A survey of empirical studies on
privatization. Journal of Economic Literature, 39(2), 321-389.
Melitz, M. J., & Ottaviano, G. I. (2008). Market size, trade, and productivity. The Review of Economic
Studies, 75(1), 295-316.
Mengistae, T., & Pattillo, C. (2004). Export orientation and productivity in sub-Saharan Africa. IMF
Staff Papers, 51(2), 327-353.
Moulton, B. R. (1990). An illustration of a pitfall in estimating the effects of aggregate variables on
micro units. The Review of Economics and Statistics, 72(2), 334-338.
Oqubay, A. (2015). Made in Africa: Industrial policy in Ethiopia. New York: Oxford University Press.
Smeets, V., & Warzynski, F. (2013). Estimating productivity with multi-product firms, pricing
heterogeneity and the role of international trade. Journal of International Economics, 90(2), 237-244.
Taglioni, D., & Winkler, D. (2016). Making global value chains work for development. Washington,
DC: World Bank.
Timmer, M. (Ed.) (2012). The World Input-Output Database (WIOD): Contents, sources and methods.
WIOD Working Paper 10.
Timmer, M. P., Miroudot, S., & de Vries, G. J. (2019). Functional specialization in trade. Journal of
Economic Geography, 19(1), 1-30.
Topalova, P., & Khandelwal, A. (2011). Trade liberalization and firm productivity: The case of India.
Review of Economics and Statistics, 93(3), 995-1009.
Van Biesebroeck, J. (2005). Exporting raises productivity in sub-Saharan African manufacturing firms.
Journal of International Economics, 67(2), 373-391.
Vogel, A., & Wagner., J. (2010). Higher productivity in importing German manufacturing firms:
Self-selection, learning from importing, or both? Review of World Economics, 145(4), 641-665.
Wang, Z., Wei, S. J., & Zhu, K. (2013). Quantifying international production sharing at the bilateral
and sector levels (Working Paper Number w19677). Cambridge, MA: National Bureau of Economic
Research.
World Bank. (2019a). World development report 2020: Trading for development in the age of global
value chains. Washington, DC: World Bank.
World Bank. (2019b). World development indicators. Retrieved from https://databank.worldbank.org/report
s.aspx?source=world-development-indicators#
World Economic Forum. (2018). Global value chain policy series: Competition. Retrieved from https://ww
w.weforum.org/whitepapers/global-value-chain-policy-series-competition-ad360d8f-fcf6-4d91-89fe-27
453427bb14
 Global Value Chain Participation, Competition, and Markups: Evidence from Ethiopian Manufacturing Firms 517

Appendix

Table A1. Summary Statistics

Variable No. of obs. Mean SD Min Max

Log of markup 10324 0.51 0.42 -0.69 2.46
Participation dummies
GVC 10324 0.05 0.22 0.00 1.00
OnlyX 10324 0.01 0.10 0.00 1.00
OnlyM 10324 0.62 0.49 0.00 1.00
NoXM 10324 0.32 0.47 0.00 1.00
State ownership 10324 0.10 0.30 0.00 1.00
Firm size (log of labor) 10324 3.56 1.37 0.00 8.98
Log of capital 10324 13.63 2.78 -0.38 21.54
Upstream Index 10324 1.54 0.51 1.00 2.60
Trade intensity 10226 0.20 0.25 0.00 1.00
Notes: The figures reflect only the observations included in the regressions. The observations without two consecutive
years are dropped since the ACF method requires them.
(Source) Ethiopian Manufacturing Census Data 2000-2014.

Table A2. Presence of GVC and State-owned Firms in Horizontal, Backward and forward Linkages, by Industry

Presence of GVC firms (%) Presence of Stat-owned Firms (%)

GTAP Description Horizontal Backward Forward Horizontal Backward Forward
(GVC) (GVC) (GVC) (State) (State) (State)
b_t Beverages & tobacco 36.510 0.000 0.060 33.600 0.000 0.835
Crp Chemical rubber products 3.023 6.836 0.177 23.796 3.963 0.564
Ele Electronic equipment 0.429 0.046 0.477 0.000 0.671 1.140
Fmp Fabricated metal 1.320 0.280 0.034 22.184 1.904 1.212
i_s Iron & steel 2.220 0.022 0.099 26.061 0.397 0.491
Lea Leather 73.323 0.175 0.249 21.303 0.100 0.959
Lum Lumber 0.800 0.072 0.263 20.041 0.298 0.933
Mil Dairy products 3.039 0.029 0.042 24.046 0.123 0.748
Mvh Motor vehicles and parts 0.000 0.000 0.090 12.071 0.001 0.960
Nfm Non-ferrous metals 0.000 0.488 0.134 2.222 5.783 0.887
Nmm Non-metallic minerals 1.258 0.209 0.045 52.259 0.294 0.317
Ofd Other food 6.359 0.008 0.144 27.501 0.008 0.754
Ome Other machinery 0.203 0.002 0.047 0.230 0.006 0.567
Omf Other manufacturing 0.000 0.005 0.297 0.000 0.015 1.220
Omt Other meat (pig meat & offal) 9.080 5.099 0.084 32.689 4.828 0.347
Ppp Paper & paper products 0.290 1.972 0.164 44.927 2.773 1.321
Sgr Sugar 55.560 2.108 0.116 97.973 2.312 1.155
Tex Textiles 41.877 0.505 0.289 37.168 0.303 1.273
Wap Wearing apparel 31.557 0.001 0.226 17.777 0.000 0.952
Total 14.045 0.940 0.160 26.097 1.251 0.875

(Source) Ethiopian Manufacturing Census Data 2000-2014; Global Trade Analysis project (GTAP) database (Version 9).