LABOR AND FINANCIAL MARKET FRICTIONS
IN DEVELOPING ECONOMIES
by
HOA Q DUONG
A DISSERTATION
Presented to the Department of Economics
and the Graduate School of the University of Oregon
in partial fulfillment of the requirements
for the degree of
Doctor of Philosophy
June 2020
DISSERTATION APPROVAL PAGE
Student: Hoa Q Duong
Title: Labor and Financial Market Frictions in Developing Economies
This dissertation has been accepted and approved in partial fulfillment of the
requirements for the Doctor of Philosophy degree in the Department of Economics
by:
Shankha Chakraborty Co-chair
Anca Cristea Co-chair
Bruce Blonigen Core Member
Jiao Zhang Institutional Representative
and
Kate Mondloch Interim Vice Provost and Dean of the
Graduate School
Original approval signatures are on file with the University of Oregon Graduate
School.
Degree awarded June 2020
ii
©c 2020 Hoa Q Duong
iii
DISSERTATION ABSTRACT
Hoa Q Duong
Doctor of Philosophy
Department of Economics
June 2020
Title: Labor and Financial Market Frictions in Developing Economies
This dissertation investigates the implications of frictions in labor and
financial markets, with a focus on developing economies. Through theoretical and
empirical analyses, I first analyze the effects of a large informal labor sector on
labor policies and outcomes. I then provide evidence on the effects of access to
credit on export performance measured along the extensive and intensive margins
of trade. Lastly, I study the implications of various finance measures at the
country level on export quality upgrading. The results highlight the importance
of policies that promote financial soundness at the macroeconomic level.
iv
CURRICULUM VITAE
NAME OF AUTHOR: Hoa Q Duong
GRADUATE AND UNDERGRADUATE SCHOOLS ATTENDED:
University of Oregon, Eugene, OR, USA
Brandeis University, Waltham, MA, USA
National Economics University, Hanoi, Vietnam
DEGREES AWARDED:
Doctor of Philosophy, Economics, 2020, University of Oregon
Master of Art, International Economics, 2015, Brandeis University
Bachelor of Economics, 2012, National Economics University
AREAS OF SPECIAL INTEREST:
Applied Macroeconomics
International Trade
Development Economics
Financial Economics
GRANTS, AWARDS AND HONORS:
Graduate Teaching Fellowship, 2015-2020, University of Oregon
v
ACKNOWLEDGEMENTS
I want to express my sincere gratitude to all the faculty and staff of the
Department of Economics for their tremendous support. I am extending my
heartfelt thanks to professors Shankha Chakraborty and Anca Cristea. They have
provided me with invaluable insights, endless encouragement, and unmatched
kindness. They are the exemplars of teaching, researching, and advising. My
special thanks go to professors Jeremy Piger, Bruce Blonigen, Michael Kuhn, Jiao
Zhang, and our department coordinator Sharon Kaplan. It was a great honor and
joy to work with them. Lastly, I am incredibly grateful to my family, partner, and
friends for their unshakable faith in my abilities.
vi
To you
vii
TABLE OF CONTENTS
Chapter Page
I. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
II. VOLUNTARY INFORMALITY IN AN EQUILIBRIUM SEARCH
MODEL WITH HETEROGENEOUS LABOR . . . . . . . . . . . . . . 5
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Calibrating Model Parameters . . . . . . . . . . . . . . . . . . . . 33
Equilibrium Analysis: Comparative Statics . . . . . . . . . . . . . 40
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
III. FINANCIAL DEVELOPMENT, EXPORT PERFORMANCE,
AND FIRM PRODUCTIVITY: EVIDENCE FROM
CROSS-COUNTRY DATA . . . . . . . . . . . . . . . . . . . . . . . . . 56
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Robustness Check . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
viii
Chapter Page
IV. FINANCE AND EXPORT QUALITY UPGRADING . . . . . . . . . . 111
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Related Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
V. CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
APPENDICES
A.. COMPARATIVE STATICS GRAPHS FOR CHAPTER II . . . . . . 148
B.. SUPPLEMENTAL MATERIALS FOR CHAPTER III . . . . . . . . 153
A Model of Credit Constraints . . . . . . . . . . . . . . . . . . . . 153
Data Exploration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
C.. SUPPLEMENTAL MATERIALS FOR CHAPTER IV . . . . . . . . 167
REFERENCES CITED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
ix
LIST OF FIGURES
Figure Page
1. Flows of High-skilled Workers . . . . . . . . . . . . . . . . . . . . . . . 15
2. Determination of θ∗, w∗ ∗HF , and uh . . . . . . . . . . . . . . . . . . . . . 32
3. General Equilibrium Effects of a Budget Neutral Policy . . . . . . . . . 51
4. Private Credit and Income Per Capita . . . . . . . . . . . . . . . . . . 81
5. Cross-sectional Distribution of Private Credit . . . . . . . . . . . . . . 83
6. Temporal Changes in Private Credit across Income Groups . . . . . . . 84
7. Two Measures of Financial Vulnerability . . . . . . . . . . . . . . . . . 86
8. Firm Size by Export Value . . . . . . . . . . . . . . . . . . . . . . . . . 87
9. Private Credit and Quality Growth . . . . . . . . . . . . . . . . . . . . 123
10. Cross-sectional Private Credit and Quality Growth . . . . . . . . . . . 124
11. Initial Quality and Quality Growth . . . . . . . . . . . . . . . . . . . . 124
12. Cross-section of Initial Private Credit and Initial Quality . . . . . . . . 126
x
LIST OF TABLES
Table Page
1. Parameter Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2. Steady State Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . 39
3. ISIC Revision 3 2-digit divisions . . . . . . . . . . . . . . . . . . . . . . 88
4. Effects of Private Credit on Total Export Sales . . . . . . . . . . . . . 105
5. Effects of Private Credit on Extensive Margin or Firm Counts . . . . . 106
6. Effects of Private Credit on Intensive Margin
or Average Sales per Firm . . . . . . . . . . . . . . . . . . . . . . . . . 107
7. Industries Most and Least Benefiting from Financial Development . . . 108
8. Marginal Effects of Private Credit by Country . . . . . . . . . . . . . . 109
9. Nested Regression: Marginal Effects of Private Credit . . . . . . . . . . 110
10. Finance Measures and Quality Upgrading . . . . . . . . . . . . . . . . 139
11. Private Credit and Quality Upgrading by Income Group: Baseline . . . 140
12. FDI and Quality Upgrading by Income Group: Baseline . . . . . . . . 141
13. Financial Reform and Quality Upgrading
by Income Group: Baseline . . . . . . . . . . . . . . . . . . . . . . . . 141
14. Private Credit and Quality Upgrading by Income Group . . . . . . . . 142
15. FDI Net Inflows and Quality Upgrading by Income Group . . . . . . . 143
16. Financial Reform and Quality Upgrading by Income Group . . . . . . . 144
xi
CHAPTER I
INTRODUCTION
This dissertation examines the implications of labor and financial frictions in
developing countries. Financial markets in developing countries are characterized
by a lack of credit provision and distorted credit allocation. Meanwhile, inefficient
labor regulations give rise to a typically large informal labor sector. Competition
from the informal sector and access to finance are among the top three challenges
faced by small and medium-sized firms in developing countries. According to the
World Bank’s 2018 Enterprises Survey (WBES), 15% of the enterprises consider
access to finance the biggest obstacle to their business, and 12% find practices
from informal firms their biggest threat. A better understanding of these two
types of frictions will help policymakers design the appropriate policies to promote
growth in developing countries.
The informal labor market refers to the labor market that may be legal but
not subject to government control. Because of its nature, there are not much direct
data on informal labor. However, we know that the informal labor market is large
and relevant based on firm and household surveys. In the WBES 2018, nearly 88%
of establishments started unregistered, and up to 50% acknowledge competition
from informal firms. In Latin American countries where the informal sectors are
unusually large, household surveys suggest that workers deliberately switch from
1
formal employment to informal employment if they want to become business
owners to avoid tax payments (Maloney, 2004). Based on these observations, I
set up an equilibrium search model with both sectors co-existing in Chapter II.
The model allows high-skilled workers from the formal labor sector transit to the
informal labor sector and become entrepreneurs. It predicts that activities in the
informal labor sector could create misleading signals about the efficiency of the
formal labor sector. While the government can use policies to reduce informal
employment, such policies have critical drawbacks that outweigh their benefits.
Instead, the government should work towards improving labor skills as a long term
strategy to diminish the informal sector.
Chapters III and IV study the importance of finance for exporting firms.
Compared to domestic firms, exporting firms are often larger and more productive,
but also have greater financial needs to accommodate extra costs and uncertainties
associated with global transactions. Even for multinational firms which tend
to be less resource constrained, financing in the domestic markets is among
the top ten factors affecting their investment decisions, according to the 2018
Global Investment Competitiveness Report of the World Bank. The main
financial measure employed in these two chapters is private credit to GDP ratio,
a commonly used indicator in the literature on finance and economic development.
A larger ratio implies a less frictional - or more developed credit market. As bank
2
loans are the major source of external funding to firms, especially in developing
countries, this indicator is relevant to the dissertation theme.
Chapter III quantifies the impact of financial development along two margins
of exports: the extensive margin (measured by firm counts) and the intensive
margin (measured by average sales per firm). Using a detailed industry-level
dataset from the World Bank, the chapter shows that financial development has
differential effects on export performance, after taking into account heterogeneity
across countries, industries, and firm types. Overall, financial development has
a statistically and economically significant impact: for example, a one standard
deviation increase in financial development in Cambodia is associated with
an increase in total exports by all firms of 58.4%. Almost half of this increase
comes from the rise in the number of exporting firms. Financial development
has a more substantial effect on exports of the lowest income country group, and
exports of certain industries such as Pharmaceuticals and Medical appliances.
Continuing exporting firms stand out from the entrant as their exports react more
strongly to finance development in both margins. For any firm types, the extensive
margin is more sensitive to financial development, implying fixed cost financing
is more financially challenging to exporting firms. Fixed cost financing is more
important to first-time exporters, as opposed to continuing exporters who may
have more internal liquidity. These results are robust to the control for state-
3
owned enterprises, of which presence can bias downward the effect of financial
development on exports.
In Chapter IV, I analyze the effect of finance on export quality upgrading,
using a dataset from the International Monetary Fund. Export quality upgrading
measures how fast the quality of a product from a given country moves towards
the world frontier. Finance has been credited in the literature on economic
convergence as an engine for innovation-led growth. Its role as a source of
comparative advantage has also been studied extensively in the international trade
literature. Guided by these findings, I examine whether finance influences export
quality upgrading as it does economic growth and export margins. I use three
country-level finance indicators: private credit to GDP ratio, FDI net inflows to
GDP, and an index of financial reform. I find that finance does promote quality
upgrading, and more strongly for products with low level of quality. The effects
vary by country groups: financial reform can explain more of quality upgrading
in the highest and lowest income country groups compared to the middle income
group. FDI, in contrast, seems more important to middle income countries than
the groups at extreme. Further analysis indicates that private credit promotes
quality upgrading more strongly in sectors that require relatively more external
finance or have relatively little tangible assets. More research is needed to isolate
the effects of country-level heterogeneity (e.g., human capital, institutional quality)
and product-level heterogeneity(e.g., product differentiation).
4
CHAPTER II
VOLUNTARY INFORMALITY IN AN EQUILIBRIUM SEARCH MODEL
WITH HETEROGENEOUS LABOR
Introduction
The term “informal sector” refers to economic activities that are not reported
to governments. In the developing world, the informal sector can be as large as
the formal sector. Between 2010 and 2014, the proportion of informal employment
as a percent of non-agricultural employment was 49% in Northern Africa, 74%
in Sub-Saharan Africa, 57% in Latin America, and 57% in South Asia (Charmes,
2016). According to the International Labour Organization, the informal sector is
not directly influenced by governmental policies, and is often seen as a hindrance
to economic development. Informal employment is quick and easy, yet it offers
low and unstable income, besides limited accessibility to social security and
professional training.
Many schools of thought have emerged to explain the existence of the
informal sector. The “dualist school” views the informal sector as an inferior
market segment with no direct links to the formal sector and expects the
informal sector to diminish with economic growth (Harris and Todaro, 1970).
The “structuralists” suggest that the informal sector consists of micro-firms and
workers that supply cheap inputs to larger capitalist firms. Since the two sectors
5
are tied in a production chain, the informal sector is unlikely to be eliminated
by growth (Castells and Portes, 1989). Another major school of thought, the
“legalists”, defines the informal sector as a collection of micro-entrepreneurs who
choose to operate informally to avoid registration cost and tax obligations (de
Soto, 1989).
Empirical evidence shows that each theory can explain some but not all
features of the informal sector. The multi-segmented theory thus arises as a
combination of the three theories above (Chen, 2006; Fields, 2005). It stipulates
that the informal sector is formed of sub-segments, among which the more
advanced tier associates with more structured and capital-intensive production
types. The lower-tier segment is home to subsistence activities, as suggested
by the dualists. The upper-tier segment contains micro-entrepreneurs that
either collaborate or compete with formal firms, as suggested by the legalists
and structuralists. Workers of various productivity levels commingle and follow
different channels to circulate across segments and sectors.
This research, in a similar vein, aims to capture the diverse nature of the
informal sector by letting this sector employ both low-skilled and high-skilled
workers. What makes this research stand out from previous studies is the specific
way in which high-skilled workers enter informal employment and the relationship
between the two types of workers. High-skilled workers are the only type to be
employed in the formal sector. They can choose to leave formal jobs to become
6
entrepreneurs and hire low-skilled workers in the informal sector. The informal
sector is therefore indispensable in this economy: it accommodates part of the
labor force that cannot find jobs in the formal sector but also competes with the
formal sector for high-quality labor. To the best of my knowledge, this research is
the first to model such an informality structure.
The stance that this research takes on informality is a blend of the theories
discussed previously. The theoretical part in this research speaks to a body of
research that builds upon the equilibrium search framework of Diamond (1982),
Mortensen and Pissarides (1994). In the second half of the chapter, a calibration
analysis is conducted to match the theoretical model with informality facts in
developing countries that are taken from Fields (1990) and Maloney (2004). In
the calibrated economy, 7% of the high-skilled are unemployed, 36% working
informally, and 57% working formally. Comparative analysis suggests that
higher worker productivity can reduce informal employment while more generous
unemployment benefit can encourage informality.
Using the equilibrium search model to address informality is not novel, two
examples being Zenou (2008) and Ulyssea (2010). The former takes the dualists’
view: workers who are neither formally hired nor unemployed are assumed to
be working informally. The informal sector pays competitive wage, and is not
frictional. Such feature marks the difference between two papers: Ulyssea (2010)
models an informal sector almost identical to the formal one. Both sectors are
7
frictional and produce intermediate goods that become inputs for a common final
good. Ulyssea (2010) illustrates the legalists’ idea: firms operate in the informal
sector to avoid regulation costs.
This research also leans on the strand of literature where equilibrium search
models combine with heterogeneity and on-the-job search. Albrecht, Navarro, and
Vroman (2009) extends Zenou (2008) by introducing ex-ante worker heterogeneity,
and explicit labor transitions between unemployment and the informal sector.
Gautier (2002) models both worker heterogeneity and job heterogeneity, though
not addressing informality per se. In Gautier (2002), firms can open either simple
jobs that accept both worker types or complex jobs that only accept high-skilled
workers. A high-skilled individual can search for a complex job while undertaking
a simple occupation.
Voluntary informality under the form of entrepreneurship has been observed
in developing countries. A survey by Fields (1990) notes that many people in
Malaysia and Costa Rica left the formal sector to work in the informal sector by
choice. Exploiting the detailed database on worker transitions in Argentina, Brazil,
and Mexico, Maloney (2004) suggests that the informal sector in these countries
should be seen as an unregulated micro-entrepreneurial sector. Approximately
35% of informal self-employed workers in Mexico leaving their previous jobs for the
reason of “being independent”. Such findings challenge the more traditional view
8
that the informal sector is where high-skilled individuals work temporarily before
being able to get a formal job.
In this research, high-skilled workers revolve from unemployment to formal
employment, then to informal employment. Such moves are not unrealistic.
Maloney (2004) reports that 75% of the unemployed in Mexico and 64% in
Argentina were informal workers previously. Fields (1990) argues that the formal
sector provides opportunities for workers to accumulate savings to start up their
own business, and the formal sector may prefer to employ workers who have
training in the informal sector. This tends to be true for better-educated workers,
for example, managers of appliance stores who previously worked in small family
businesses. The flow of high-skilled workers across employment states is restricted
to be one-way for simplification.
Putting the low-skilled workers in a symbiotic relationship with the high-
skilled opens up new questions on the implications of worker heterogeneity. A
heterogeneous informal sector reflects the “internal dualism” notion of Fields
(1990). Compared with other studies (such as Albretch et al., 2009) where cross-
matching of skill and job types are allowed, workers of different productivity levels
are rivals as they compete for the same job type. Such studies examine worker
heterogeneity as an input for firms’ decision on which types of jobs to open. This
research, on the other hand, will explore the influence of worker heterogeneity on
formal and informal market outcomes.
9
The chapter finalizes with an analysis of policies toward informality.
Governments, for practical reasons, might want to exercise stronger regulatory
control toward the informal sector. Informal activities and illegal activities
sometimes conflate, and informal workers generally lack benefits and protections
(Chen, 2012). Given the specific roles of the informal sector in this model, I
will formulate a policy and test the effectiveness of a tax hike. Throughout the
analysis, the size of the informal sector will be evaluated by both employment and
income.
The remainder of this chapter is structured as follows: Section 2 provides
the basic structure of the model, the definition of a steady state equilibrium and
condition for uniqueness. Section 3 discusses the details of a calibration exercise.
Section 5 presents results from a comparative statics analysis, including an effect of
a budget neutral policy. Section 4 concludes.
The Model
Agents
The economy is populated by a continuum of infinitely lived workers of mass
L = 1 and a continuum of firms of undetermined mass. Workers and firms are
subjected to a common exogenous real interest rate r. Use LH and LL to indicate
the number of high-skilled and low-skilled labor force, so L = LH + LI . The
exogenous ratio of high-skilled workers in the labor force is φ, then LH = φL and
10
LL = (1− φ)L. Let LH = EH + IH + UH where EH , IH , and UH correspond to the
number of high-skilled workers who are employed in the formal sector, employed in
the informal sector, and unemployed. I use lower case characters to denote labor
ratios with respect to LH : eh + ih +uh = 1. The productivity of each worker type is
a constant, denoting y for high-skilled type and z for low-skilled type (0 z  y).
Matching function
The frictional nature of the formal labor market is embodied in the matching
function M(UH , V ). This function returns the number of matches M per unit of
time from two inputs: the number of vacancies opened by firms in the formal
sector V and the number of unemployed high-skilled workers UH . The matching
function is increasing in both arguments (MU > 0,MV > 0), concave, andH
constant return to scale. Given the last property, the matching function can
transform into functions of a single input as follows
( ) ( )
≡ M(V, UH) UH 1q(θ) = M 1, = M 1, (2.1)
V V θ
≡ M(V, UH) M(V, UH)/Vp(θ) = = θq(θ) (2.2)
UH UH/V
where θ ≡ V/UH is the indicator of “market tightness” from firms’ viewpoint.
q(θ) is the vacancy filling rate and p(θ) is the job finding rate (q′(θ) < 0,
p′(θ) > 0). These functions satisfy Inada conditions: limθ→∞ q(θ) = 0, limθ→0 q(θ) =
11
∞, limθ→∞ p(θ) = 0, limθ→0 p(θ) = ∞. The elasticity of q(θ) is a number
between −1 and 0. Its absolute value is denoted η(θ). The elasticity of p(θ) is then
1− η(θ) ≥ 0.
Here q(θ) and p(θ) are Poisson intensities or Poisson rates that indicate
the average number of matches arriving to a firm and to a worker per unit time,
respectively. Particularly, in a period of ∆t the average number of matches realized
is q(θ)∆t to firms and p(θ)∆t to workers. Put differently, for firms, the flow of
matches is described by a Poisson process with parameter λ = q(θ)∆t. For
workers the process is Poisson λ = p(θ)∆t. Such difference highlights the frictional
characteristic of the formal labor market.
The flow of matches can be described by a Poisson process because matches,
or job/worker pairs, are counted in whole numbers. Assuming a Poisson process
implies these conditions hold true: i. each match occurs independently from each
other, ii. the time interval can be divided into non-overlap small enough sub-
intervals, where the chance that a match is formed in each sub-interval is constant
and proportional to the sub-interval’s length, and iii. two matches cannot form at
the same instant.
By definition of Poisson rates and Poisson parameters, the mean duration
of a vacant job is 1/q(θ), and the mean duration of being unemployed is 1/p(θ),
measured as a fraction of the time interval ∆t. When there are more vacancies
relative to unemployed workers firms find workers more slowly while workers get
12
hired more quickly. The opposite happens when there are more unemployed labor
relative to vacancies.
Another result stemming from the constant return to scale property of the
matching function is min{UH , V } ≤ M(UH , V ) ≤ max{UH , V }. This result
implies that the labor market will never move “too fast” or “too slow” to both
parties. When UH < M(UH , V ) < V , the unemployed are better off as a whole
since there is at least one worker who receives more than one job offer. When
V < M(UH , V ) < UH , firms can easily choose one unemployed worker to fill
vacancies and exit, while workers have to compete for getting matched.
Parameters
In the formal sector, a firm must pay a sunk cost κ ≥ 0 to enter the market.
The firm opens a vacant job and wait until it is filled. A filled job or a match will
generate gross output y, which is also the productivity of high-skilled individuals.
An ongoing match can be destroyed if it is hit by an adverse productivity
shock. The shock arrives at rate δF > 0, and is severe enough to make the
match no longer profitable to firms. After the match is dissolved, firms will either
withdraw from the market or re-enter the market with a vacancy. When the match
is destroyed by a negative productivity shock, the worker becomes unemployed.
A current match will also dissolve if the employee by choice breaks the
contract to join the informal sector. The informal sector is not frictional like the
formal sector; therefore, the workers need not spend time and effort on finding a
13
business opportunity. The average number of high-skilled workers starting informal
ventures is α > 0 per unit time. The mechanism of informal job arrival is left
undiscussed here. However, it can be justified by the “very reduced form”, “no
micro foundation” spirit of the equilibrium search framework.
High-skilled workers in the formal sector receive wHF if employed and b if
unemployed. Formal wage wHF is determined endogenously by a Nash bargaining
scheme. The equilibrium wage depends on the bargaining power of workers,
denoted β ∈ [0, 1). Unemployment income b is a fixed parameter. I assume high-
skilled workers earn more from working in any sector than from being unemployed
b < min{wHF , wHI}.
In the informal sector, high-skilled individuals earn wHI and pay low-skilled
workers wl. Both values are endogenous. Wage wHI is defined as the maximum
profit attained by the informal micro-enterprises. Wage wl is the marginal product
of low-skilled labor. The informal labor market is perfectly competitive, and
given that the supply of low-skilled workers is perfectly inelastic, there is no
unemployment among low-skilled labor.
The informal enterprises close at rate δI - a Poisson intensity measured at
the same temporal unit with α and δF . Unlike α and δF , δI is not a constant. The
destruction rate of informal business is rather a decreasing function of wHI (see
Calibration section). Intuitively, the informal owner is more likely to quit and
14
return to unemployment when business profits decline. Lastly, I assume that the
processes of job destruction and job arrival in two sectors are independent.
Transition across states
Following Albrecht et al. (2002) and Gautier (2002): in steady state, for each
worker type, the inflows and outflows of each employment state must balance.
In other words, we are looking at the long-run equilibrium where the allocation
of high-skilled workers into formal employment, informal employment, and
unemployment is constant.
FIGURE 1. Flows of High-skilled Workers
i. High-skilled workers and formal employment
The outflow from this pool includes high-skilled individuals who either lose
formal jobs due to adverse shocks or leave for the informal sector by choice EHδF +
EHα. The inflow is made of unemployed high-skilled individuals who find jobs in
the formal sector UHp(θ). The change in the pool of employed high-skilled worker
15
per unit time, written in labor ratios, is ėh = ehδF + ehα − uhp(θ). In steady state
ėh = 0 or
ehδF + ehα = uhp(θ) (2.3)
ii. High-skilled workers and informal employment
The pool of informal high-skill workers expands by EHα and reduces by IHδI
per unit time. The net flow is ėh = ehα− ihδI . In steady state ėh = 0 or
ehα = ihδI (2.4)
iii. High-skilled workers and unemployment
The inflow to this pool consists of high-skilled workers who get displaced
from informal and formal occupations EHδF + IHδI . The outflow contains high-
skilled workers who find formal jobs UHp(θ). In steady state, u̇h = 0 or
ehδF + ihδI = uhp(θ) (2.5)
Value Functions of Formal Firms and High-skilled Labor
Value functions are the present-discounted values of income or profit
accumulated by agents during the infinite horizon. Agents are assumed to hold
an asset of which value varies by the state of agents in the labor market. For the
16
following equations, the left-hand side can be interpreted as the flow of capital cost
and the right-hand side as the return of capital. The two flows equate each other
since we assume a perfect capital market.
i. Formal firms: Value of a matched job VJ for a firm is the sum of the
net output produced by the match and the potential loss if the match dissolves.
Since adverse productivity shock and informal job formation are two independent
Poisson process, their rates can be combined (term δF + α) to indicate the number
of job-worker pairs separated per unit time. This setup implies that the informal
sector competes against the formal sector for high-quality labor
rVJ = y − wHF − (δF + α)(VJ − VV ) (2.6)
Value of a vacant job VV to a firm is merely the expected capital gain when
the vacancy is filled. The firm does not pay a cost to maintain the vacancy per
unit time, so the value of a vacancy is only the product of a filled job’s premium
and the mean arrival rate of a match per unit time
rVV = q(θ)(VJ − VV ) (2.7)
ii. High-skilled worker: The employment status of a high-skilled worker is
comparable to an asset with three states. An asset in the formal employment state
can switch to two other states: informal employment and unemployment. Hence,
17
the return on working formally VE has three components: wage wHF , the expected
capital loss from being unemployed, and the expected value change if the worker
moves to the informal sector. This setup of VE (and VJ) is based on studies with
on-the-job search modeling (e.g., Gautier, 2002).
An asset in the informal employment or unemployment state can switch to
only one other state. The return on an informal job VI is the sum of the current
yield wHI and the expected capital loss from being unemployed. Value of an
asset in the unemployment state VI is the unemployment income b plus with the
expected capital gain if the worker finds a formal job.
rVE = wHF − δF (VE − VU) + α(VI − VE) (2.8)
rVI = wHI − δI(VI − VU) (2.9)
rVU = b+ p(θ)(VE − VU) (2.10)
The term wHI refers to the salary that high-skilled workers earn from
entrepreneurship in the informal sector.
Job creation relation
A firm must pay a sunk cost (or creation cost) κ to enter the market and
open a vacancy. We can think of this sunk cost as firms buying machinery and
equipment before hiring labor (Acemoglu, 2001). A vacancy remains open until
the firm finds a suitable worker and turns the vacancy into a filled job, which is
18
more profitable for the firm. The vacancy has a non-negative value as described in
Equation 2.7.
Profit maximization dictates that firms would continue to post vacancies
until the value of vacancies is driven down to the value of the sunk cost: VV = κ.
At that time, the firm can no longer extract rent from the vacancy. Substitute this
value of VV into the value functions of a filled and a vacant job for a formal firm
(Equation 2.6 and Equation 2.7) and rearrange, then the value of a filled job can
be expressed in two ways:
y − wHF + κ(δF + α)
VJ = (2.11)
( r + δF)+ α
r
VJ = κ 1 + (2.12)
q(θ)
In Equation 2.11, the denominator is the effective discount rate of a filled
job, which is the sum of the real interest rate and the risks of match destruction.
The term κ(δF + α) is the amount “saved” by firms when a match remains intact.
In Equation 2.12, the value of a filled job balances with the paid upfront cost κ
and the expense incurred during the life of the vacancy 1/q(θ).
As it happens, the sunk cost is not indeed a one-time cost. After entering
the market, firms continue to pay a fraction of the sunk cost (term κr) to maintain
vacancies. The presence of a sunk cost not only creates an extra “flow cost” but
also opens up an additional channel via which the real interest rate can affect
19
a match’s value. In the baseline model where sunk cost is absent (so VV = 0)
and vacancies are maintained by a fixed cost (similar to the “flow cost”, so in
equilibrium VJ = VV 6= 0), this channel is turned off.
In equilibrium, VJ is constant. Hence, two expressions of VJ must equal.
Combine them we have the zero-profit or job-creation condition
( )
r + δF + α
y − wHF − κr 1 + = 0 (2.13)
q(θ)
The job-creation condition illustrates a negative relationship between formal
wage and market tightness. A higher market tightness extends the duration
of the vacancy, increasing hiring cost. Firms must offer a lower wage to retain
profits. The job-creation condition is comparable to the labor demand function
in Walrasian economics (Pissarides, 2000).
The term α represents the influence of the informal sector. This term, along
with the arrival rate of adverse shock δF and real interest rate r, determines
the slope of the job-creation curve. An increase in this term can make firms’
willingness to create jobs become more sensitive to changes in market tightness.
Wage setting relation
A formed match benefits both workers and firms. The total surplus of a
match is given by S = (VE − VU) + (VJ − VV ). This expression already accounts for
the value of an informal job to workers (i.e., we need VI to derive VE − VU). Nash
20
bargaining sets a wage rule that splits S linearly between the firm and the worker.
The share of each party depends on the worker’s bargaining power β. Here β is
non-negative and strictly less than one otherwise firms have no incentive to open
vacancies.
Wage in the formal sector solves
wHF = arg max(V
β 1−β
E − VU) (VJ − VV )
wHF
which can be simplified into
(1− β)(VE − VU) = β(VJ − VV ).
The right-hand side term represents the share of the firm in the total match
surplus S. This term is derived by subtracting the value function of a vacant job
(Equation 2.7) from the value function of a filled job (Equation 2.6) and rearrange
− y − wHFVJ VV = (2.14)
r + α + δF + q(θ)
The left-hand side term is the share of the worker in the total surplus S.
This term is derived by combining three steady state values of the worker: being
employed VE (Equation 2.8), unemployed VU (Equation 2.10), and informally
21
employed VI (Equation 2.9)
− (r + δI)(wHF − b) + α(wHI − b)VE VU = (2.15)
(r + δI)(r + α + p(θ) + δF ) + αp(θ)
Nash Bargaining implies a wage rule
( )
− (r + δI)(wHF − b) + α(wHI − b) y − wHF(1 β) = β
(r + δI)(r + α + p(θ) + δF ) + αp(θ) r + α + δF + q(θ)
The job-creation condition (Equation 2.13) gives
κr(r + δF + α)
q(θ) =
y − wHF − κr
Use this expression to substitute q(θ) and p(θ) out of the wage rule and rearrange
to get
[ ( )] [ ]
− α αwHF = β (y κr) + θκr + 1 + (1− β) b− (wHI − b) (2.16)
r + δI r + δI
This is the wage-setting curve, an equivalence of the labor supply
curve in the Walrasian market. The wage-setting curve portrays a positive
relationship between formal wage and market tightness, conditional on some
parameter restrictions. Consistent with the baseline model, formal wage increases
proportionally with workers’ productivity and unemployment benefit.
22
Workers can claim a better wage at a higher market tightness essentially
because of the term θκr. This term is the average flow cost per unemployed worker
θκr = vκr/u (Pissarides, 2000). As pointed out in the previous section, this is the
expense firms pay to maintain an unfilled job. Once the job is filled, the expense
stops incurring, and firms reward workers for saving firms such expense.
Being formally employed also gives the worker access to the informal sector.
The worker might break the contract at some point, and firms respond to that
risk by taking away part of the worker’s compensation as a penalty. The amount
deducted from the compensation is proportional to the worker’s informal income,
i.e. term (1 − β)wHI . All else being equal, a higher bargaining power raises the
reward portion while reduces the penalty portion.
The wage-setting condition (Equation 2.16) can be rewritten as
( )
α
wHF = β(y − κr) + βθκr + (1− β)b+ βθκr + (1− β)b− (1− β)wHI
r + δI
Without the presence of the informal sector, the term α/(r + δI) would not
exist. This term accentuates the difference between formal benefits (rewards and
unemployment insurance) and informal income. The first three terms are standard
for a single sector model.
23
Informal sector
In the informal sector, high-skilled workers are voluntary participants.
They can choose to leave the formal sector to become informal business owners.
Low-skilled workers are not as much privileged. They have no other choice than
working for the high-skilled owners.
All informal enterprises adopt an identical technology that combines the
productivity of the high-skilled owner and the collective productivity of low-skilled
labor. Each enterprise hires li number of low-skilled workers, each of whom has
productivity of z. The owner’s productivity y can be interpreted as entrepreneurial
skill in the technology f(li) = Ay(zli)
ρ. The technology exhibits diminishing
returns in low-skilled workers’ productivity z, i.e. the marginal product with
respect to low-skilled labor ρ ∈ (0, 1). The good produced by these micro-
enterprises is sold on competitive markets with price normalized to one.
In the informal sector, the entrepreneur i maximizes
πi = f(li)− wlli
where A is the scale parameter. In the calibration exercise, A is constrained to
keep f(li) always smaller than y. A high-skilled individual working for a formal
firm is always more productive than herself as an entrepreneur. Empirical evidence
suggests that the formal sector uses more capital-intensive technology (Zenou
24
2008). This feature is reflected in this model where informal firms do not pay for
capital (i.e., sunk cost).
The income of the high-skilled owner is the highest profit of her business:
w∗HI = max πi. Let wl be the wage received by every low-skille(d worker). The first1
w ρ−1l
order condition of this profit maximization problem gives li = .
ρAyzρ
Denote N the number of micro-enterprises or number of high-skilled workers
operating in the informal sector. To pin down wl, I equate labor demand to labor
supply ∑ ∑( ) 1w ρ−1l
li = Nli = = Lρ LρAyz
i i
where N = ihLH = ihφL. Rearrange the term to yield the solution for w
∗
l and l
∗
i
( )
− ρ−11 φ
w∗l = ρAz
ρ (2.17)
φi∗h
∗ 1− φli = ∗ (2.18)φih
Term l∗i is the optimal size of informal firms. Since l
∗
i = LI/IH , this term
should be best interpreted as the optimal amount of low-skilled workers hired by
each high-skilled worker. It follows that a high-skilled worker earns in the informal
sector
( )ρ
z(1− φ)
w∗HI = π
∗ = A(1− ρ)y ∗ (2.19)φih
25
It is clear that the low-skilled workers gain when there are relatively more
high-skilled workers in the economy (higher φ) or more high-skilled workers
participate in the informal sector (greater ih). Either change leads to a larger
IH = N , shifting the demand curve for low-skilled workers outward. Each
entrepreneur will hire fewer workers and pay them better.
Steady state equilibrium
The steady state equilibrium is a 9-tuple (θ∗, w∗ , w∗ , w∗, v∗, e∗ , i∗ , u∗ , l∗HF HI l h h h i )
satisfying the following conditions
i. Equilibrium worker flows
(2.3) uhp(θ) = ehδF + ehα
(2.4) ehα = ihδI
(2.5) uhp(θ) = ehδF + ihδI
ii. Job-creation or zero-profit
( )
− r + δF + α(2.13) wHF = y κr 1 +
q(θ)
iii. Wage setting
[ ( )] [ ]
α α
(2.16) wHF = β (y− κr) + θκr + 1 + (1− β) b− (wHI − b)
r + δI r + δI
26
iv. Informal labor market clearing
( )
− ρ−11 φ
(2.17) wl = ρAz
ρ
φih
1− φ
(2.18) li =
φih ( )
− z(1−
ρ
φ)
(2.19) wHI = A(1 ρ)y
φih
Solving for equilibrium
i. Solving for θ∗, w∗HF , and w
∗
HI
Three unknowns, equilibrium market tightness θ∗, earnings of high-skilled
workers in the formal sectorw∗HF and in the informal sector w
∗
HI , can be solved
from a nonlinear system consisting of Eqs.(13), (16), and (19). They are the job-
creation condition (formal labor market demand), wage-setting condition (formal
labor market supply), and informal labor market clearing condition. The detailed
methodology is described in Section 3.
ii. Solving for e∗ , i∗h h, and u
∗
h
The analytical solution for the composition of the high-skilled labor force can
be derived from Eqs.(3),(4) and (5). These equations are linear combinations, so
only two of them are needed to yield the following expressions:
27
p(θ)δI
eh = (2.20)
p(θ)(α + δI) + δI(α + δF )
δI(α + δF )
uh = (2.21)
p(θ)(α + δI) + δI(α + δF )
p(θ)α
ih = (2.22)
p(θ)(α + δI) + δI(α + δF )
These labor ratios are true for all values of θ. It is easily seen that uh =
eh(α + δF )/p(θ) and ih = ehα/δI(wHI) (recall that δI depends on wHI). The
relationships among eh, uh, and ih are nonlinear. The equilibrium values of the
worker ratios e∗h, i
∗ ∗
h, and uh are pinned down at the equilibrium value of market
tightness θ∗.
Substituting ih out of Equation 2.19 to have wHI expressed in terms of θ and
parameters:
( )ρ( )ρ
− ρ 1− φ δI(wHI) δI(wHI)(α + δF )wHI = A(1 ρ)yz 1 + + (2.23)
φ α p(θ)α
Since p(θ) increases in θ, and δI decreases in wHI , by implicit function
theorem it is established that wHI decreases in θ. Intuitively, a more inefficient
formal market (more unemployment at the same level of unfilled jobs) makes
the option of joining the informal sector more attractive. Equation 2.23 will be
included in the system of three unknowns (θ, wHF , wHI), replacing Equation 2.19.
iii. Solving for v∗, w∗l , and l
∗
i
28
∗
By definition of θ, we have v∗ = V = θ UH = θ∗uh. The equilibrium values ofLH LH
low-skilled workers’ earning w∗l , and optimal informal firm size l
∗
i are pinned down
with i∗h.
Uniqueness of equilibrium
Equate the job-creation condition (Equation 2.13) and wage-setting condition
(Equation 2.16) and rearrange to yield an equilibrium condition for θ:
( ) ( ) ( )
r + δF + α − − α(wHI − b) ακr = (1 β) y b+ − κr − κrβθ + 1
q(θ) r + δI r + δI
where wHI = wHI(θ) and δI = δI(wHI). It is not possible to derive a
closed-form solution for θ∗; however, we know a unique θ∗ exists if two sides of
this equation move in opposite direction when θ changes monotonically. Since
q′(θ) < 0, the left hand side strictly increases in θ. To establish that the right-hand
side strictly decreases in θ, I impose
(1− β)(wHI − b) > βκθr (2.24)
The imposition of this condition also ensures the wage-setting curve slopes
upward in the (θ, wHF ) plane (see Figure 2). Recalled that wage bargaining
implies both gain and loss from a match will be distributed to the worker and
the firm. The wage paid to the worker accounts for reward and punishment made
29
by the firm. The worker is rewarded for saving firms vacancy maintenance fee
but penalized for potential voluntary leave. The penalty is proportional to the
difference in income between informal employment and unemployment. This
condition implies that at a greater market tightness, workers can claim a higher
wage only if they accept to be penalized more than rewarded.
This condition is required because the exit rate of informal firms δI is set
to be decreasing in informal income. Simply put, a higher θ makes the option of
moving to the informal sector less attractive because profit is lower and business
failure is more likely. A smaller informal profit has a leverage effect on formal
wage, but the effect of a higher failure rate on formal wage is undetermined.
This effect is positive only if the cost savings portion is smaller than the penalty
portion, which is precisely Equation 2.24. On the contrary, this model has a single
unique equilibrium unconditionally when δI is independent of wHI .
Unique values of θ∗ and w∗HF pin down a unique wHI . Other endogenous
variables are monotone functions of θ∗, so their equilibrium values are unique too.
Overall, this model conditionally has a single, unique equilibrium.
Informal sector and Beveridge curve
Figure 2 depicts how formal sector outcomes (θ∗, w∗ ∗HF , v , u
∗
h) are
determined. The left panel plots the wage-setting and job-creation curves in the
(θ, wHF ) plane. The right panel plots the Beveridge curve (also called the UV-
curve) in the (uh, v) plane.
30
The formula for the Beveridge curve is derived from the unemployment
equation (Equation 2.21)
[( ) ] 1
δ (α + δ 1−ηI F ) (1− uh)
v =
α + δ ηI uh
where δI = δI(wHI(θ)). When the elasticity of matching rate is η = 0.5, the
Beveridge curve is given by:
( )2
δI(α + δF ) (1− u 2h)
v =
α + δI uh
How do informal sector and sunk cost change these curves? In a baseline
model without the informal sector and sunk cost (as in Pissarides, 2000), the
formulae for the job-creation condition, wage-setting condition, and Beveridge
curves are
w = (1− β)b+ βy(1 + γθ)
w = y − (r + δ)yγ/q(θ)
δ
u =
δ + p(θ)
where δ is the exogenous job destruction rate, and γ is the fixed cost to maintain
vacancy per unit time. When the matching elasticity is 0.5, the formula of
31
Beveridge curve is
2 (1− u)2v = δ
u
FIGURE 2. Determination of θ∗, w∗HF , and u
∗
h
Given the presence of a sunk cost, the intercept of the job-creation curve
lowers from y to y−κr. The upper bound for wages is not y anymore: a permanent
portion of a match’s productivity is spent on vacancy maintenance. The wage-
setting curve used to have an intercept b + β(y − b) in the baseline setting. With
the presence of the informal sector, it does not have an intercept. It grows tangent
to the y-axis (wHF axis) as θ approaches zero. Its shape becomes nonlinear since
wHI is nonlinear and decreasing in θ.
The Beveridge curve, an indicator of labor market efficiency, retains its shape
and x-axis intercept despite the existence of a sunk cost and the informal sector.
Its slope, meanwhile, takes into account flow dynamics in the informal sector.
32
Determinants of the slope include the formation and destruction rate of informal
firms α and δI , in addition to the arrival rate of adverse productivity shock δF .
Flow dynamics in the informal sector can make the formal labor market
appear more or less efficient. When informal firms exit more frequently (higher
δI), the Beveridge curve steepens or shifts rightwards. Such movement indicates
a more inefficient labor market since there is more unemployment associated with
the same level of vacancies. The same result establishes if adverse shock arrives
more quickly.
When high-skilled workers move into the informal sector at a greater rate
(higher α), a similar “inefficiency” effect appears only if informal jobs dissolve
faster than formal ones do (δI > δF ). When δI < δF , a higher α shifts the
Beveridge curve leftwards. In words, a growing informal sector where informal
businesses survive better than formal firms will make the formal labor market
appear more efficient.
Some texts (such as Pissarides 2000) do not discuss the intercepts of the job-
creation curve and the Beveridge curve. Realistically it is never the case that no
vacancies are created, or there is no employment.
Calibrating Model Parameters
In the first part of this section, I describe a numerical method to solve the
theoretical model. The results are calibrated to match several empirical facts
on the informal sector in Latin American nations, taken from Maloney (2004).
33
In the second part, I analyze the sensitivity of equilibrium outcomes to changes
in the model’s parameters. Lastly, I discuss the effectiveness of a budget neutral
policy where firms with filled jobs are taxed to subsidize firms with vacancies and
unemployed workers. Quantitative and graphical results are presented at the end
of this section.
Methodology
The first step in solving for the model’s equilibrium is to compute
(θ∗, w∗HF , w
∗
HI). The system of these three unknowns are nonlinear; therefore, it
must be solved with the numerical method. The methodology involves defining
functional forms, choosing model parameters, and setting up constraints for the
system.
i. Functional forms and parameter values
The functions of market tightness are q(θ) = Kθ−η and p(θ) = Kθ1−η.
By common practice I set η = β = 0.5. With β = 0.5, we have a symmetric
situation where the match surplus is divided equally between workers and firms.
These two parameter are held fixed throughout the analysis, and so comparative
statics regarding β and η will not be discussed.
Other parameters are either borrowed from existing studies or picked
subjectively. The productivity of high-skilled workers, real interest rate,
unemployment benefit, scale parameter of the matching function, and the arrival
rate of adverse shock are based on Zenou (2008). The productivity of low-skilled
34
workers, the sunk cost for formal firms, the marginal product of low-skilled labor,
and the formation rate of informal firms are selected subjectively.
The Hosios condition η = β ensures the decentralized search equilibrium
is efficient, i.e. the two externalities of matching process cancel out each other
(Hosios, 1990). As mentioned in Section 2.1.2, when θ increases workers find
jobs more easily while firms find workers more slowly. The former effect is called
‘thick-market externality’, and the latter ‘congestion externality’. Under the Hosios
condition, a social planner facing the search frictions will maximize the utilitarian
welfare with θ and v chosen at their decentralized equilibrium values, given that
she knows the income of all agents and the exact composition of the labor force. It
is a general result that when η 6= β one externality will dominate the other.
Previously I assume δI is a decreasing function of wHI . Let δI = d(wHI)
−0.5,
where d  0. Prespecifying all other parameters, I then pick the smallest values
of the scale parameter of δI and the scale parameter of the informal production
technology (A, d) such that there exists an unique (θ∗, w∗ ∗HF , wHI) satisfy job-
creation condition (Equation 2.13), wage-setting condition (Equation 2.16), and
informal sector market clearing (Equation 2.19).
The setup of parameters gives a hypothetical economy with an equal number
of high-skilled and low-skilled workers. The elasticity of the matching rate is 0.5,
implying that when market tightness θ doubles, the job finding rate increases by
50% while the vacancy filling rate decreases by 50%. Firms pay a sunk cost of
35
4 to be able to hire workers, each of whom can produce 10. At every instant,
0.2 high-skilled worker is fired because of an adverse shock while 0.8 worker
purposely breaks the contract to move to the informal sector. To firms, that makes
a total of one job match destruction per unit time. A low-skilled worker exerts
a productivity of 2 and never earns more than an unemployed high-skilled who
receives a benefit of 3.
ii. Constraints of the nonlinear system
I add four constraints to the system of (θ∗, w∗HF , w
∗
HI): The high-skilled
workers strictly prefer i. working informally for the lifetime to being unemployed:
VI > VU ; and ii. working formally to being entrepreneur: VE > VI ; iii. The system
has a unique solution (Equation 2.24); and iv. High-skilled entrepreneurs would
never pay low-skilled workers more than their unemployment benefit: w∗l ≤ b.
These conditions correspond to four inequalities
Kθ1−η(wHF − wHI)− (r + δF + α)(wHI − b) < 0
(r + δ +Kθ1−ηF )(wHI − wHF )− (δI − δF )(wHF − b) < 0
b+ θκr − wHI < 0
wHIihρφ− b(1− ρ)(1− φ) < 0.
The first two constraints, along with tailoring parameters (A, d, ρ), are
sufficient to maintain informal wage below formal wage. A pecuniary gap does
not imply workers are worse off in the informal sector. Working in the informal
36
sector gives the workers unquantifiable utility such as the sense of being one’s own
boss or reputation of being business owners (Maloney 2004, Fields 2004).
The wage of low-skilled workers is set below unemployment benefit for
a technical purpose; however, the constraint also accords with empirical facts.
Technically, without this constraint, the informal sector will eventually ‘explode’,
i.e., the whole high-skilled labor force will move into it. Empirically, Maloney
(2004) has noticed a suppressing wage of low-skilled workers in Latin American
countries. The author observes that informal family businesses usually provide
their low-skilled employees with lodging and food. The wage of the workers is
artificially low because it does not account for these implicit benefits.
iii. Results and interpretation
Numerical simulation gives the equilibrium composition of the high-skilled
labor force: 57% working in the formal sector, 36% working in the informal sector,
and 7% unemployed. According to Maloney (2004), the informal self-employment
in Latin America is generally the largest source of employment among men after
formal salaried employment, in some cases exceeding 40% of the labor force.
The equilibrium market tightness is approximately 27, meaning the number
of vacancies is 27 times greater than the unemployment pool and more than triple
the high-skilled labor force. This situation is favorable for the workers: per unit
time firms can find only q = 0.29 worker for each vacancy, while each unemployed
worker can match with p = 7.8 firms. A representation firm with vacancy needs to
37
pay 5.4 for vacancy maintenance. The “flow cost” charged to firms per unit time is
even higher than the sunk cost, highlighting the adverse influence of a high market
tightness value to firms.
TABLE 1. Parameter Values
Parameters Definition Values
η Elasticity of the matching rate 0.5
β Bargaining power of high-skilled workers 0.5
φ Ratio of high-skilled workers in the population 0.5
ρ Marginal product w.r.t labor in informal sector 0.50
δF Arrival rate of adverse productivity shock 0.20
α Formation rate of informal micro-enterprises 0.80
κ Sunk cost for formal jobs 4
r Real interest rate 0.05
y Productivity of high-skilled workers 10
z Productivity of low-skilled workers 2
b Unemployment benefit of high-skilled workers 3.0
K Scale parameter of matching function 1.5
Once employed, high-skilled workers are paid 9.1 by formal firms. By
choosing to switch to the informal sector, the high-skilled worker accepts an
income loss of 8.5%. He can earn 8.4 as an informal entrepreneur, yielding an
income premium of 5.4 over unemployment. The ratio of formal to informal
wage is close to the case of Mexico and Colombia in late 1999 (Maloney 2004).
In Colombia, relative formal to informal wage was 1.2 in 1999 : Q3. In the same
period, the formal wage was twice as high as the informal wage in Mexico.
Calculated from the equilibrium informal wage of the high-skilled, the
equilibrium exit rate of informal firms is 1.28, higher than the effective destruction
38
rate of formal firms. Each informal enterprise recruits 2.8 low-skilled labor in
equilibrium. This ratio is slightly higher than in Maloney (2004) who observes
that 80% of micro-firms in Mexico have only one or two employees.
TABLE 2. Steady State Equilibrium
Variables Definition Values
θ∗ Market tightness or v∗/u∗h 26.78
v∗ Vacancies to high-skilled labor force ratio 1.96
w∗HF Formal wage of high-skilled workers 9.08
w∗HI Informal wage of high-skilled workers 8.37
w∗l Wage of low-skilled workers 2.99
e∗h Proportion of formally employed high-skilled workers 0.57
i∗h Proportion of informally employed high-skilled workers 0.36
u∗h Proportion of unemployed high-skilled workers 0.07
l∗i Optimal firm size in informal sector 2.80
VE Lifetime value of being formally employed 167.83
VI Lifetime value of being informally employed 167.15
VU Lifetime value of being unemployed 167.14
GNIF GNI of formal sector 2.70
GNII GNI of informal sector 2.99
A Scale parameter of informal production function 0.71
d Scale parameter of informal production function 3.69
Over the infinite horizon, the representative high-skilled worker is not much
worse-off being unemployed. The difference between the lifetime values of being
informally employed (VI) or being unemployed (VU) and being formally employed
(VE) is less than 1%. Also, the gap in the income of a low-skilled worker and an
unemployed high-skilled worker is also less than 1%. These are results from the
subjective restrictions I imposed when solving the system.
39
Equilibrium Analysis: Comparative Statics
This section discusses how equilibrium values react when I vary each of
the following parameters: y, b, K, δF , α, κ, r, φ, and z. I record the change in
equilibrium values of endogenous variables: market tightness, vacancy ratio, wages,
high-skilled work force composition, and measures of Gross National Income
(GNI).
I construct three measures: GNI in the formal sector (denoted GNIF ), GNI
in the informal sector (GNII), and total GNI (GNIT ).
( )
GNIF = φ e
∗
h(y − κr − w∗ )− u∗HF hθκr + e∗w∗ + u∗h HF hb (2.25)
GNII = φi
∗w∗h HI + (1− φ)w∗l (2.26)
GNIT = GNIF +GNII (2.27)
GNI in the formal sector is the sum of the income earned by the formal
firms, formal workers and the unemployed in equilibrium. GNI in the informal
sector is the sum of the income generated by owners of informal enterprises and
their low-skilled employees in equilibrium. Total GNI is the sum of both sectors’
income. It is also GNI per capita since the population is normalized to one.
Regarding GNIF , the term (y − κr − w∗HF ) is the profit from a match in
each period. Term θκru∗h is the flow cost firms pay for a vacant job, or vacancy
maintenance fee. The number of firms existing with a filled job is e∗h, and the
40
number with a vacant job is v∗ = θu∗h. Combine and rearrange the term, we have
GNIF simply be the total surplus of matching plus the income of unemployed
workers. The term wHF drops out of the expression, a natural consequence of
surplus linear splitting.
( ∗ ∗ )GNIF = φ eh(y − κr) + uh(b− θκr) (2.28)
Productivity of high-skilled workers
A boost in the productivity level from 8 to 12 (Appendix A, Figure 13)
almost triples θ∗ and v∗, while wages and employment in both sectors double. The
effect of higher productivity can be shown graphically with Figure 2. An increase
in y shifts the job-creation curve upward by y and the wage-setting curve upward
by βy. With β strictly smaller than one, rising productivity always raises w∗HF and
θ∗. At a higher productivity, the profit from a filled job increases, leading to more
job creation v∗. Workers stay unemployed for a shorter period and can claim a
better wage. These findings are consistent with the baseline model.
In contrast to the baseline model where unemployment is predicted to decline
monotonically, here, unemployment initially surges then declines after y = 8.5.
The plots of trajectory show that at low levels of productivity, the informal sector
is the major source of employment. It also dominates the inflow to unemployment
in this economy. As productivity rises, the informal sector loses its significance
41
to the formal sector. There is a sharp increase in formal employment and a sharp
decrease in informal employment.
This result suggests that the informal sector can shrink with development.
However, in this model, the informal sector would never be eliminated because
low-skilled workers are the inherent part of the sector. Hence, the evolution
of the informal sector, as implied by this result, is aligned with the dualists.
The final status of the informal sector, on the other hand, is leaning toward the
structuralists.
Wage gap as a portion of formal wage narrows down slightly, implying the
formal employees suffer a smaller income loss from their occupational switch.
When high-skilled workers are significantly more productive, they choose to stay
in the formal sector although they can earn equally well in the informal sector.
This notion can be related to the known fact that less developed countries have
lower productivity and more pronounced informal sector than developed ones.
Unemployment benefit
As benefit level b rises from 2 to 5 (Appendix A, Figure 14), market tightness
reduces by 40% while formal salary is 2% higher. Similar to the case above, we
can illustrate the influence of unemployment benefit on market outcomes with
Appendix A, Figure 2. A higher b leaves the job-creation curve unaffected while
shifts up the wage-setting curve shifts, leading to a smaller θ∗ and a larger w∗HF .
42
With a higher benefit, unemployed workers are better off. Hence, they are
willing to wait longer for a match. At a lower cost of unemployment, workers
can claim a higher wage, causing firms less willing to create vacancies (60%
fewer). High-skilled workers in the informal sector also get a salary rise (30%),
which significantly minimizes the formal-informal wage gap. The informal sector
outgrows the formal sector by both employment and income.
This result suggests that generous benefit in the formal sector can promote
informality. This mechanism can be related to informal experience in Latin
American countries. Maloney (2004) observes that in many families some members
participate in the formal sector while others work informally. Formal worker-
members receive benefits which are also enjoyed by informal worker-members.
Unambiguously, the formal benefit is a more stable alternative of income; hence,
a more generous benefit can encourage further informality.
The behavior of market tightness, formal wages, and formal employment
are standard. The difference, again, is in the reaction of unemployment.
Unemployment decreases by 30% instead of increasing as in the baseline model.
Low-skilled workers benefit directly from this change: their salary wl increases as
much as unemployment benefit. The sharp rise in informal GNI surmounts the
decline in formal GNI, helping the economy achieve a higher aggregated GNI.
43
Matching efficiency
A triple in K from 1 to 3 (Appendix A, Figure 15) gives the best outcomes
to all parties. As the matching process becomes more efficient, firms can create
fewer vacancies (54% fewer) and save on hiring cost. Salaries in both sectors rise,
and the wage gap narrows down by seven percentage points. The formal sector
grows by both income and employment. The informal sector loses its workforce by
14%, but its income does not change. As a result, total GNI rises by 11%.
Destruction of formal job matches
Two independent factors can destroy a formal job match: an adverse
productivity shock arriving at rate δF or the worker voluntary opting out at rate
α. Analytically, these factors affect supply and demand on formal labor market
differently. A change in δF does not move the wage-setting curve, but a change in
α makes this curve rotated in an uncertain direction. A faster rate of any factor
steepens or tilts downward the job-creation curve. Firms consider both factors
detrimental: the factors shorten the life of a match, making firms more reluctant
to open new vacancies. At a higher δF , market outcomes are clear: w
∗ and θ∗HF
become smaller. Market outcomes are undetermined when α changes.
The reaction of all endogenous variables to the two factors are similar in
both direction and magnitude (Appendix A, Figure 16 and Appendix A, Figure
17). When informal micro-enterprises form at a faster rate, equilibrium formal
44
wage and market tightness contract. This result is identical to the case of rising
δF . In the baseline model, the behavior of vacancy ratio is uncertain. Here, in both
cases, the vacancy ratio increases, possibly because unemployment increases faster
than market tightness declines. While the formal sector remains the main source
for employment, its income drops dramatically. Unemployment and informal
employment expand. Informal entrepreneurs are not better off as their salary
lowers and drops further below formal salary.
Sunk cost and real interest rate
Market outcomes react similarly to an increase in either parameter
(Appendix A, Figure 18 and Figure 19). Vacancy ratio and market tightness
have identical trajectories across both cases. Consistent with standard results,
unemployment rate rises when jobs become more costly for firms to afford. Wages
in both sectors and formal employment reduce while the informal sector and wage
gap expand. By income, the formal sector and the economy perform poorly as a
whole.
A higher sunk cost raises not only the entry barrier for firms with unfilled
jobs but also the flow cost incurring after the vacancy is filled. An increase in
sunk cost κ shifts the job-creation curve downward by κ and rotates it clockwise.
Firms’ demand for labor weakens and match surplus becomes more limited.
The wage-setting curve also shifts down, but by a smaller extent (βκr) than the
job-creation curve does. The wage-setting curve also becomes steeper i.e. tilted
45
counterclockwise. It can be shown analytically that the effect of rising sunk
cost is adverse for θ∗ but ambiguous for w∗HF . A higher interest rate affects the
job-creation curve similarly, but its influence on the wage-setting curve is also
undetermined. Real interest rate varies in a very narrow range, yet its effect is
still significant, as the effect is magnified with κ.
Other parameters
It can be shown analytically that when the exit rate of informal enterprises
δI is non-increasing in profits π = wHI , an increase in low-skilled workers’
productivity z raises informal firms’ profits, while an increase in the relative size of
high-skilled labor force φ has an opposite effect. We can conclude that both types
of workers in the informal sector benefit from a productivity boost to either type.
The informal sector becomes more attractive to high-skilled workers in that case,
but not so when the high-skilled are relatively more abundant than the low-skilled.
Plots of market outcomes for the latter situation are presented in Appendix
A, Figure 20. A doubling φ from 0.3 to 0.6, in fact, has a negligible effect on
formal and informal salaries. We can witness major changes happening to
vacancy ratio and formal labor (employed and unemployed), each of which almost
quadruples. Informal employment also shrinks by more than 70%. Despite such
significant changes, market tightness only increases by 0.4%. This result happens
due to the constant return to scale property of the matching function.
46
Put differently, the change in labor skill composition does not affect matching
outcomes in the formal sector. The rise in unemployment can be explained by a
higher exit rate of informal businesses, despite minimal change in w∗HI . We can
explain that entrants to the informal sector are fewer, but it is not sensible since
we have more individuals working for formal firms. By income, the formal sector
expands and surpasses the informal sector.
The endogenous variables do not react when I vary the low-skilled workers’
productivity, even at the 10−5 level. Also, we can noticed that in all comparative
analysis cases, w∗l does not change except when I vary the unemployment benefit.
This “unresponsiveness” should be seen as a technical issue rather than a modeling
issue. Possibly too many constraints are imposed to the problem. Alternatively,
the constraints are not set up strategically to allow for sufficient variation of the
endogenous variables. This issue is a shortcoming of my paper.
Budget neutral policy
In this model, the informal sector has two contrasting roles. First, it
competes against the formal sector for high-skilled labor. Second, it provides
employment to low-skilled labor. Put differently, the informal sector is a threat to
the formal sector but is beneficial to the low-skilled who are not directly reached
by the government. The government might want to balance the two roles by
reducing informal employment among the high-skilled and improving the income
47
of the low-skilled. This section analyzes the efficiency of a budget neutral policy
towards these two goals.
The budget neutral policy will be structured as follows: A tax will be
imposed on firms with filled jobs to finance firms with vacancies and unemployed
workers. As suggested by comparative analysis results, a reduction in sunk cost
increases formal employment and decreases both informal employment and
unemployment. Besides, raising unemployment benefit lowers unemployment and
boosts the informal sector’s income.
I denote t as the dollar tax amount imposed on the profits of the formal firms
with filled jobs, and σ as the ad valorem subsidy on firms’ vacancy maintenance
fee. This setup follows Zenou (2008) where a similar tax is used to subsidize
vacancy maintenance fee and unemployment benefit. After testing for different
specifications, I find that the policy should subsidize the flow cost κr instead of
the sunk cost κ. Financing sunk cost, in fact, hurts formal employment. Sunk
cost determines flow cost, and sunk cost subsidy also includes flow cost subsidy.
A greater subsidy requires to be financed by a more onerous tax which discourages
formal firms from job creation.
The experimenting also shows that the policy has a stronger effect overall
when setting unemployment benefit proportional to the formal wage, instead of
holding it fixed. Recall that in this model, the high-skilled entrepreneurs are not
willing to pay the low-skilled employees more than their unemployment benefit.
48
Therefore, giving a greater allowance to the high-skilled might be the right way
to improve the low-skilled’s income. How unemployment benefit is modeled in the
budget equation differs my methodology from Zenou (2008).
The government collect tax revenue from e∗h firms that are operating with
filled jobs. The total flow cost for all firms with vacancies is uhθκr. With an ad
valorem subsidy σ ∈ (0, 1), these firms only need to pay (1 − σ)uhθκr for vacancy
maintenance. The balanced budget of the government is
tEH = bUH + σκθrUH
From Equation 2.20 and Equation 2.21, EH/UH = eh/uh = p(θ)/(α + δF ), I
rearrange the above condition as
tp(θ)/(α + δF )− σκθr = 0 (2.29)
where b = rr ∗ wHF and rr is the replacement rate.
The nonlinear system needs to undergo some modifications before solving
numerically. First, I add the budget equation to the list of equality constraints.
Second, for all constraints, I change every incidence of κr to (1 − σ)κr, and
replace y with (y − t). Wage-setting, job-creation, and requirement for uniqueness
equilibrium are three constraints to be modified. In the calibration exercise, I set
the dollar tax amount t = 0.5 and a replacement rate rr = 0.2. I include ad
49
valorem subsidy σ among the endogenous variables of the system. This means σ
is solved along with (A, d) and (θ, wHF , wHI). The solution for σ is 0.33, implying
that the government pays roughly one third of the vacancy maintenance fee.
Raising the tax level has a strong positive effect on labor composition and
a weak negative effect on incomes. Figure 3 plots the trajectories of market
outcomes following a tax raise from 0.4 to 0.7. Unemployment drops significantly
by 21%. Formal employment increases by 3.5% while informal employment
decreases by 1.8%. The salary of informal high-skilled workers improves slightly
by 1%, and the wage gap between two sectors is narrowed down by 1.7%.
Meanwhile, salaries of formal workers and low-skilled workers, informal GNI,
and total GNI all shrink, though by less than 1%. Reconstructing unemployment
benefit to be proportional with formal wage produces an interesting side effect.
Equilibrium salary of the low-skilled no longer hits its boundary (which I set to be
unemployment benefit less 0.01), unlike in the initial setting where the benefit is
considered a parameter.1
Based on the calibration results, the budget balance policy is effective in
reducing unemployment and informal employment, while increasing efficiency in
the formal labor sector. However, the policy has unwanted consequences: formal
high-skilled workers and informal low-skilled workers earn lower wages, and gross
income in either labor sector decreases. Further examination shows that tax
1Pegging unemployment benefit to wage helps the model attain a desirable property, that is
to have an unemployment rate constant and independent of workers’ productivity in the long-run
equilibrium (Pissarides, 2000).
50
FIGURE 3. General Equilibrium Effects of a Budget Neutral Policy
revenue and subsidy for vacancy maintenance increase with the tax raise, but
unemployment benefit decreases. The tax hike damages firms’ profits which leads
to a smaller wage paid to the high-skilled employees. Low-skilled workers also earn
less since their compensation is capped by the high-skilled’s unemployment benefit.
A tax hike in this context helps the government achieve its goal of retaining high-
skilled workers in the formal sector. However, the policy hurts both types of
working labor in terms of income, and consequently, worsens aggregate national
income. This policy hence is not suitable to treat informality.
51
Conclusion
This chapter contributes to the related literature on informality with a fresh
view inspired by observations on Latin American countries. Previous studies
often treat the informal sector as an “absorbent” state, and few attempt to
account for its diverse nature. I fill this gap by (i) including in the model an
occupational switch as formal sector workers become informal sector employers,
and (ii) modeling the informal sector as an arena for micro-entrepreneurial firms
that hire low-skilled workers. Using an equilibrium search framework, I investigate
the interaction between two labor sectors and potential outcomes of policies
towards informality. The study yields several theoretical results, some of which
are substantiated with empirical evidence from Latin American countries.
This model assumes a perfectly competitive informal sector coexists with a
frictional formal sector. Both types of workers participate in the informal sector.
Working informally is mandatory for the low-skilled, as their productivity is too
low to be qualified for formal employment. On the other hand, taking informal job
is optional to the high-skilled. In every period, a fraction of the formally employed
high-skilled choose to switch to the informal sector. Given that it competes
directly with the formal sector for high-skilled labor, the informal sector is able to
influence formal market conditions. The Beveridge curve, an indicator of efficiency
in the formal labor sector, can shift due to changes in profits, formation, and
failure rates of informal firms. Specifically, I find that a growing informal sector
52
where informal businesses survive better than formal firms will make the formal
sector appear more efficient, as signaled by the Beveridge curve shifting inward.
The model yields a unique equilibrium under two key assumptions. First, the
survival rate of informal firms is assumed to increase with their profits. Second,
the wage paid to a worker by a formal firm should account for the possibility of
worker departure to the informal sector in a certain way. Formal salary depends on
the productivity of the match, plus two components with different purposes. The
formal firm rewards the newly hired workers for saving its vacancy maintenance
fee but also penalizes them for potential quitting in the future. The sizes of these
components depend on the bargaining power of the workers and the income
premium of informal employment over unemployment. Uniqueness is established
when the penalty component is greater than the reward component.
The model is calibrated to match empirical observations on informality
in Latin American countries. In the calibrated economy, 7% of the high-
skilled workers are unemployed, 36% work informally, and 57% work formally.
Comprehensive comparative analysis shows that informal employment among
high-skilled workers can decrease with the productivity of the high-skilled, more
efficient job matching process, lower levels of unemployment benefit, and sunk
cost of formal jobs. This finding aligns with the fact that developing countries
generally have lower labor productivity and more pronounced informal sectors
than developed countries. The result suggests that developing countries should
53
improve labor productivity through education and training programs. This can be
an effective strategy to reduce informality in the long run.
The informal sector in this model has two contrasting roles: (i) providing
income to the low-skilled workers who otherwise cannot work formally, and (ii)
taking high-skilled labor away from the formal sector which the government can
regulate. Given this informality structure, the government would want to balance
these roles with a policy that allows it to keep more high-skilled labor stay in
formal employment while improving the income of the low-skilled. In the last
comparative analysis exercise, I test whether this goal can be achieved with a
budget neutral policy. This policy has a tax on profits of firms with filled jobs,
and tax revenues are spent on unemployment benefits and subsidy to hiring fees
of formal firms with vacancies. Fixing the levels of subsidy, I show that a tax
hike in this policy package is effective in reducing unemployment and informal
employment among high-skilled workers. However, other equilibrium outcomes are
not desirable. The policy worsens labor and aggregate national income, hurting
both types of workers. Thus, I conclude, it is ineffective to tackle informality with
active government intervention in the form of labor policies.
In this chapter, the dynamics of informal labor sector and its ties with the
formal sector are modeled based on a non-standard but empirically sound view.
The research presented has certain limitations. First, occupational choice is set up
passively: a formal sector worker does not choose to become informal entrepreneur,
54
they simply become one. Second, the model is quite rigid with many constraints
imposed in the numerical algorithm. While these constraints are necessary to
yield an equilibrium, they inhibit further interesting analyses. For example, I
can show analytically that both worker types earn more when the low-skilled are
more productive. This result, because of restrictions imposed, cannot be obtained
numerically.
The analysis can be extended to include financial markets. This paper
indeed has a result in favor of financial development: the informal sector shrinks
when the output of the formal match increases, which could be due to workers
becoming more productive or formal firms having better access to finance. It will
be interesting to evaluate the contrasting roles of financial development: more
productive matches retain more high-skilled labor in the formal sector, but also
allow high-skilled workers to save up effectively to invest in their future informal
business. The calibration exercises may be enhanced too by taking into account
the distribution of worker skills, proxied by educational levels, and observations on
labor flows across employment states.
55
CHAPTER III
FINANCIAL DEVELOPMENT, EXPORT PERFORMANCE, AND FIRM
PRODUCTIVITY: EVIDENCE FROM CROSS-COUNTRY DATA
Introduction
Access to finance is a major challenge for the operation of firms of all kinds.
According to the World Bank Enterprises Survey 2018, the majority of firms
consider it to be their most significant obstacle, irrespective of their sizes. Ability
to acquire capital cheaply and quickly is particularly critical for exporting firms.
For these firms, extra expenses arise from activities such as exploring foreign
markets, setting up business relations, and paying for international transport fees.
The firms also need liquidity to prepare for adverse events such as trading partners
failing to make payment or shipment lost during transit (Foley and Manova, 2015).
The mechanisms whereby a well-functioning financial market promotes exporting
activities, however, remain inconclusive.
Financial development increases export by either allowing more firms to
export (extensive margin) or raising the level of export sales per exporting firm
(intensive margin), or both. Whether the influence of financial development
occurs along both margins of trade and if so, which margin is dominant has
important implications. The extensive margin depends on the fixed trade cost
firms have to pay upfront when enter the global market, while the intensive margin
56
is determined by the variable trade cost firms face once they begin exporting.
The reaction of the margins to financial frictions help us understand better the
financing pattern of exporting firms. Related literature agrees that financial
development affects the extensive margin yet provides mixed results regarding
the intensive margin. Studies that use country- or industry-level data mostly
find a positive effect of financial development on the intensive margin (Beck,
2002, 2003; Besedes et al., 2011; Manova, 2003, 2013). Studies that use firm-level
data find such effect to be either positive (Muul, 2008; Bricongne et al., 2012) or
statistically insignificant (Bellone et al., 2010; Berman and Hericourt, 2010). While
the inconsistency of evidence results from differences in data used by the studies,
it suggests that the impact of financial frictions on export relies on the interplay of
country, sector, and firm characteristics.
In this paper, I empirically explore the relationship between domestic
financial development and export performance, accounting for factors at the
country, sector, and firm levels. I use a detailed dataset from the World Bank, the
Exporter Dynamics Database (EDD). Domestic financial development, measured
by the private credit to GDP ratio, has a significant effect on export activities of
manufacturing industries. This effect is more pronounced in industries that require
more external finance or have less collateralizable assets. Export sales by industry
increase mainly through the extensive margin, i.e., more firms are selected into
57
exporting. Financial development does expand the intensive margin, i.e., export
sales per firm, but only for firms with higher productivity.
A comprehensive sensitivity analysis reveals that improving domestic
financial markets has statistically and economically significant benefits to exports.
For example, Cambodia, the country at the 25th percentile of the distribution of
financial development, would be able to raise the export revenue in its average
industry by 70% from a one standard deviation increase in private credit. Some
sectors benefit more from financial development, such as Pharmaceuticals,
Computing machinery, Television transmitters, and Medical appliances. The
rise in export sales by these sectors that results from a one standard deviation
improvement in private credit is almost 80 percentage points higher than the
increase in export sales by Basic ferrous metal, Fertilizers, and Starch products.
Given the richness of EDD, I verify that the effect of financial development
on export varies by firm productivity. The EDD reports measures of export
performance and export dynamics at firm-level for four firm types: entrants,
surviving entrants, incumbents, and exiters. Though EDD does not provide further
details to estimate the productivity of these firms, I can infer that incumbents and
surviving entrants are the most productive types based on their continued survival
and large sizes. Firm size can proxy for firm productivity, as established in the
seminal work of Melitz (2003).
58
My empirical analysis shows that more productive firms benefit more from a
deeper financial market. Total export sales of incumbents and surviving entrants
rise by 170% and 58%, respectively, given a one standard deviation increase in
private credit. Financial development affects export predominantly through the
extensive margin. However, the relative importance of the extensive margin
differs between incumbents and surviving entrants: Extensive margin explains
roughly 56% the increase in total export sales of incumbents, as opposed to 72%
for surviving entrants. This result implies that in terms of fixed cost financing,
financial development has a more influential role over first-time exporters. Fixed
trade costs are not limited to the entry cost borne by domestic firms to become
exporters. Existing exporters also need to pay for activities such as exploring new
destinations or developing new products. The ability to cover these costs is vital
to their survival, as indicated by a significant marginal effect on their extensive
margin. However, since existing firms tend to be mature firms, they are more likely
to meet these costs with internal liquidity.
This paper extends the literature that studies the role of finance on export
performance. Studies using country-level data measure the margins of trade
at the product-variety or sectoral level. Using the number of varieties as proxy
for the number of firms is an acceptable practice, as the canonical trade model
(Melitz, 2003) assumes each firm produces one variety. However, this method
can lead to misleading estimation of the actual extensive margin. Overestimation
59
occurs when one firm produces more than one variety, underestimation when the
variety classification is too coarse to differentiate between two distinct products
provided by two separate firms. Besides, if a firm begins exporting with the same
variety it offers for the domestic market -a stylized fact in international trade-
then no change in extensive margin is recorded. By using detailed sector-level
data, I can give a more rigorous test on the theory put forth by Manova (2013).
Manova (2013) extends the heterogeneous firm framework by Melitz (2003) into
a multiple-sector, multiple-country setting. However, using data at the sectoral
level (COMTRADE) prevents her from thoroughly testing her hypotheses. Besides
being more theoretically well-suited, my results are comparable to Manova (2013)
and other cross-country studies: the total export sales by industry in this paper
are similar to the measure of intensive margin at the industry level. This paper
also contributes to the rather small body of evidence on developing countries:
notable works include Feenstra et al. (2011) on China, Kapoor et al. (2012) on
India, and Aghion et al. (2007), Berman and Héricourt (2010), Wang (2011) on a
subset of less developed countries. My sample covers 52 countries, among which 41
are lower-middle or low income.
This research also speaks to the broader literature on finance and economic
growth (see Levine, 2005, for example). The empirical strategy in this paper
follows the pioneering work of Rajan and Zingales (1998), who use U.S. data to
explain industrial growth in non-U.S. countries. Under this method, U.S. industry-
60
level features interact with non-U.S. country-specific characteristics to form a
combined explanatory factor. While resting on strong assumptions, this method
has the advantage of exploiting industry-level variation and avoiding potential
endogeneity. The later part of this paper adopts a specification that models
heterogeneity across countries explicitly. This setting is guided by discussions on
empirical growth methodologies (Durlauf et al., 2005) and the role of institutional
quality in financial markets (La Porta et al. (1998), for example).
The remaining of this paper proceeds as follows: The next section
summarizes related work, Section 3 outlines the theoretical framework and
methodology, Section 4 presents data sources and description, Section 5 discusses
empirical results and Section 6 concludes.
Literature Review
Prominent work at the country aggregate level includes Beck (2002), Manova
(2008), Besedes et al. (2011), and Manova (2013). Beck (2002), using a 30-year
panel of manufacturing exports from 65 countries, proves that more financially
developed countries have a comparative advantage in manufacturing goods whose
production requires more external capital. Besedes et al. (2011) yield a similar
conclusion with product-level data on exports to the United States and some
European countries between 1989 and 2008. Manova (2013) confirms that more
financially advanced countries export relatively more in terms of product counts
and quantity traded. These papers all use the ratio of private credit to GDP as a
61
measure of a country’s financial development. Manova (2008), using data for 91
countries and the 1980-1997 period, finds that opening equity markets to foreign
capital increases exports, especially in sectors with higher external financing needs.
Several papers with country-level approach exploit the event of the Financial
Crisis from 2008-09. Among them are Chor and Manova (2012), Kiendrebeogo
(2013) and Iacovone and Zavacka (2009). Chor and Manova (2012) observe a surge
in interbank rates across many developed and some developing countries during
the Financial Crisis, indicating worsening financial conditions in those countries.
Examining data on 3-digit NAICS monthly imports to the United States between
November 2006 and October 2009, Chor and Manova (2012) find that countries
with higher interbank rates exported less to the United States, especially in more
financially vulnerable sectors. Kiendrebeogo (2013), analyzing crises in both origin
and destination countries, remarks that the negative effect of crises is stronger in
manufacturing industries relying more on external finance. Iacovone and Zavacka
(2009) study manufacturing export growth of 21 countries during 23 crises between
1980 and 2006, and conclude that sectors more dependent on external banking
finance grow significantly less than other sectors. The banking finance channel is
highlighted here as opposed to the inter-firm financing channel. The latter is called
trade credit in accounting term. Ronci (2004) and Berman and Martin (2012) both
use the amount of outstanding short-term credit in USD to proxy for trade credit
at the country-level. Ronci (2004) studies crises in 10 developing and emerging
62
market economies and concludes that a fall in trade financing in connection with
domestic banking crisis can lead to substantial loss of trade. Berman and Martin
(2012) focus on how exports by African countries are affected by banking crises
occurring to their trading partners.
Most evidence at the firm-level is from developed countries where firm-level
data are of good quality. Amiti and Weinstein (2011) construct a unique dataset
where Japanese manufacturing firms are matched with the banks providing trade
finance to them. Using the market-to-book value of banks as an indicator of banks’
health, Amiti and Weinstein (2011) are able to establish direct causal relationship
between bank health and exports. Their most important finding is that exports
are more susceptible to financial shocks than domestic sales. Similar evidence
comes from Paravisini et al. (2011) that also utilizes matching firms and banks
data in Peru. Additional evidence is found by Muul (2008) and Behrens, Corcos,
and Mion (2011) for Belgian manufacturers, and by Eck, Engermann and Schnitzer
(2012) for German firms, for British manufacturing firms by Greenaway, Guariglia,
and Kneller (2007), and Chilean firms by Alvarez and Lopez (2012). Minetti and
Zhu (2011) use data on Italian firms and find that credit rationing affects both
the extensive and the intensive margins of exporting. Firm-level evidence from
developing countries are Feenstra et al.(2011) on China and Kappoor et al. (2012)
on India. Berman and Héricourt (2010) provides evidence from nine emerging
economies. With access to firm-level data, these authors can compute actual need
63
of external capital and other measures of firms’ financial health such as liquidity
ratio, leverage ratio, or credit scores.
Another closely related literature investigates the role of financial
development for industrial growth. Seminal papers include Rajan and Zingales
(1998), Fisman and Love (2003), and Braun (2003). These papers are in agreement
that sectors intensive in external finance and sectors with fewer collateralizable
assets grow faster in financially developed countries. Aghion, Fally and Scarpetta
(2007) examines the effect of financial constraints on firm dynamics using data
from 16 industrialized and emerging economies. Their results show that financial
development matters most for the entry of new firms and post-entry growth in
sectors that are more dependent on external financing. In terms of methodology,
Rajan and Zingales’ interacting variable method has been used extensively by
many papers, including those cited above such as Fisman and Love (2003), Braun
(2003), Beck (2002), Berman and Héricourt (2009), Alvarez and Lopez (2012),
Manova (2013). A more extensive survey of the literature in trade and finance can
be found at Contessi and Nicola (2013) and Foley and Manova (2015).
All the literatures discussed so far confirm the importance of access to
finance and financial health in explaining the extensive margin of international
trade, consistent with the role of fixed costs defined in trade theory. The role
of finance on the intensive margin is somewhat mixed. The implication of firm
productivity on the intensive margin when firms face financial frictions have
64
been theoretically delineated in an untested hypothesis by Manova (2013), and
empirically in Berman and Hericourt (2010). This paper, therefore, attempts
to shed more light on this matter. Particularly, I examine whether productivity
correlates with the sensitivity to financial development of exporting firms.
Methodology
Theoretical framework
I use the formal framework by Manova (2013) to examine the channels of
financial development on export performance at the industry level. Manova (2013)
sets up a partial equilibrium model with multiple sectors, multiple countries, and
firms heterogeneous in productivity. Her model is a variation of the influential
work by Melitz (2003). In her model, exporting firms must borrow in a competitive
market to finance the fixed and variable trade costs, and pledge collateral to secure
their loans. The model generates several testable hypotheses, two of which will be
tested empirically with finer data in this paper.1 A more detailed version of this
model is presented in Appendix B.
Assume that countries differ by the degree of contractual enforcement λi.
Higher values of λi imply investors are more likely to receive repayment from
the firms. Sectors differ by the degree of financial vulnerability: A sector is more
financially vulnerable if it requires to cover a greater portion of trade costs with
1Proposition 1 and Proposition 5 in Manova (2013)
65
external finance (higher ds) or it relatively lacks of tangible assets that can be used
as collateral (lower ts). The sectoral difference in ds and ts assume to be driven by
differences in production technology across sectors. Firms consider λi, ds, and ts
exogenous.
The first testable hypothesis relates the extensive margin of trade with
financial development and sectoral financial vulnerability. Financial development
lowers the productivity cut-off ais relatively more in financially more vulnerable
sectors. This means more firms can enter the export market.
∂2(a ) ∂2is (aijs)
< 0, > 0
∂λi∂ds ∂λi∂ts
The second testable hypothesis relates the intensive margin of trade with
financial development and sectoral financial vulnerability. Financial development
(weakly) increases the level of firms’ exports (revenue ris) from relatively more
in financially more vulnerable sectors. The term ”weakly” is involved because
financial development only increases the export level of firms that have a
productivity between aLis and a
H
is . The liquidity constraint is binding for firms with
productivity lower than aLis and not binding for firms with productivity at least a
H
is .
∂2ris ∂
2ris
> 0, < 0
∂λi∂ds ∂λi∂ts
66
To check these propositions with the new data, I set up a regression equation
where the dependent variable is either firm counts (for the first proposition) or
average export sales per firm (for the second proposition). The set of independent
variables is the same across regressions. This set includes the level of financial
development (λi) and two interaction terms between financial development with
each financial vulnerability measure (λi#ds and λi#ts). According to the first
proposition, we should expect the coefficients of the first and second interaction
terms to be positive and negative, respectively. However, based on the second
proposition, these signs would only hold for the intensive margin of firms with
intermediate level of productivity, i.e., non-surviving entrants in EDD. This means
the signs may not hold in regressions on average export sales of incumbents and
surviving entrants.
Industry-level financial measures
The financial vulnerability of each industry is assessed by two criteria: the
need for external finance and the ability to borrow. The former is captured by
the portion of capital expenditure (CAPEX) that is financed by external funds.
CAPEX is the amount spent on tangible assets that will be used for more than one
fiscal year. Thus, CAPEX is important to the long-run viability of the business.
The main source of internal fund that firms use to finance CAPEX is the net
amount of cash flow generated from its operations (CFOP). A large, positive
external financial dependence implies that the industry does not generate enough
67
liquidity internally relative to its fixed capital investment. The former is reflected
in asset tangibility (TANG) which is the ratio of hard assets (i.e. net property,
plant and equipment) to total assets. An industry with a greater amount of
tangible assets relative to total asset has better potential to borrow, as it can
secure its debt with more collateral. The variable external financial dependence
is defined as in Rajan and Zingales (1998) and asset tangibility is defined as in
Claessens and Laeven (2003).
CAPEX - CFOP
EXTFINk =
CAPEX
Tangible Assets
TANGk =
Total Assets
In the literature of corporate finance, firms have two sources of funds to
finance CAPEX. Firms can fund CAPEX internally from cash flow from operations
(CFOP) or from the sale of long-term assets (cash inflow from investing activities).
Firms can also fund CAPEX externally by borrowing or issuing capital stocks
(cash inflow from financing activities). Among three types of cash flows, CFOP is
the most important for financial analyst to determine the efficiency of the business.
CFOP is calculated as the sum of funds from operations plus changes in working
capital.2
2Funds from operations are, at the simplest, net income plus depreciation and amortization
and deferred taxes. Working capital is current assets less current liabilities, and the term
”current” implies ”less than one fiscal year”. Changes in working capital that raise CFOP include
68
asset tangibility strictly lies between zero and one, since tangible assets is a
component of total assets and never receives negative values. Meanwhile external
financial dependence can be either positive or negative. When both CAPEX
and CFOP are positive, if the firms have more capital expenditures than its own
liquidity and have to seek external funds, then external financial dependence will
be positive. A negative external financial dependence implies that firms are able
to generate more than sufficient cash to fund CAPEX. A more negative external
financial dependence (or highly positive CFOP) indicates a healthier business
as it shows this business has substantial liquidity of its own to survive and less
dependence on external finance. The firm can use free cash flow to develop new
products, make acquisitions, pay dividends and reduce debt. CAPEX is always
positive but CFOP can be negative. Some industries typically have negative CFOP
such as Pharmaceuticals, Oil and gas, Biotechnology, Drug manufactures and
Medical instruments. Firms in these industries often make large investments which
promises pay-offs only in the long term.
increase in account payables, decrease in account receivables, and decrease in inventory. Increase
in account payables (a current liabilities item) is the payment amount the business can defer to
make to its customers. Account receivables (or trade receivables, or trade credit, a current asset
item) arises from selling merchandise or providing services to customers on credit. A decrease
in account receivables means customers have made payment on their unpaid bills. Finally, a
decrease in inventory (a current assets item) implies that the firm has more sale, thus receiving
more cash.
69
Interaction variable method
Exporting firms perform worse when facing financial imperfections because
they could not raise as much financial capital as needed. Data on external financial
usage and borrowing, if available, reflect the equilibrium values in a frictional
market. Using such data and measure of financial frictions in the same equation
to explain export performance could lead to severe endogeneity bias. Besides, such
data are usually not available for a large number of countries and industries.
To isolate the effect of financial frictions across countries while still exploiting
the variations in financial need across industries, Rajan and Zingales (1998)
developed a technique known as ”interaction variable method”. According to
this method, researchers can use financial characteristics of public firms in the
U.S. instead. Because the U.S. arguably has the least frictional financial market
in the world, the equilibrium values observed in U.S. firms are likely their true
demanded value of financial capital. Put differently, as financial markets in less
developed countries become less frictional, the level of financial capital attainment
of firms in such countries will approach that of U.S. firms. As a result, the capital
structure (the mix of debt and equity) and asset structure (the mix of tangible and
intangible asset) of non-U.S. firms will eventually resemble those of U.S. ones.
Why U.S. sectoral data can be applied on similar sectors of non-U.S.
countries also rests on assumption of consistent technological difference. Rajan
and Zingales (1998) argue that the difference in industry-level financial structures
70
is driven by the technological difference among industries. Such difference
is preserved across time and across countries. For example, a typical firm in
Chemical products industry in the U.S. may demand relatively more external
finance than a typical firm of the same industry in Mexico. However, in either
country, the demand for external finance of firms in Chemical products is greater
than that of any other industries. This method thus provides a rank of sectoral
financial vulnerability rather than the actual demand of industries globally.
This method requires the industry-level financial measures to be computed
from a 10-year period preceding the sampling period of the research and the
U.S. to be excluded from the sample. Doing so helps prevent potential channels
of endogeneity: there might be omitted factors that influence both financial
characteristics of firms in the U.S. and export performance. The 10-year time
span helps smooth out any noises during the business cycle and credit cycle. This
interaction variable approach, nonetheless, is not perfect. The industry measures
come from large public firms which are less liquidity constrained than small firms.
Baseline specification
Export performance is modeled using a standard gravity setting (Anderson
and van Wincoop, 2003) that allows for variation at the level of country by
industry by firm type. I employ the interaction variable method by Rajan and
Zingales (1998), in which the country-level measure of financial development,
private credit, is interacted with two industry-level measures of financial
71
characteristics, external financial dependence and TANG. I also include the GDP
of the exporting country and a set of fixed effects to control for any omitted
variables. The basic regression equation takes the form:
LogY ikt = β0 + β1LGDPit + β2PRIV CREDit
+ β PRIV CRED × TANG + β PRIV CRED × EXTFIN (3.1)3 it k 4 it k
+ αi + αk + αt + ikt
where LogY ikt takes on different measures of export performance for industry k in
country i, in year t. Y will be total export flow, number of exporting firms, and
average export level of firms. External financial dependence and asset tangibility in
industry k are denoted EXTFIN and TANG, respectively.
This equation will be used to explain total export flow, extensive margins
and intensive margins. Export flow (total export sales) can be decomposed into
extensive margin, i.e., the increase in number of firms entering international
trade, and intensive margin, i.e., the average export sales of exporting firms.
The measure of extensive margin is the number of exporting firms by ISIC Rev
3 industry, and the intensive margin is measured by the average export sales of
exporting firms. EDD does not provide the level of export volume, therefore for
each industry, I multiply the number of exporters with average exporter size to get
total export volume.
72
EDD reports extensive and intensive margins by different firm types:
exporters, entrants, incumbents, survivors, and exiters. Exporters at time t
(Exportert) refers to all firms that export in year t. Among these firms, the
ones that did not export last year are called entrants (Entrantst), the ones that
exported last year are incumbent (Incumbentst). Exporters in year t is the sum of
the entrants and incumbents of the same year, minus those that exported last year
but cannot continue this year (Exitert). Among the entrants, a surviving entrant
(Survivort) is the one that will continue to export in the following year (t + 1).
Survivort thus becomes a subset of Incumbentt+1.
The coefficients of interest are β2, β3 and β4. The sign of private credit
(β2) is expected to be positive. In financially more developed countries where
firms can easily secure a loan, the productivity cut-off for exporting is likely
lower. The coefficient of the interaction term between private credit and asset
tangibility (β3) is expected to be negative. Industries with relatively more tangible
asset can borrow more, since loans are pledged with firms’ tangible assets. Being
able to borrow more implies that industries are less dependent on the financial
development of the country. The coefficient of the interaction term between private
credit and external financial dependence (β4) is expected to be positive. Greater
needs on external finance suggests that firms in those industries should rely more
on the country’s financial system to obtain more liquidity.
73
Sensitivity to financial development
Differentiating both sides of Equation 3.1 with respect to private credit yields
the marginal effect of exports to private credit:
∂LogY ikt = β2 + β3 ∗ TANGk + β4 ∗ EXTFINk (3.2)
∂PRIV CREDit
This step is necessary since we want to account for both measures of
sectoral financial vulnerability. As presented later in the Data section, there is
no correlation between these measures. A sector can have a large ratio of asset
tangibility, while requiring external finance for its capital expenditure, and vice
versa. The calculation of marginal effect is also helpful, as we can identify the
heterogeneity in the sensitivity to financial development. In particular, we can
evaluate which trade margin is the primary channel through which financial
development affects exports, whether different firm types display unequal
sensitivity to financial development, or which sectors can benefit more from
financial improvement. Lastly, given the same degree of financial improvement,
we can assess whether exports from countries of different financial conditions will
experience the same benefit.
Modified specification I
The baseline model (Equation 3.1) has a full set of fixed effects to control
for factors that are invariant for a country, for an industry or for a particular
74
year. The time fixed effect accounts for factors that apply to all countries and
industries such as a global business cycle, the industry fixed effect accounts for
time-invariant difference in industrial technology, and the country fixed effect for
time-invariant country-specific underlying conditions. Such factors are numerous
and sometimes not measurable, and failing to account for them can lead to biased
estimates. Having fixed effects thus is necessary.
One problem with using fixed effects is that the fixed effects can absorb too
much variation from the data, leading to imprecise estimates. In this exercise, I
remove the country fixed effects, allowing private credit display a stronger effect.
private credit may not have substantial time-series variation, since the sample
is fairly short by temporal dimension. Indeed, relaxing the gravity equation by
omitting country fixed effects is not uncommon in the literature.3
That fixed effects identification strategy should be approached with care
has been discussed in empirical growth literature. Most studies in this literature
employ panel data analysis where income growth is regressed on various country
factors, including financial development. Including country fixed effects removes
between-country variation of the regressors, thus identification of the slope
parameters only rely on within-country variation. If the regressors are highly
persistent over time, which is not rare for a handful of country factors, the within-
3For example, the first paper that employs EDD (Fernandes et al. 2015) does not use country
fixed effects either in their gravity equation estimations. Choi and Lugovskyy (2019), in a study
that examines the effect of private credit on export quality of U.S. Imports, also follow this
practice. They argue that having country fixed effects would remove meaningful cross-country
differences that they are interested in.
75
country variation would be too meager to yield significant estimation. Having
country fixed effects in such cases could lead to a misleading conclusion that a
given country factor ”does not matter” while it should better say that the effect
cannot be identified from the data at hand (Durlauf et al., 2005).
To balance the trade-off between the benefit of bias reduction and the
cost of variation reduction in using country fixed effects, some studies offer an
alternative: that is to proxy for omitted variables (Temple 1999, Griliches and
Mairesse 1995). Controlling for observed sources of cross-country heterogeneity
leaves more identifying variation in the regressors of interest and is informative
itself. Furthermore, having country fixed effects makes inference based on large
changes of the regressors (for example, ”one standard deviation increase in private
credit”) implausible, since such large changes are unlikely to occur within-country
(Mummolo and Peterson, 2018).
Based on the above discussion, I replace country fixed effects in the baseline
equation with country factors that are known to influence financial development
and likely export performance. I use GDP per capita (in log) to isolate the effect
of overall economic development, and legal rights index, rule of law index and
former legal origin to proxy for institutional quality. Legal rights index measures
the stringency of legal protection towards the rights of borrowers and lenders,
while rule of law index reflects the quality of contractual enforcement, property
rights and crime likelihood. Legal origin refers to the historical legal origin of the
76
country, which can be British, Scandinavian, German, French or Socialist. The role
of these factors on shaping financial development has been confirmed in La Porta
et al. (1998), Acemoglu et al. (2001), Beck et al. (2003), and Do and Levchenko
(2007). The updated regression equation is as follows:
LogY ikt = β0 + β1LGDPit + β2PRIV CREDit
+ β3PRIV CREDit × TANGk + β4PRIV CREDit × EXTFINk
+ LEGALit + LGDPPC (3.3)it +RULELAWit
+ SC ORi + FR ORi + SC ORi + UK ORi
+ αk + αt + ikt
where LGDPPC denotes log of GDP per capita, RULELAW rule of law,
LEGAL legal right index, SO OR Socialist origin, FR OR French origin, SC OR
Scandinavian origin, and UK OR British origin. The coefficients of interest β3
and β4 are identified using variation in financial vulnerability across industries and
variation in financial development between countries, controlling for aggregated
time trends.
Modified specification II
To allow for statistical comparison of marginal effects across firm types, I
modify Equation 3.3 into two versions as follows
77
LogY ikt = β0 + β1LGDPit + β2PRIV CREDit
+ β3PRIV CREDit × TANGk + β4PRIV CREDit × EXTFINk
+ β5IEntrant + β6PRIV CREDit × IEntrant
(3.4)
+ β7IEntrant × PRIV CREDit × TANGk
+ β8IEntrant × PRIV CREDit × EXTFINk
+ αk + αt + ikt
and
LogY ikt = β0 + β1LGDPit + β2PRIV CREDit
+ β3PRIV CREDit × TANGk + β4PRIV CREDit × EXTFINk
+ β51ISurvivor + β61PRIV CREDit × ISurvivor
+ β61ISurvivor × PRIV CREDit × TANGk
+ β71ISurvivor × PRIV CREDit × EXTFIN (3.5)k
+ β52INon−survivor + β53PRIV CREDit × INon−survivor
+ β61INon−survivor × PRIV CREDit × TANGk
+ β62INon−survivor × PRIV CREDit × EXTFINk
+ αk + αt + ikt
78
I add to Equation 3.3 dummies of firm types: entrants, surviving entrants
(or survivor), and non-surviving entrants (or non-survivor). I also include the
interaction of those dummies with the main regressors. Incumbent is the reference
group in both versions. These equations will be estimated with a transformed
dataset where observations of different firm types are stacked on each other.
I will apply OLS method to estimate four equations: Equations 3.1, 3.3, 3.4,
and 3.5. All GDP and trade variables are in log level. The standard errors are
clustered at country level.
Data
Export performance data
Export performance at the firm level is obtained from the World Bank
Exporters Dynamic Database (EDD). The EDD reports exporter statistics based
on customs data collected from more than 50 countries. This dataset allows me
to exploit different measures of extensive and intensive margins for different
exporter types in various manufacturing industries (except oil and gas). The
EDD is available at different disaggregation levels; however, as this research
focuses on the characteristics of exporting countries, I choose the version at the
country-year-industry level. Export values are measured in US Dollars (USD). I
also limit the sample to manufacturing exporters who export at least 1000 USD a
year. The EDD has a fairly short time-series (1997-2014) and strongly unbalanced
79
cross-sectional coverage; thus I choose the nine-year period 2004-2012 when most
countries are represented. The final sample includes a total of 54 countries and 55
three-digit ISIC Revision 3 manufacturing industries.
Country-level measures
Among these 52 countries, 5 are high-income, 8 are upper-middle income,
18 are lower-middle income and 21 are low-income. This classification is based on
GNI per capita in 2003. Country size is measured by GDP in PPP (constant 2011
international dollar). The smallest countries by GDP in the sample are Rwanda in
2004 and 2005, and Timor Leste in 2006. The largest is Mexico for the year 2008,
2011 and 2012. The largest country is almost 180 times larger than the smallest
country in terms of GDP.
Financial development is measured by the financial resources issued to
the private sector by deposit money banks as a share of GDP, or private credit
to GDP ratio (in short, private credit). Domestic money banks are defined as
commercial banks and other financial institutions that accept transferable deposits,
such as demand deposits. Private credit excludes credit extended by central bank
and other financial institutions such as insurance firms. It also excludes credit
offered to state enterprises.
private credit is a widely used measure of financial development. The
difference in the components of banks’ balance sheet across countries of different
income levels has been noted in Langfield and Pagano (2015), and Demirguc-
80
FIGURE 4. Private Credit and Income Per Capita
Kunt and Huizinga (2010). Banks in high-income countries offer more long-term
loans to households and small and medium-sized enterprises. Banks in low and
middle income countries focus more on government bond holdings and corporate
lending, often short-term. However, private credit to GDP is a crude measure
at best. It only indicates the quantity, not the quality of the financial system or
the ease to access credit of consumers and firms. Its major advantage over other
financial measures is availability for many developing countries. private credit is
also highly correlated with other indicators of financial system efficiency, if data
81
on such indicators are available (Beck, 2015).4 private credit is obtained from the
Global Financial Development database from the World Bank.
There is a wide variation in the ratio of private credit to GDP in our sample.
The average value of private credit is 36.9%. Timor Leste has the smallest value
of private credit at only 2.96%, while Denmark has the highest private credit of
182.6%. Other countries with very low value of private credit are Guinea (4.13%),
Yemen (6.3 %), Zambia and Uganda (8.7%). All are small low-income countries.
Unsurprisingly, the top countries with highest private credit are large high-income
countries: Spain (157%), Portugal (141.8%), and Norway (92.6%). However, the
fifth position belongs to Mauritius (77.5%), a small upper-middle income country.
private credit has a fairly strong correlation with GDP per capita (almost 0.7) but
weak correlation with GDP (0.3).
Data on institutional quality are publicly available. Rule of law, an index
from -2.5 to 2.5, is taken from Worldwide Governance Indicators (Kaufman et al.
2010). Legal rights, an index ranging from 0 to 10, is from Getting credit module,
Doing Business database, the World Bank. Historical legal origin is from CEPII.
These data are not available for all countries and all years in the sample. Our
sample does not contain any country of former German origin.
4For example, bank net interest margin and lending deposit spread.
82
FIGURE 5. Cross-sectional Distribution of Private Credit
Industry-level measures
External financial dependence (EXTFIN) and asset tangibility (TANG) are
constructed from Compustat, a database of balance sheets data of public firms
in the United States. The measures are averaged for each firm between 1993 and
2003, then the median value is chosen for each industry. The standard practice
popularized by Rajan and Zingales (1998) is to calculate these measures for a
10-year period before the starting point of the sampling period. Averaging over
ten years will cancel out the temporal noises in the data, which may be results of
business cycles and financial cycles. I then use Haveman’s concordance table to
convert Compustat data, which comes in 3-digit and 4-digit 1987 U.S. SIC code, to
83
FIGURE 6. Temporal Changes in Private Credit across Income Groups
3-digit and 4-digit ISIC Rev.3. After matching, I reduce any 4-digit ISIC code into
3-digit to ensure compatibility with EDD data.
On average, 30% of an industry’s capital expenditure needs to be covered by
external finance, while tangible assets constitute 26% of the industry’s total assets.
The sector that requires the most external finance is ISIC 242, ”Other chemical
products”. The capital expenditure in this sector is three times larger than the
internal cash flow (294%). Examples for products in this sector are pesticides,
pharmaceuticals, soap and detergents. ISIC code 242 belong to the same Division
24 with ISIC 241 ”Basic chemicals” (e.g. fertilizers) and ISIC 243 ”Man-made
fibers” (e.g. nylon). On the other end of the spectrum, ISIC 160 Tobacco products
84
can generate the internal liquidity four times more than its capital expenditure.
ISIC 160 Tobacco is a atypical industry because it is considered most vulnerable
by one measure (TANG) but least vunerable by the other measure (EXTFIN).
Television and radio receiver (ISIC 323) is considered more financially
vunerable by both measures. This sector has less than 10% of total asset in
tangible form, while the need for external finance is 1.5 times greater than its
capital expenditure. The sector with the highest asset tangibility is ISIC 201 ”Saw
milling and planing of wood”. Based on the top five and bottom five industries
measured by asset tangibility (Table A9 in Appendix B), we can tell that the
industries with low asset tangibility are those having more intangible assets e.g.
intellectual property. The pattern for external financial dependence (Table A8 in
Appendix B) is not as clear. For example, three industries ISIC 359, 353 and 352
are from the same Division 35 Manufacture of other transport equipment. Yet
ISIC 352 Railway and ISIC 353 Aircraft do not require external finance, while ISIC
359 Transport equipment not elsewhere classified does.
Export pattern
Table A2 in Appendix B summarizes the descriptive statistics of the main
variables. On average, half of the exporters in a year are entrants, half are
incumbents, and half of the exporters would exit the market in the following
year. Among the entrants, less than half can survive until the next year. The
average survival rate of entrants after the first year is 28%. In terms of firm size,
85
FIGURE 7. Two Measures of Financial Vulnerability
the average exporter exports 0.9 million USD, nine times larger than an average
entrant, but lower than the average size of an average incumbent that exports over
1.8 million USD. Surviving entrants, on average, are twice the size of entrants. Not
surprisingly, exiters have the smallest average size at only 0.08 million USD.
Despite being outnumbered by low and lower-middle income countries, high
and upper-middle income countries export more in all ISIC Rev. 3 divisions (see
Table 3). The top income countries export the most in Division 27, 30 and 34, and
the least in Division 16, 22 and 23. The low and lower-middle income countries
export the most in Divisions 17 and 18, both of which are related to textiles. The
86
FIGURE 8. Firm Size by Export Value
low-income countries export the least in Division 30, 31, 32. The following table
reports exports value averaging across all countries by ISIC division
Empirical Results
In this section, I will discuss the OLS estimation results from four
specifications described in Methodology section. Each specification is regressed
on three measures of export performance: total export flows, intensive margin,
extensive margin. Each specification is repeated for four firm types: exporting
firms, entrants, incumbents, and surviving entrants. Table 4 to 6 report the results
from a basic gravity equation with a full set of fixed effects (Equation 3.1) and
87
TABLE 3. ISIC Revision 3 2-digit divisions
Average Exports Value (million USD)
15 - Food products and beverages 437
16 - Tobacco products 64.4
17 - Textiles 376
18 - Wearing apparel; dressing and dyeing of fur 649
19 - Tanning and dressing of leather 144
20 - Wood and of products of wood and cork 125
21 - Paper and paper products 397
22 - Publishing, printing and reproduction of recorded media 45.7
23 - Coke, refined petroleum products and nuclear fuel 29.5
24 - Chemical products 706
25 - Rubber and plastics products 318
26 - Other non-metallic mineral products 227
27 - Basic metals 1480
28 - Fabricated metal products, except machinery and equipment 315
29 - Machinery and equipment n.e.c. 600
30 - Office, accounting and computing machinery 550
31 - Electrical machinery and apparatus n.e.c. 234
32 - Radio, television and communication equipment and apparatus 427
33 - Medical, precision and optical instruments, watches and clocks 188
34 - Motor vehicles, trailers and semi-trailers 1080
35 - Other transport equipment 146
36 - Furniture and manufacturing n.e.c. 295
a modified version of Equation 3.1 where origin country fixed effects have been
replaced with additional country-level controls (Equation 3.3).Tables 7 to 9 present
the marginal effects of private credit.
Regression results
In Tables 4 to 6, LGDP and private credit have positive coefficients,
indicating that larger countries with more developed financial system have more
and larger exporting firms of all types. Because the measure of total export flows
is the product of the number of exporting firms and the average firm size and these
measures are log transformed in the regression, it is true for every explanatory
88
variable that its coefficient from regressing on total export flows is the sum of
its coefficients from regressing on firm counts and regressing on firm sizes. The
relative magnitudes of the coefficients from regressions with firm counts and
firm sizes reveal which margin of trade account for more of the variation in total
trade flows. Given the regression results, whether extensive margin or intensive
margin plays the dominant role depends on the specifications. The regression
specifications with origin country fixed effects (Column 1 to 4) predict that larger
countries can export more mainly because they have larger firms. Intensive margin
thus explains more than 60% of the variation in total export flow in this case.
Replacing origin country fixed effects with additional country controls (Column 5
to 8) balances out the roles of the two margins of trade. Extensive margin explains
slightly more than 50% of the variation in total export flows across countries of
different sizes.
Decomposing the overall effect of financial development by extensive and
intensive margin is not straightforward as in the case of LGDP since private credit
also enters the regression in interactive form. Isolating the effect of private credit
needs sensitivity analysis, of which results are discussed in the next subsection.
Nevertheless, comparing results from the first and last four columns in Table
4 to 6 reveals that the inclusion of origin country fixed effects substantially
influences the estimated coefficient of the stand-alone private credit. While the
coefficient of private credit appears large, positive and statistically significant
89
across all regressions without country fixed effects (Column 5 to 8), this is not
the case when country fixed effects are added (Column 1 to 4). In these more
stringent specifications, private credit only displays the expected positive sign
and a considerable magnitude when it explains the export flows of entrants and
survivors, the number of survivors, and the average size of entrants. In other
cases, the coefficients of private credit are small and insignificant. The change in
estimated coefficients can be explained as a result of data characteristics. The
ratio of private credit to GDP is fairly stable for each country, which means it
does not have much time-variation.5 Including the origin country fixed effects
in the regression, meanwhile, demeans the data. This makes private credit,
due to its little temporal variation, seem to not have any effect over export
data. The adequately large effect of private credit in regressions with certain
export measures, such as the number of surviving entrants, may occur due to
the relatively small time-variations of these export measures. Low variations in
dependent variables, in contrast, magnify the impact of the explanatory variables.6
This explanation is confirmed by the fact that the coefficients of LGDP are
also larger in regressions where private credit has statistically significant effect.
Such explanation is not sufficient, since LGDP does not lack time-variation
yet the relative role of the margins of trade still varies with different regression
5For each country, I calculated the standard deviation of private credit and LGDP over time.
A paired t-test shows that the mean standard deviation of private credit is significantly smaller
than that of LGDP at the 1% level.
6Overall standard deviation of the average firm size is largest for incumbents, followed by
survivors and then entrants.
90
specifications. The country fixed effects not only take away the mean of data but
also absorb time-invariant country-specific unobservable factors. These factors
capture the inherent difference among nations that influences the development of
the financial system, and subsequently, exporting pattern.
Across all specifications, the coefficients of the double interaction terms
between private credit and measures of sectoral financial conditions are consistent
and display expected signs. The magnitudes of these coefficients are somewhat
smaller in the absence of origin country fixed effects, yet their signs and
significance levels are unaffected. The coefficients of the interaction term between
private credit and asset tangibility are negative and highly significant, confirming
that domestic firms in sectors with relatively less tangible asset find it easier to
export and increase export volume when their origin country are more financially
developed. This interaction term is not statistically significant when explaining the
intensive margin of exporting firms (Table 6, Columns 1 and 4).
The dependent variable here is the average size of all firms that export in a
certain year without regard to their status in the year before or after. This result
suggests that the heterogeneity among exporting firms is not negligible and if we
regard all exporting firms to be the same, we would not see the combined effect
of financial development and asset tangibility on firm-level export volume. The
coefficients of the interaction term between private credit and external financial
dependence are positive and also highly significant in regressions with total export
91
flows and extensive margins (Table 4 and 5) but not with intensive margins
(Table 6). This confirms a lower threshold of selection into exporting for firms
in more financially dependent sectors in more financially advanced economies, yet
whether such firms are able to export more is statistically uncertain. Comparing
by absolute value, the coefficients of private credit and external financial
dependence interaction term tend to be smaller than those of private credit and
asset tangibility interaction term. Possibly this occurs due to external financial
dependence being more disperse, as opposed to asset tangibility which is limited
between zero and one.
Sensitivity to financial development by industry and country
Because private credit enters the regression in three terms: a stand-
alone level and interacting with two measures of sectoral financial conditions, a
sensitivity analysis is necessary to isolate the effect of private credit on export
performance across sectors, countries and exporting firm types. Particularly, I
will estimate the growth of total exports flows, number of exporting firms, and
average exporting firm size given a certain increase in private credit. To enable
comparability, I calculate the marginal effect when separately varying the degree
of asset tangibility TANG, external financial dependence EXTFIN, and financial
development private credit, while keep other variables at their means. The
92
sensitivity analysis is a post-estimation procedure, and I will conduct sensitivity
analysis for the regression without origin country fixed effects (Equation 3.3).7
Table 7 reports the marginal effects of private credit calculated from the
regression on total export flow for each industry. The industries that benefit
the most from an improvement in the country’s financial development are
Other chemical products, Television and radio receivers, Office, accounting and
computing machinery, Television and radio transmitters, and Medical appliances.
These industries are predicted to raise their export volume by 2.5% up to 3%, and
increase the mass of exporting firms by 1.4% up to 2.2%, given a one percentage
point increase in private credit. These industries are either the most intensive in
external finance or the least intensive in asset tangibility.
The country that benefits the most from improvement in exports by these
five sectors is Costa Rica where the five sectors make up 32% of its exports. It
is followed by Mexico (27% exports) and Denmark and Dominican Republic
(19% and 18%). The industries that least benefit from financial improvement
include ”Grain mill and starch products”, ”Printing”, ”Man-made fibres”, ”Basic
chemicals”, and ”Basic iron and steel”. These industries are predicted to raise
their export volume and number of exporting firms by 1%, given a one percentage
point increase in private credit. Basic chemicals contribute to one-third of Iran’s
exports while Basic iron makes up one half of Macedonia’s exports. Total export
7The marginal effects of private credit calculated from the regression with country fixed effects
(Equation 1) are mixed, since the origin country fixed effects absorb too much variations in the
data, as discussed before.
93
share of the five industries is the largest for Macedonia (47%), Iran (40%), Georgia
(32%) and Senegal (32%). A low sensitivity means little gain from financial
improvement, but it also means if there is a disruption in the domestic financial
system, the exports of such countries may not be hampered as much.
In the distribution by external financial dependence, the sector at the 25th
percentile is ”Domestic appliances”, at the 75th percentile is ”Optical instruments
and photographic equipment”. A one standard deviation increase in private
credit ratio would raise total exports in these two sectors by 53.3% and by 60.5%,
and increase the number of exporting firms by 35.1% and 40%, respectively. In
the distribution by asset tangibility, the sector at the 25th percentile is ”Other
textiles”, at the 75th percentile is ”Basic non-ferrous metal”. A one standard
deviation increase in private credit ratio would raise exports in these two sectors
by 74.7% and by 43.2%, and expand the mass of exporting firms by 43.7% and
32.6%, respectively. This means the sector at the 75th percentile in the distribution
of external financial dependence can raise their exporting volume by 7 percentage
point higher than the sector at the 25th percentile. Similarly, the difference in how
many more exporting firms can stay exporting is 5 percentage point between two
sectors. The sector at the 25th percentile in the distribution of asset tangibility
can raise their exports by 31.5 percentage point more than the sector at the 75th
percentile. The difference is 11 percentage point for the number of exporting firms.
These differences are statistically significance at the 5% level at the least.
94
Table 8 reports the sensitivity to financial development of exports in an
average sector (one with the values of external financial dependence and asset
tangibility at the means of the data). I calculate the marginal effects for countries
at different percentile of private credit. It is clear that improvements in financial
development improve trade in all countries. Further tests show that the effect
on total export volume and extensive margins is significantly different for two
countries at the bottom of the distribution (Timor Leste and Guinea) compared
to the top country on the list (Denmark).
A one standard deviation improvement in financial development will bring
Cambodia, the country at the 25th percentile of the distribution by private
credit ratio, to the position of El Salvador. The impact on Cambodian exports
is statistically and economically significant: its total exports would grow by 58.4%
and the number of exporting firms would grow by 26.7%. Cambodian exports is
dominated by two sectors: Knitted and crocheted fabrics and articles (29% country
total exports) and Wearing apparel (63% total exports). A one standard deviation
increase in private credit ratio would raise exports sales in the first sector by 53%
and in the second sector by 86%. More than one half of the increase comes from a
higher number of exporting firms.8
8I further estimate the marginal effects of private credit on export performance between two
country groups: ”Low Income” and the rest. The marginal effects are not statistically different
between two groups (for more details see the last 3 tables at the end of the text).
95
Sensitivity to financial development by firm types
Table 9 reports marginal effects of private credit on different firm types that
are calculated from estimation results of Equations 3.4 and 3.5. Results from the
first section of this table shows that the effect of financial development mostly
occur along the extensive margin. A one standard deviation increase in private
credit associates with an increase in total industry sales of 70%. Over 65% of
this effect comes from a greater number of firms becoming or surviving in the
export market. Not considering firm productivity, we can conclude that financial
development has a small and insignificant effect on the intensive margin, i.e.,
average export sales per firms. However, when we decompose the pool of exporting
firms into incumbents and entrants, a new pattern emerges. Financial development
helps incumbents and surviving entrants expand their export sales. Both margins
of trade measured in these two firm types react positively to financial development,
with a stronger degree for incumbents. The effect of financial development on non-
surviving entrants is small and not significant. Interestingly, if we treat all entrants
the same, we may incorrectly conclude that financial development only benefits
incumbents.
The relative importance of extensive margin also differ between incumbent
and surviving entrants. Extensive margin explains roughly 56% the increase in
total export sales of incumbents, as opposed to 72% the increase in total export
sales of surviving entrants. Incumbents, even after breaking into the global market,
96
still incur some fixed cost from exporting in new destinations or introducing new
products. These costs, however, may not be as vital as the cost incurred by first-
time exporter. Also, if incumbents are mostly mature firms, then the results
suggest that exporting firms may have different financing pattern among stages
of development. Younger exporting firms may rely more on outside capital while
older firms are able to cover these costs with internal liquidity. Rajan and Zingales
(1998) points out a common wisdom in corporate finance that firms are more
dependent on external financing early in their life than later. Using data from
Compustat, the authors show that external financial usage and investment on
fixed capital are the highest in the first four years of corporate life, then converge
rapidly to zero after passing year 10th.
Empirical evidence on the intensive margin diverges from the predictions of
Manova (2013). The theory predicts that only firms of intermediate productivity
can grow under the influence of financial development. The signs of the second
derivatives, i.e., the interaction terms between private credit and each sectoral
financial measures, are predicted to hold for firms of intermediate level only. These
are not true for my results. The signs do not hold for intensive margin of average
exporting firm, yet appear mixed for specific firm types (Table 6). Financial
development has positive impact on average export sales of specific firm types
97
through the asset tangibility channel. Results on marginal effect show that even
most productive firms can still benefit from financial development.9
Robustness Check
In this section, I conduct three exercises, using the regression specification
with country controls. In the first exercise, I replace private credit ratio with
its initial level (measured at the beginning of the sample period) to control
for potential reverse causality. In the second exercise, I control for potential
presence of state-owned enterprises which would undermine causal inference. In
any exercise, the results remain robust. I conclude this section with a detailed
discussion on other threats to the identification strategy. As most of the threats
cannot be addressed with the data at hands, the results in this chapter should be
best interpreted as strong associations.
Potential reverse causality
As private credit and export performance are measured over the same
time period, a valid concern arise that private credit rises due to more intense
export activities from sectors that require more credit. I dismiss this concern
9The effect of financial development on trade, according to my analysis, mainly operates
through the extensive margin. This is different from Manova (2013) as she finds one-third of the
effect of credit constraints on trade coming from limited firm entry into exporting (extensive
margin), and two-thirds coming from contractions in exporters’ sales (intensive margin).
However, our results are not quite comparable, because (i) she measures extensive margin by
the number of product variety, (ii) she controls for the effect of credit constraint on selection into
domestic market and (iii) she uses a different method to separate the effect between two margins
of trade. She was not able to test the second hypothesis, while I did.
98
by re-estimating Equation 3.3 with the initial value of private credit, i.e.,
PRIV CREDi,2003. Fixing private credit at its initial value does not affect
regression results (see Tables A14 to A17 in Appendix B).
Controlling for bias from state-owned enterprises
A major threat to the identification strategy here is the presence of state-
owned enterprises (SOEs). Enterprises are defined as SOEs if the state owns,
directly or indirectly, more than 50% of shares at the national or sub-national
level. SOEs are able to receive government financial support, which makes their
credit constraints very different from private firms. SOEs’ access to credit is not
captured by our measure of financial development - private credit to GDP, while
the contribution of SOEs in our export data could be substantial. Failing to
account for the presence of SOEs, therefore, would bias downward the effect of
private credit on export performance.
I will address this source of bias by identifying industries that typically
have large SOE presence and countries with large SOE presence, based on the
findings of Kowalski et al. (2013).10 Kowalski et al. (2013) identify economic
sectors and countries where SOEs are most prevalent, using financial information
from the world‘s largest 2000 public companies from Forbes Global 2000 list. The
10Data on SOEs’ shares by country and by sector is not available for the majority of my
sample. The most detailed dataset is provided by the OECD, which reports the size of national
SOE sectors in 2012, with a breakdown by broad sector and corporate form. Only seven out of 34
countries appear in EDD: Colombia, Spain, Mexico, Norway, Chile, Portugal, and Denmark. The
last three do not have data on SOEs in the manufacturing sector.
99
analysis also includes more than 330,000 domestic and foreign subsidiaries of these
companies, which span across 37 countries. The authors use data on shares of
sales, assets, and market values of each country’s top ten companies to construct
two indicators: country SOE shares (CSS) and sectoral SOE shares (SSS). CSS
evaluates the importance of the state in the country’s main international business
players, rather than in the country’s overall economic activities.
According to Kowalski et al. (2013), ten countries with highest CSS are
China, the United Arab Emirates, Russia, Indonesia, Malaysia, Saudi Arabia,
India, Brazil, Norway, and Thailand. Norway is the only country among these to
be in my sample. In terms of SSS, some manufacturing sectors with moderate SSS
are manufacture of fabricated metals, tobacco, basic metals, electrical equipment,
machinery, and transport equipment. These sectors make notable contribution in
the world merchandise trade. Their SSS are around 7-10%. Some services sectors
(telecommunications, banking, and engineering) have similar SSS values. All of
them seem less significant when compared to the energy and heavy industries
where SSS can reach up to 40%.
As a robustness exercise, I will drop the country and industries with high
presence of SOEs from the sample and re-estimate the main equation. The
Norwegian share of observations in my sample is around 2.5%. The manufacturing
sectors identified above correspond to Division 16, 27, 28, 29, 31, 34, and 35 in
Table 3. These sectors contribute to 40% the sample observations. Table A18
100
reports the regression results from the original and new samples. Results from
the new sample are weaker: estimates have smaller magnitudes and lower levels
of significance. However, we can still conclude that private credit directly and
indirectly benefits export performance, measured by total export sales and number
of firms.
Policy discussion
This research may be useful for designing policies as it provides a guideline
based on which policy makers can select the right industry and firms that optimize
the outcomes of financial support program. Policymakers can choose large,
continuing exporters, and subsidize their fixed trade costs. This would be a robust
policy recommendation if the causality from finance to export can be established
more carefully.
Several mechanisms undermine that causal relationship and have been
documented by firm-level research as follows.
The effects of financial development on exports could be biased downward
if the exporting firms do not rely much on borrowings in the private markets.
This description is true for two firm types: SOEs and foreign owned firms. We
have controlled for the former case but not the latter. Foreign owned firms are
less dependent on private credit due to intra-group lending mechanisms from
their affiliates. Manova et al. (2009) use Chinese firm-level data, and Alvarez and
101
Lopez (2012) use Chilean plant level to show that multinational affiliates and joint
ventures had better export performance than private domestic firms.
Another potential source of downward bias is some type of “learning from
exporting”. Incumbents can diversify financial resources and risks, as extended
overseas operations help them establish more ties with the local institutions.
They thus become less dependent on the domestic financial market. This view is
supported by Greenaway et al. (2007) who analyze data on British manufacturing
firms. This view is not shared by several other studies. Bellone et al. (2010), using
data from French manufacturing firms, argue that better financial health is not an
outcome of exports. Indeed, firms committing more to international trade appear
to be financially more constrained. Feenstra et al. (2009) also provide evidence
that Chinese manufacturing firms experience a tighter credit constraint as their
export share grows. Shipment time takes longer with more exports, increasing
firms’ financial need.
The effects of financial development could also be potentially biased upward.
Incumbents could have better access to finance, as they tend to be larger and
banks prefer larger firms. Due to information asymmetry, banks cannot observe
firms’ productivity. Banks opt to lend out more to large firms as they believe
these firms are more credit trustworthy. Banks could also prioritize public firms
for the same reason. In a study of a subsidized credit program in Pakistan, Zia
(2008) finds that even though the program created incentives for banks to lend to
102
smaller private firms, banks are still reluctant to do so. Banks end up providing
more loans to big firms than small ones, despite equal productivity.
The results in this chapter may best be interpreted as providing a strong
association rather than causality. It is important to include firm ownership and
firm productivity data that I do not have in future works on export and finance.
Conclusion
In this chapter, I show that financial development, measured by private
credit to GDP ratio, has differential effects on export performance. The effects
are stronger for the lowest income country group, for industries that require
more outside capital or have relatively less collateralizable assets, and for
continuing exporting firms. Among exporting firms, continuing exporters and
first-time exporters are the most productive types if using sales as a proxy for
firm productivity. First-time exporters rely relatively more on financial markets
to finance fixed trade costs compared to continuing exporters. This result implies
that the fixed cost firms pay to enter the export market for the first time is the
most challenging to finance, compared to other fixed costs incurred once they
survive in the market. This result highlights the importance of considering firm
productivity in answering which margin of trade is more sensitive to financial
development. This question is important to policy design: as the extensive margin
is associated with fixed trade cost and the intensive margin with variable trade
103
cost, a detailed understanding of how each margin reacts to finance would help
policymakers provide proper financial support to firms.
104
TABLE 4. Effects of Private Credit on Total Export Sales
Specification 1: With Country FE Specification 2: With Country controls
(1) (2) (3) (4) (5) (6) (7) (8)
VARIABLES Exporter Entrant Incumbent Survivor Exporter Entrant Incumbent Survivor
Log GDP 1.747*** 1.996*** 1.828*** 1.902*** 1.108*** 1.033*** 1.119*** 1.106***
(0.277) (0.358) (0.396) (0.453) (0.109) (0.093) (0.109) (0.110)
Private credit 0.295 0.780** 0.448 1.255*** 1.445*** 1.243*** 1.762*** 1.810***
(0.337) (0.368) (0.431) (0.429) (0.436) (0.335) (0.441) (0.416)
Private credit × Asset tangibility -2.026** -1.701*** -2.881*** -2.655*** -1.868** -1.572*** -2.669*** -2.571***
(0.833) (0.452) (0.873) (0.584) (0.817) (0.415) (0.848) (0.526)
Private credit × External finance 0.125** 0.051 0.176** 0.137** 0.135** 0.046 0.186*** 0.125**
(0.055) (0.047) (0.067) (0.068) (0.056) (0.044) (0.068) (0.054)
Legal rights 0.171*** 0.118*** 0.181*** 0.103***
(0.046) (0.030) (0.052) (0.035)
Log GDP per capita 0.600*** 0.194 0.791*** 0.347**
(0.168) (0.132) (0.192) (0.145)
Rule of law 0.267 0.223 0.229 0.099
(0.244) (0.206) (0.277) (0.234)
Constant -27.819*** -36.771*** -30.028*** -35.310*** -20.131*** -16.656*** -22.641*** -20.156***
(7.058) (9.088) (10.116) (11.564) (2.381) (1.543) (2.570) (2.128)
Observations 19,447 17,546 17,080 13,940 17,885 16,044 15,626 12,590
R-squared 0.727 0.673 0.686 0.617 0.684 0.640 0.642 0.587
Country FE Yes Yes Yes Yes No No No No
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
Sector FE Yes Yes Yes Yes Yes Yes Yes Yes
Results on dummies of legal origin (Socialist, French, British, Scandinavian are suppressed
Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1
105
TABLE 5. Effects of Private Credit on Extensive Margin or Firm Counts
Specification 1: With Country FE Specification 2: With Country controls
(1) (2) (3) (4) (5) (6) (7) (8)
VARIABLES Exporter Entrant Incumbent Survivor Exporter Entrant Incumbent Survivor
Log GDP 0.698*** 0.659** 0.686** 0.385 0.599*** 0.574*** 0.620*** 0.575***
(0.238) (0.308) (0.269) (0.267) (0.103) (0.105) (0.104) (0.115)
Private credit 0.309 0.329 0.328 0.611** 0.942** 0.853** 1.158*** 1.085***
(0.243) (0.331) (0.234) (0.261) (0.353) (0.373) (0.354) (0.394)
Private credit × Asset tangibility -0.923*** -0.865*** -1.231*** -1.323*** -0.834*** -0.783*** -1.140*** -1.205***
(0.279) (0.245) (0.339) (0.307) (0.262) (0.228) (0.325) (0.286)
Private credit × External finance 0.126*** 0.107** 0.152*** 0.132** 0.120*** 0.094** 0.150*** 0.123**
(0.037) (0.041) (0.040) (0.051) (0.037) (0.040) (0.040) (0.047)
Legal rights 0.154*** 0.136*** 0.155*** 0.108***
(0.038) (0.037) (0.040) (0.034)
LGDP per capita 0.462*** 0.445*** 0.582*** 0.521***
(0.166) (0.165) (0.174) (0.161)
Rule of law 0.098 0.081 0.048 0.008
(0.221) (0.227) (0.227) (0.228)
Constant -13.213** -12.774 -13.651* -6.978 -17.212*** -17.099*** -19.399*** -18.154***
(6.000) (7.773) (6.832) (6.786) (1.740) (1.749) (2.088) (2.226)
Observations 19,447 17,546 17,080 13,940 17,885 16,044 15,626 12,590
R-squared 0.868 0.864 0.847 0.831 0.770 0.755 0.752 0.734
Country FE Yes Yes Yes Yes No No No No
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes
Results on dummies of legal origin (Socialist, French, British, Scandinavian are suppressed
Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1
106
TABLE 6. Effects of Private Credit on Intensive Margin
or Average Sales per Firm
Specification 1: With Country FE Specification 2: With Country controls
(1) (2) (3) (4) (5) (6) (7) (8)
VARIABLES Exporter Entrant Incumbent Survivor Exporter Entrant Incumbent Survivor
Log GDP 1.049*** 1.337*** 1.142*** 1.517*** 0.510*** 0.459*** 0.500*** 0.531***
(0.231) (0.226) (0.327) (0.341) (0.078) (0.044) (0.077) (0.048)
Private credit -0.015 0.451* 0.119 0.644** 0.503 0.390** 0.604* 0.724***
(0.335) (0.226) (0.393) (0.314) (0.336) (0.172) (0.312) (0.195)
Private credit × Asset tangibility -1.103 -0.836** -1.650** -1.333*** -1.033 -0.789** -1.529** -1.366***
(0.720) (0.331) (0.705) (0.403) (0.702) (0.318) (0.681) (0.373)
Private credit × External finance -0.001 -0.056 0.024 0.005 0.015 -0.048 0.036 0.003
(0.047) (0.037) (0.051) (0.038) (0.048) (0.042) (0.053) (0.036)
Legal rights 0.017 -0.018 0.026 -0.006
(0.030) (0.029) (0.032) (0.028)
LGDP per capita 0.138 -0.251*** 0.209 -0.174*
(0.123) (0.087) (0.129) (0.096)
Rule of law 0.168 0.142 0.181 0.090
(0.148) (0.096) (0.148) (0.089)
Constant -14.606** -23.997*** -16.377* -28.332*** -2.920* 0.443 -3.242** -2.002*
(5.883) (5.734) (8.374) (8.710) (1.614) (0.902) (1.589) (1.087)
Observations 19,447 17,546 17,080 13,940 17,885 16,044 15,626 12,590
R-squared 0.502 0.393 0.443 0.348 0.450 0.348 0.393 0.316
Country FE Yes Yes Yes Yes No No No No
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes
Results on dummies of legal origin (Socialist, French, British, Scandinavian are suppressed
Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1
107
TABLE 7. Industries Most and Least Benefiting from Financial Development
Total export flow Extensive margins Intensive margins
Marginal Growth Marginal Growth Marginal Growth
ISIC ISIC ISIC
effect (%) effect (%) effect (%)
Most benefiting industries
242 1.567*** 3.79 242 1.172*** 2.23 323 0.429 0.54
(0.428) (0.417) (0.314)
323 1.478*** 3.38 323 1.049*** 1.85 242 0.395 0.48
(0.413) (0.384) (0.327)
300 1.259*** 2.52 359 0.912** 1.49 300 0.388 0.47
(0.394) (0.385) (0.295)
322 1.247*** 2.48 322 0.883** 1.42 181 0.388 0.47
(0.388) (0.363) (0.297)
331 1.208*** 2.35 300 0.871** 1.39 331 0.368 0.44
(0.388) (0.357) (0.289)
Least benefiting industries
153 0.728** 1.07 315 0.684 0.98 201 -0.247 -0.22
(0.362) (0.347) (0.401)
222 0.73** 1.08 221 0.684 0.98 210 -0.015 -0.01
(0.362) (0.337) (0.302)
243 0.735** 1.09 333 0.71 1.03 251 0.066 0.07
(0.369) (0.34) (0.279)
241 0.738** 1.09 191 0.717 1.05 171 0.085 0.09
(0.37) (0.351) (0.278)
271 0.752** 1.12 293 0.718 1.05 269 0.101 0.11
(0.372) (0.346) (0.274)
Note: Table generated from the three OLS regressions without origin country fixed
effects. The dependent variables in the three regressions are total export volume by
exporters, extensive margins of exporter, and intensive margins of exporters. The
marginal effects are calculated at means (MEM). Since the dependent variables are
logged level, marginal effects are transformed into growth rate (%) according to
exp(MEM)-1. The growth values are interpreted as the extent to which the dependent
variable changes given a one percentage point (0.01) change in private credit.
108
TABLE 8. Marginal Effects of Private Credit by Country
Private credit Marginal effects Growth rate (%)
Total Extensive Intensive Total Extensive Intensive
Percentile Value (%) Countries
Export Margins Margins export Margins Margins
1% 2.96 Timor Leste (TLS) 0.964*** 0.715** 0.249 1.623 1.045 0.282
(0.366) (0.354) (0.268)
5% 6.31 Guinea (GIN) 0.931** 0.702* 0.229 1.538 1.018 0.258
(0.365) (0.355) (0.267)
25% 14.78 Cambodia (KHM) 0.917** 0.691* 0.226 1.502 0.996 0.253
(0.364) (0.354) (0.267)
50% 23.85 Pakistan (PAK) 0.918** 0.688* 0.23 1.505 0.991 0.258
(0.364) (0.353) (0.267)
75% 41.26 El Salvador (SLV) 0.919** 0.689* 0.229 1.506 0.992 0.258
(0.364) (0.353) (0.267)
95% 141.78 Portugal (PRT) 0.918** 0.689* 0.23 1.505 0.991 0.258
(0.364) (0.353) (0.267)
99% 182.58 Denmark (DNK) 0.918** 0.69* 0.228 1.504 0.993 0.257
(0.364) (0.353) (0.267)
Note: Table generated from the three OLS regressions where country fixed effects are replaced with country controls and the dependent variables are
measured for generic exporters. The marginal effects are marginal effects at means (MEM). Since the dependent variables are logged level, marginal
effects are transformed into growth rate (%) according to (exp(MEM)-1). The growth values are interpreted as the extent to which the dependent
variable changes given a one percentage point (0.01) change in private credit. Standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1.
109
TABLE 9. Nested Regression: Marginal Effects of Private Credit
(1) (2) (3)
VARIABLES Total export sales Extensive margin Intensive margin
Regression with Exporters (all types)
All firm types 1.039*** 0.693* 0.346
(0.320) (0.355) (0.246)
Observations 12,375 12,375 12,375
Regression with two firm types
Incumbents 1.613*** 0.933*** 0.680***
(0.338) (0.362) (0.202)
Entrants 0.329 0.480 -0.152
(0.310) (0.366) (0.190)
Observations 24,750 24,750 24,750
Regression with three firm types
Incumbent 1.730*** 0.974*** 0.755***
(0.349) (0.369) (0.178)
Surviving entrant 0.926*** 0.670* 0.257*
(0.341) (0.377) (0.153)
Non surviving entrant 0.277 0.474 -0.197
(0.312) (0.372) (0.178)
Observations 37,027 37,027 37,027
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Note: These results are generated from a transformed version of the dataset
where different firm types (exporters, incumbents, entrants, non-surviving
entrants) are stacked on top of each other (the original data set was transformed
from wide to long form). The regression equation has country controls in place
of country fixed effects. Each section of this table reports results from a separate
regression. In the first section, regression is only conducted on Exporter type,
i.e., any firms exporting in period t. In the next two sections, the regression
has Incumbents as the reference group. The marginal effects among the firm
types (row) but within the same regression (column) are statistically different.
Regression results are reported in Tables A11, A12, and A13 in Appendix B.
110
CHAPTER IV
FINANCE AND EXPORT QUALITY UPGRADING
Introduction
Established work in international trade typically emphasizes export
performance along two margins: the extensive margin (number of exporting
firms) and the intensive margin (exporting size per firm). However, a country can
possibly achieve greater export sales by exporting higher quality products that
garner higher prices, without raising the number of exporting firms or exporting
quantities per firm. Export quality is a margin of trade yet to receive adequate
attention in the literature. Three questions arise regarding how to measure export
quality, what the determinants of export quality are, and what drives export
quality upgrading. This chapter will focus on the last question.
It is known from the literature on economic development that quality
upgrading accompanies economic growth. Over time, less developed countries
export products with better quality, and export more in industries that used to
be produced predominantly by developed countries (Hausmann et al., 2007, Sutton
and Trefler, 2011, IMF 2014). Some view the ability of a country to produce better
quality goods as necessary to export success and economic growth (Grossman
and Helpman, 1991). However, as quality upgrading requires better technology
and inputs (Kremer, 1993), it is a challenge for poor countries to accumulate
111
physical and human capital stock (Hausmann and Rodrik, 2003). In researching
the mechanisms of quality upgrading, this chapter can help us understand the
design of growth-promoting policies in developing countries.
It is worth bearing in mind that export quality and export quality upgrading
are not the same: the former refers to the level of export quality in absolute
terms and the latter to the improvement in quality ranking relative to a certain
threshold. Put differently, studying the determinants of export quality helps us
characterize the country or industry profiles that associate with better quality
exporting goods. Identifying the determinants of export quality upgrading tells
us the conditions under which a certain country or industry can reach the highest
level of export quality at a faster rate. So far, it is unclear whether export quality
and export quality upgrading are driven by similar mechanisms. For example,
while high-income countries export better products, the rate of quality upgrading
is less pronounced for them compared to less developed countries (International
Monetary Fund report, 2013).
In this chapter, I will explore how finance drives export quality upgrading
using the Export Diversification and Quality Database provided by the IMF. This
database is the most comprehensive to date, as it covers over 170 countries, 50
years, and six hundred product categories at its most disaggregated level. It is
especially suitable to study export quality upgrading across countries and over
time. In this database, annual export quality is calculated for each country-
112
product following Hallak (2006). Countries are then ranked by their export quality
for each product-year. The rank of a country is the ratio of its export quality
to the best export quality - which is taken as “the world frontier”. The growth
in the rank of a certain country-product measures its export quality upgrading.
The Database itself is a notable contribution since the main challenge to studying
export quality is a measurement of quality that is accurate and comparable across
countries and over time.
The connection between finance and export quality upgrading has not
been documented in depth, either theoretically or empirically. Based on existing
literature, we could propose a simple connection: export quality upgrading results
from innovation in production technology, and finance is conducive to innovation.
The international trade literature provides different explanations for why firms
innovate and upgrade product quality. Firms produce better products as they face
stronger market competition (Amiti and Khandelwal, 2013), or take advantage
of knowledge spillovers (Harding and Javorcik, 2011), or simply choose better
inputs (Verhoogen, 2008). While studies on export quality upgrading are scant,
the literature on how finance facilitates technological innovation is vast. Financial
systems can stimulate innovation-led growth through various mechanisms, as noted
in Beck and Levine (2005). In essence, a well-functioning financial system helps
ease the financial constraints that restrict entrepreneurs from fully engaging in the
113
process of innovation. Considering these two literature together, we might expect a
positive overall influence of financial development on export quality upgrading.
A closer examination of the existing literature suggests that the effect
of finance may be more complicated. As export quality upgrading can be a
direct outcome of technological innovation, similar to productivity growth, we
can follow the Schumpetarian view. According to the Schumpetarian growth
paradigm with financial constraint (Aghion, Howitt, and Mayer-Foulkes, 2005,
Beck and Levine 2005), how fast a country (or a sector) catches up with the global
technology frontier depends on its distance to the frontier and financial conditions
in the country. The country could enjoy an “advantage of backwardness”: which
means the further it is away from the frontier, the more easily it progresses
technologically by adopting new technologies from countries that are close to
the frontier. However, there is also empirical evidence on “disadvantage of
backwardness”: to catch up with the world frontier, firms in “backward” countries
need to pay an R&D cost, which is harder to finance under a weak financial system
and low labor productivity. These two conditions are prevalent in “backward”
countries. Beck and Levine (2005) conclude that rapid advancement for backward
countries occurs only when the country reaches a certain level of institutional
quality. A less developed financial system, therefore, can act as a source of
divergence. The same reasoning can hold true here, meaning that the effect of
finance on export quality upgrading is ambiguous.
114
We may also argue that finance is beneficial to export quality upgrading as
to export performance. Improvement in financial conditions has a direct effect
on exports: firms expand exporting activities because it is cheaper and easier
to finance these activities with outside capital. There is also an indirect and
asymmetric effects: export performance intensifies more markedly for industries
that are more dependent on outside capital. These industries experience an
increase in the number of exported products and the exported volume per product
(Manova, 2013), the number of exporting firms and export sales per firm (Duong,
2020), and average product quality (Crino and Ogliari, 2017). We can anticipate
that if certain conditions allow firms to export higher-quality goods, then these
conditions may also help the products of these firms advance faster towards the
world frontier.
To sum up, I explore the role of finance on export quality upgrading based
on established knowledge from two strands of literature - comparative advantage
and economic convergence. I will investigate the direct and indirect effect of
country-level financial conditions, measured by private credit to GDP ratio, FDI
net inflows to GDP ratio, and an index of domestic financial reform. Among the
three indicators, private credit to GDP ratio has been widely used in related
literature to proxy for domestic financial development. It reflects the ease with
which the private sector of an economy borrows from depository institutions.
Private credit does not account for any funding resources other than bank loans;
115
therefore, I include two other finance measures to address this shortcoming. FDI
not only brings in foreign capital, it may also bring foreign technologies that
associate with more sophisticated or higher quality products. Quality upgrading
in the host countries is made possible by two channels: (i) local firms learn about
new technologies from multinationals, and (ii) local firms in the supplying sector
have to innovate themselves to meet higher product standards from multinationals
(Harding and Javorcik, 2011). The last finance measure - financial reform index
- captures advancement of governance in various financial markets, including
domestic finance, banking, securities, and capital account. In my analysis, the
three indicators will be included in the specification as independent terms and as
interaction terms with initial export quality, and product characteristics (external
financial dependence and asset tangibility). The former way of interaction would
be inspired by research on economic convergence, and the latter way by studies in
comparative advantage.
I find that all three indicators of finance have nonlinear effects on quality
upgrading. The rate at which the quality of certain product catches up with the
world frontier depends negatively on the initial gap between the product and
the leading quality level. This rate will increase if the product is exported by a
country with better initial financial conditions. The role of finance is not the same
across country groups. FDI inflows have a stronger impact on quality upgrading of
upper-middle income countries, while financial reform can explain more of quality
116
upgrading in the highest and lowest income groups. The direct performance of
private credit is not precise when considered individually. Once controlling for
indirect channels, the estimates of private credit and its interaction terms are
significant with signs as expected. Both hypotheses mentioned above can be
confirmed for private credit. Only the first hypothesis can be confirmed for FDI
and reform index.
This chapter proceeds as follows: Section 2 discusses related literature.
Section 3 describes the data and methodology. Section 4 presents empirical results.
Section 5 concludes.
Related Literature
Some studies have shown the positive effect of FDI inflows on quality
upgrading (Henn et al., 2017, Harding and Javorcik, 2011), and one explanation
for such effect is the spillover of innovations from multinationals. Amiti and
Khandelwal (2013) study the evolution of export quality under domestic
competition - proxied by import tariffs. They build a theoretical model based
on the distance-to-the-frontier framework by Aghion et al. (2005, 2009). The
theoretical hypotheses are tested on U.S. import data. They find that countries
with poorer export quality have a slower upgrading rate as competition dampens
their incentives to innovate. While the authors do not directly mention financial
development, they find that the nonmonotonic relationship between import tariffs
and export quality upgrading only holds for countries that reach a certain level of
117
institutional quality, as measured by the “good business climate” indicator from
Doing Business Survey Database.
A model is “Schumpeterian” when it has profit-maximizing and product
obsolescence in innovation. The innovation, if successful, will give the entrepreneur
a monopoly on an improved production technology. Successful innovations make
existing technologies obsolete, and drive economic growth. Schumpeterian growth
models have been constructed for multi-country or firm-level frameworks. For
the version set in the latter context, innovations come from both entrants and
incumbents. Credit constraints can either increase innovation by eliminating
less efficient incumbents, or decrease innovation by preventing potentially good
innovators from entering the market.
There are different versions of Schumpetarian models, starting from Aghion
and Howitt (1998). A version of their model with credit constraint is discussed in
Beck and Levine (2005) while a newer version with market competition is adapted
in Amiti and Khandelwal (2013). Aghion, Howitt, Mayer-Foulkes (2005) present
a model of cross-country convergence with financial constraints with several
theoretical implications. First, financial development has a positive direct effect on
the economic convergence rate. Second, such an effect vanishes once the country
reaches a certain level of GDP per capita. In the empirical analysis the authors
use the distance in the growth rate of GDP per capita from the U.S.’s as the
dependent variable. The most important explanatory variable is the interaction
118
term between initial relative GDP per capita and private credit. Empirical results
show an insignificant point estimate of private credit, and a negative coefficient
for the interaction term. This implies that given the same income level, a more
developed financial system is associated with a slower speed of convergence. The
authors calculate the cutoff or critical level of GDP per capita above which the
role of private credit starts to diminish. They rule out other explanations for
economic divergence.
Data
Export Quality Database
Product quality itself is unobservable. Unit prices or unit values are
conventionally used as a proxy for product quality due to data availability and
ease of construction. However, unit prices are not an accurate measure of product
quality for several reasons. They may be influenced by factors unrelated to
product quality such as production costs, transportation costs, or consumer
preferences. Many authors, for example Khandelwal (2010), Hallak and Schott
(2011) and Feenstra and Romalis (2014), have proposed various methods to
construct a more accurate measure of export quality. These methods, while
arguably sophisticated, are inapplicable to most developing countries due to
lack of data. Henn et al. (2017) sidestep this limitation by building upon an
earlier method developed by Hallak (2006) to construct the IMF’s Export Quality
119
Database - the most extensive dataset of export quality in terms of countries,
sectors, and time periods. Its wide coverage allows for more in-depth analysis on
export quality of developing countries, and further facilitates the formation of
growth-promoting policies.
The methodology of Henn et al. (2017) requires three steps: estimation,
aggregation, and normalization. In the estimation step, the authors applied a two-
stage least squares (developed by Hallak, 2006) to UN-NBER trade data. The
estimation equations can be simplified as follows
( )
p = f 1mxt (θmxt, yxt, dmx )
Imports = f 2mxt δθmx, Kmxt, dmx, α
where m, x, t denote, respectively, importing country, exporting country, and year.
The first equation assumes unit price (p) depends on three factors: the unobserved
product quality (θ), exporter income per capita (yxt), and distance between
the trading partners (dmx). The second estimation stage involves a quality-
augmented gravity equation. It contains the set of standard “gravity” factors and
an interaction term of exporter-specific quality θmxt with the importer’s income
per capita. Parameter α denotes importer and exporter fixed effects. Parameter
δ reflects the importers’ taste for quality. Combining these two equations yields a
120
raw estimate of export quality
( )
Qualitymxt = δlnθmxt = f
3 pmxt, yxt, dmx
In this estimate, product quality and importers’ taste for quality are non-
separable. The functional form of the quality estimate indicates that quality still
depends on unit prices, but not on production costs (proxied by yxt) and selection
bias (proxied by dmx). Selection bias happens when exporters choose to ship
products with higher quality to further destinations.1
After the estimation step, the authors normalize the quality indices by the
90th percentile of the cross country distribution in the relevant product-year.
The quality estimates are aggregated using current trade values as weights, first
consolidated across importers to obtain aggregated measures at the SITC 4-digit
exporting country-year level, then further aggregations provide estimates at 3-, 2-
or 1-digit, respectively, and at the country level.2 Normalization is repeated for
each aggregation step. The quality index of a product category in a country thus
depends on the distribution of export quality values of all other countries for that
category. It is impossible for users to derive the quality value of a product category
using quality values of its subcategories.
1Henn et al. (2017) make a detailed comparison between this quality estimate and unit prices.
2Table A20 in Appendix C shows an example of the SITC classification.
121
Due to normalization, the resultant quality indices do not reflect the absolute
quality level of an exporting category in a given year. Instead, it should be best
translated as the inverse of the distance of a country to the world frontier (the 90th
percentile) regarding a given exporting good. The quality indices are comparable
across countries within a year, and across years for one specific country. A higher
quality index implies a quality closer to the best quality available. The dependent
variable in this chapter is the growth rate of the quality indices calculated for a
specific country-product pair. It is the rate at which the quality of this country’s
product catches up with that of the world’s best product.
This quality database, despite its wide coverage, is not convenient to
analyze in conjunction with other types of data. The main limitation is that the
database follows SITC Rev. 1, an outdated and more generic product classification
system that has very little concordance with recent systems. Conversion among
the systems is still possible, yet involving multiple steps that result in greater
matching error and information loss. This is an issued directly faced by the
analysis in this chapter. In particular, several SITC-2 product categories
(correspondingly, around 30 SITC-4 categories) are matched to the same ISIC
Rev.2 2-digit category for data on financial characteristics (i.e., external finance
dependence and asset tangibility). The cross-product variation in financial features
is effectively reduced, which lessens the explanatory power of these product-level
financial features.
122
After combining with data on explanatory variables, the final sample has
product quality estimates for 87 countries, 520 four-digit SITC product categories,
43 two-digit SITC categories, and 36 years (1976-2010). Half of the countries
are in the middle income group, and one-third are high-income. There are seven
five-year non-overlapping periods: 1976-1980, 1981-1985, 1986-1990, 1991-1995,
1996-2000, 2001-2005, and 2006-2010. Summary statistics of export quality
upgrading are reported in Table A19 for the whole sample, and in Table A21 for
the subsamples divided by the country groups.
Figures 9 and 10 do not show any clear pattern in the relationship between
the growth of export quality and private credit. Figure 11 display a clear negative
relationship between initial export quality and quality growth. The slopes of the
fitting lines are different among country groups, with the steepest slope observed
for the high-income group.
FIGURE 9. Private Credit and Quality Growth
123
FIGURE 10. Cross-sectional Private Credit and Quality Growth
FIGURE 11. Initial Quality and Quality Growth
124
Quality data is widely disperse as seen in 9. For later analysis, I remove
outliers from the original dataset by truncating the top and bottom 1%. Only
quality estimates between -0.316 and 0.321 are kept.
Country-level measures
The country-level controls I use in my study are GDP per capita, private
credit to GDP ratio, FDI net inflows as a ratio of GDP, and an index of financial
reform. The last three are different indicators of financial conditions. Private
credit is the variable of interest here. It has been used extensively in related
research. A higher value of private credit indicates more credit issued by
depository institutions to the private sector relative to GDP. Private credit signals
the capability of the domestic financial market in channeling capital to firms
through financial institutions. Among other funding channels of firms, FDI inflows
pertain more to this analysis. Along with private credit and GDP per capita,
FDI inflows data are obtained from the World Development Indicator database
by the World Bank. The final financial indicator is an index of domestic financial
reform. This is a normalized index taken from the Database of Financial Reforms
(developed by Abiad et al. (2010) and maintained by the IMF). The index of
financial reform accounts for advancement in various aspects of the domestic
financial markets: controls of interest rate, reserve requirement, entry barriers,
state ownership in the banking sector, capital account restrictions, supervision of
the banking sector and securities market. It is expected to be more comprehensive
125
than private credit and FDI inflows. Table A22 shows the correlation matrix
of these country-level factors. Overall, the correlation among all factors are
moderately to weakly positive.
Figure 12 plots initial conditions of private credit and export quality, both
averaging over the sample period. Country groups noticeably distinguish by the
degree of export quality. While the fitting line has low standard errors for the
high-income group, more noises and outliers are observed for other income groups,
especially for the low-income.
FIGURE 12. Cross-section of Initial Private Credit and Initial Quality
Product-level measures
Without particular measures for product characteristics, I use measures
of production technology by sectors instead. The sectoral measures (external
126
financial dependence and asset tangibility) are taken from Braun (2003) and are
available at ISIC Revision 2. They are calculated from COMPUSTAT, a database
of accounting information of public firms in the United States. Each measure
is an average for all countries in a 10-year period from 1980 to 1989, and the
sectoral value is the median of these average values. The calculation is based on
the seminal work by Rajan and Zingales (1998). I combine concordance tables
from Haveman and UNSTAT to match the ISIC industry data with SITC export
data. External financial dependence and asset tangibility are barely correlated (the
correlation coefficient is -0.12).
Including these variables in the analysis requires several assumptions to be
met (Rajan and Zingales, 1998): the financial structures of public firms in the U.S.
represent the optimal structure for firms globally, and the rank of financial need of
industries is assumed to preserve across country and time. Implementation of this
method requires excluding the U.S. from the dataset, and the sampling period
should be later than the period from which these financial need measures are
calculated. In particular, if external financial dependence and asset tangibility are
calculated from 1980-1989 data, then the sampling period should start from 1990
at the soonest. Following this practice means truncating 20 years of the quality
dataset. Doing so weakens the power of the estimation methods, even when I try
to compensate for the loss of observations by reducing the interval length from five
to three or two years. By favoring to retain the whole quality data set, I make a
127
stronger assumption than Rajan and Zingales (1998): I assume that these financial
measures are the optimal of worldwide production technology, and their rank stays
fixed for the whole length of the sampling period (35 years). Making a stronger
assumption potentially weakens the preciseness of the empirical results.
Methodology
The main specification is
Yikt =β1Initial Qualityik,t−4 + β2Xi,t−4+
β3Initial Qualityik,t−4 ×Xi,t−4 + β4X − (4.1)i,t 4 ∗ Zk+
αik + αkt + ikt
where i, k, t represent exporting country, export product category, and year,
respectively. The dependent variable, Yikt, is the log growth rate of export quality
of product k from country i between year t and t− 4
( Qualityik,t )
Yikt = ln
Qualityik,t−4
Among the main regressors, Xi,t−4 are the initial values of the country-level
factors: log GDP per capita, private credit to GDP ratio, domestic financial reform
index, FDI net inflows to GDP ratio. These regressors are measured in the initial
year of the five-year non-overlapping interval (year t − 4). Zk are product or
128
industry-level factors: asset tangibility and external financial dependence. Fixed
effects are denoted by α.3
The basics of my estimation follow the setup of Henn et al. (2017) and
Can and Gorgoz (2018), the only two papers that have explored this dataset
so far.4 The dependent variable is log growth rate of product quality which is
measured over a non-overlapping period of a certain length (10 years in Henn
et al., 2017; and five years in Can and Gorgoz, 2018). The basic explanatory
variables are initial product quality and log GDP per capita, all measured at
the first year of each non-overlapping period. Both mentioned studies show that
initial proximity to the world quality frontier is the most important determinant
of quality upgrading. Specifically, products near the bottom of the quality ladder
(defined as the difference between the highest and lowest quality, Grossman and
Helpman, 1991) will move faster towards the world frontier. GDP per capita is the
second most important driver of export quality upgrading.
The length of the non-overlapping period is fixed at five years. It should
not be shorter due to a potential increase in measurement error bias (Henn et al.,
2017). Macroeconomic data are often measured with error, due to limited data
3In an additional exercise, I explore the role of the initial level of export quality of two-digit
SITC categories. Preliminary analysis yield a positive and significant estimate for SITC-2 initial
export quality, suggesting that quality upgrading is faster for a SITC-4 product when it belongs
to a high-quality SITC category.
4Henn et al. (2017) are the authors of the EDD. Their paper discusses different attributes of
the EDD, and presents a simple panel analysis where export quality upgrading depends on initial
quality and various country-level measures such as GDP per capita, institutional quality, reform
indices, etc. They use quality data reported in 3-digit Broad Economic Categories classification.
Can and Gorgoz (2018) examine the relationship between product diversification (which is also
an attribute of EDD) and overall export quality upgrading.
129
sources (Chang and Li, 2018). The problems of measurement error worsens with
taking first differences, as doing so reduces the meaningful signal and increases
the unwanted noise. Widening the differencing intervals (from five or 10 to 20
years) may alleviate this problem (Krueger and Mikael, 2001). In this chapter,
measurement error will not be a serious issue because we only take differences of
the dependent variable. While measurement error in the dependent variable does
not bias the coefficients, it can make them less precise. Setting the interval at five
years ensures sufficient treatment towards measurement error without risking too
many observations.
The coefficients of interest here are β2, β3 and β4. Consistent with previous
research, we expect β2 to be positive, as GDP per capita and financial conditions
should promote export quality upgrading. Should the effects of the country-level
conditions be nonmonotonic, that will show in β3. A negative β3 implies that more
private credit helps low-quality products upgrade disproportionately faster. Lastly,
β4 is expected to be positive for the interaction term of country-level financial
measures with external financial dependence and negative for the interaction term
with asset tangibility.
The set of fixed effects I employ will follow Can and Gorgoz (2018) and
Henn et al. (2017). Both studies control for unobservable heterogeneity along
the country-product and product-year dimensions. Henn et al. (2017) especially
emphasize the importance of the product-year heterogeneity. Therefore, while I
130
try estimating the equation with different sets of fixed effects (individual country
and product fixed effects or interacting the two), I keep year-product fixed effects
in all regressions.5 This type of fixed effects is crucial because: (i) quality values
are estimated separately for each product category, so quality values are only
comparable within category, and (ii) there are unobservable shocks or changes
in production technology that affects all categories worldwide (Henn et al.,
2017; Amiti and Khandelwal, 2013). All regressions are estimated with OLS and
standard errors are clustered at the country level.
Empirical Results
This section presents empirical results from seven tables. In Table 10, I
compare the first-order effects of the three finance measures. I also evaluate their
performance when different combinations of them are included in one regression. I
construct Tables 11 to 13 to examine whether the effect of each individual finance
measure differ by country groups. I repeat this exercise for the full specification as
in Equation 4.1, and report the results in Tables 14, 15 and 16.
All regressions include the initial level of export quality and GDP per capita.
Initial level of export quality is the most important determinant of export quality
upgrading. The estimates of this factor are negative and their magnitudes are the
highest among all regressors. This result is consistent with Henn et al. (2017),
5The results are indeed much weaker when I estimate the specification where the country-
product fixed effect are separated into a country fixed effect and a product fixed effect.
131
Amiti and Khandelwal (2013), and Can and Gozgor (2018) despite different export
quality measures.6 Another result that is consistent with these previous studies
is the estimates of GDP per capita, which are always positive and significant.
We can conclude that export quality upgrading is faster for products with lower
initial levels of export quality, and for products exported by a country with better
economic development
In the baseline estimation using the whole sample (Table 10), the estimates
of private credit are negative and highly significant. Estimates of FDI net inflows
are also negative with varying degree of significance. FDI net inflows indicator
is significant when being estimated alone, or in conjunction with private credit.
It loses significance when the index of domestic financial reform is introduced
into the regression. This result suggests a potential multicollinearity between two
indicators. The estimates of financial reform index, on the contrary, is positive and
highly significant across all regressions. As these results are somewhat counter-
intuitive for private credit and FDI inflows, I separate the sample into four country
groups. I start from the simplest form of specification that has only initial quality,
GDP per capita, and one financial indicator. I gradually add the interaction terms
of the financial indicator with product characteristics.
6Henn et al. (2017) and Can and Gozgor (2018) employ the same IMF database, but different
indicators and frequency. Henn et al. (2017) analyze data on 3-digit Broad Economic Category
products and calculate export quality growth over 10-year non-overlapping periods. Can and
Gozgor (2018) use the 1-digit SITC product classification and five-year non-overlapping intervals.
Amiti and Khandelwal (2013) use 10-digit HS varieties and five-year intervals.
132
Effects of Private credit
The estimates of private credit do not change their negative sign when I
regress by country groups (Table 11). They only become insignificant with a
much smaller magnitude. There are two possible explanations of these results:
either private credit has little variation within each country group and/or there
are other channels through which private credit impacts quality upgrading that I
have not accounted for. Results in Table 14 support both explanations. Private
credit becomes positive and significant when all of the necessary interaction terms
are included. The interaction term between private credit and initial quality is
negative, suggesting that private credit is more important for sectors with initially
low quality. The interaction term between private credit and sectoral financial
conditions bear expected signs. Export quality upgrading is therefore faster in
more financially vulnerable sectors.
Some interesting patterns appear when estimation results are compared
across country groups. The direct effects and indirect effects through initial quality
are stronger for lower income countries, especially the low and lower-middle income
groups. The interaction term with external finance is technically zero for all four
sub-samples. The interaction term with asset tangibility is also zero for the high
income group, but statistically significant for other groups. Such differences require
a closer examination on the exporting pattern of the country groups. It is possible
133
that there is little variation in the degree of tangible assets implied in the products
of the high income.
Effects of FDI
The overall effect of FDI on quality upgrading is negative and significant,
which is mostly driven by the high income countries (Table 12). The estimate of
FDI is zero for the high income group, but positive for the others. Its magnitude is
especially large for the upper-middle group, and fairly large for the low and lower-
middle group (approximately the same magnitude with GDP per capita). For
both the highest and lowest income groups, FDI is not significant. It is possible
that FDI is not truly meaningful for the high income - they have the ability to
innovate on their own, without relying on new technologies introduced by FDI.
They may not find it necessary to adopt new technologies from FDI since their
technologies are already the most advanced. For the lowest income group, the
imprecise estimate of FDI suggests that some but not all of them benefit from
FDI. The difference between the lowest income and the upper-middle income could
be explained by other underlying factors that are known to influence the rate of
economic convergence such as institutional quality and human capital.
Similar to private credit, FDI has a stronger effect for sectors with initially
low quality (Table 15). The indirect effect through the initial quality channel is
also stronger for the upper-middle income group. The indirect effects through
sectoral financial characteristics display somewhat random pattern. These
134
interaction terms are probably misspecified: FDI and private credit are completely
different funding resources. Firms choose to apply for credit from financial
institutions depending on their need for external finance and their endowment of
tangible asset. Meanwhile, FDI is initiated by the parent company without regards
to the financial characteristics of the subsidiary firms.
Effects of Financial reform
Domestic financial reform is meaningful to export quality upgrading on
average (Table 13). Its effects are not monotonic with income level: stronger
and significant effects are observed in the highest and lowest income group. The
effect is basically zero for the upper-middle income group. Interpreting these
results require a closer look at the trend of financial reform in each country group.
Financial reform is conducive to quality upgrading, and especially more for lower
quality products (Table 16). It is unclear whether the effects of financial reform is
stronger in financially vulnerable sectors.
Conclusion
Recent research in international trade has shifted focus to export quality,
another margin of trade besides the traditional extensive and intensive margins.
The growth in export quality, or export quality upgrading, has important
implications for economic development, however, not much research attempt
has been done on this subject. In this chapter, I aim to fill this research gap by
135
analyzing the role of domestic financial conditions on export quality upgrading.
I use three financial indicators: private credit to GDP ratio, FDI net inflows to
GDP ratio, and a domestic financial reform index. Since there is a lack of research
on the topic, my approach is guided by established understanding of how finance
influences economic convergence and comparative advantage. In particular, I
investigate whether finance has a nonlinear effect on export quality upgrading,
and whether finance affects export quality upgrading in a similar fashion as it does
export volume.
I conduct an empirical analysis using data on export quality of SITC 4-digit
products from the IMF. I estimate a specification where export quality upgrading
is explained by initial values of export quality, GDP per capita, one indicator
of country-level finance, and three terms where the financial measure interacts
with initial quality, external financial dependence, and asset tangibility. In the
estimation, I control for heterogeneity along country-product and product-time
dimensions. Results from the full specification suggest that all three country-
wide measures of finance positively impact quality upgrading. Their effects are
disproportionately stronger for products with lower values of initial quality. Better
financial conditions at the macroeconomic level allows firms to invest more in the
costly process of innovation, or to be able to approach and adopt more advanced
technologies. I find that, for the case of private credit, quality upgrading also
occurs more rapidly in sectors that require more external finance or have little
136
tangible assets. Both hypotheses are hence confirmed for this indicator, while only
the first hypothesis is confirmed for FDI inflows and financial reform.
I further examine whether the effects of finance are similar for countries of
different income levels. It is possible that firms finance their innovative process
differently depending on the development stage of their home country. Estimation
results on the first-order effects show a clear pattern of differences across country
groups for FDI inflows and financial reform. FDI inflows appear to be more
important to the upgrading process for upper-middle income countries, while
financial reform is more important for the highest and lowest income groups. The
patterns become less clear when I consider the indirect effects of these indicators,
through the channels of initial quality and financial characteristics of the products.
A concrete explanation for such results would require a comprehensive analysis
using these indicators, exporting pattern by country groups, and other structural
factors. This is the main limitation of the analysis in this chapter: it needs a
stronger argument, either theoretically or empirically, as to why these channels,
especially the financial characteristics, are relevant to FDI and financial reform.
There might be other aspects of the production technologies, such as R&D
intensity, that better correspond to FDI inflows indicator.
The scope for future work is wide. One reason why the lowest income
country group does not benefit from FDI as much is their low level of human
capital. The role of human capital in helping countries catch up with the world
137
frontier has been discussed in Benhabib and Spiegel (2005). Less developed
countries are introduced to advanced technologies through FDI. However, these
countries would not be able to catch up with developed nations if they lack
sufficient human capital to retain and incorporate new technologies into their
production. By adapting this idea to the research at hand, it will be interesting
to calculate the critical level of human capital above which quality upgrading
of a nation can converge to the world frontier. Another avenue to extend this
analysis is product differentiation. The effects of finance could be stronger for
differentiated products, where firms have more scope to adjust quality (Crino and
Ogliari, 2017). There are three methods to define product differentiation: quality
dispersion (Fan et al., 2015), quality ladder (Khandelwal, 2010), differentiation
index (Rauch, 1999). These methods can be easily applied to the research in this
chapter: the first method uses the standard deviation in quality, the second is the
gap between the maximum and minimum quality, and the last is a index available
in SITC Revision 2 - the classification system closest to SITC Revision 1 as of my
knowledge.
138
TABLE 10. Finance Measures and Quality Upgrading
Dependent variable:
Quality Growth at SITC 4-digit
(1) (2) (3) (4) (5) (6) (7)
Initial Quality −0.577∗∗∗ −0.577∗∗∗ −0.577∗∗∗ −0.577∗∗∗ −0.577∗∗∗ −0.577∗∗∗ −0.577∗∗∗
(0.015) (0.015) (0.015) (0.015) (0.015) (0.015) (0.015)
Initial (log) GDP per capita 0.032∗∗∗ 0.032∗∗∗ 0.032∗∗∗ 0.029∗∗∗ 0.028∗∗∗ 0.029∗∗∗ 0.032∗∗∗
(0.005) (0.005) (0.005) (0.005) (0.004) (0.005) (0.005)
Initial Private credit −0.011∗∗∗ −0.011∗∗∗ −0.011∗∗∗ −0.011∗∗∗
(0.003) (0.003) (0.003) (0.003)
Initial FDI net inflows −0.015∗ −0.018∗ −0.015 −0.012
(0.008) (0.010) (0.009) (0.008)
Initial Domestic financial reform 0.013∗∗∗ 0.014∗∗∗ 0.013∗∗∗ 0.013∗∗∗
(0.004) (0.004) (0.004) (0.004)
Observations 132,481 132,481 132,481 132,481 132,481 132,481 132,481
R2 0.517 0.517 0.517 0.516 0.517 0.517 0.517
Adjusted R2 0.361 0.362 0.362 0.361 0.361 0.361 0.362
Country-Product FE Yes Yes Yes Yes Yes Yes Yes
Year-Product FE Yes Yes Yes Yes Yes Yes Yes
Robust standard errors clustered at the country level
∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
139
TABLE 11. Private Credit and Quality Upgrading by Income Group: Baseline
Dependent variable:
Quality Growth at SITC 4-digit
All H (UM+LM+L) UM (LM+L)
Initial Quality −0.577∗∗∗ −0.639∗∗∗ −0.560∗∗∗ −0.610∗∗∗ −0.526∗∗∗
(0.015) (0.023) (0.019) (0.027) (0.032)
Initial (log) GDP per capita 0.032∗∗∗ 0.030∗∗∗ 0.038∗∗∗ 0.037∗∗∗ 0.044∗∗∗
(0.005) (0.005) (0.007) (0.011) (0.010)
Initial Private credit −0.011∗∗∗ −0.005 −0.010 −0.008 −0.006
(0.003) (0.003) (0.011) (0.013) (0.010)
Observations 132,481 68,764 63,717 38,895 24,822
R2 0.517 0.530 0.568 0.600 0.634
Adjusted R2 0.361 0.390 0.380 0.419 0.398
Country-Product FE Yes Yes Yes Yes Yes
Year-Product FE Yes Yes Yes Yes Yes
Robust standard errors clustered at the country level
∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
Notes: Tables 11 to 13 report results from estimation on sub-samples of country
groups. Bases on country classification of the World bank in 2011, the sample is first
divided into two groups: High-income countries (denoted H) and other (UM+LM+L).
The latter is further divided into the Upper-middle group (UM) and a group summing
up Lower-middle and Low income (LM+L). Low and lower-middle income countries
are not separated, due to the small share in observations of the low income group. In
the sample, high-income countries occupy almost half of the observations, while the
share of upper-middle countries is about one-fourth. Diving the sample in two steps as
described enable more precise cross-group comparison.
140
TABLE 12. FDI and Quality Upgrading by Income Group: Baseline
Dependent variable:
Quality Growth at SITC 4-digit
All H (UM+LM+L) UM (LM+L)
Initial Quality −0.577∗∗∗ −0.639∗∗∗ −0.561∗∗∗ −0.610∗∗∗ −0.526∗∗∗
(0.015) (0.023) (0.019) (0.026) (0.032)
Initial (log) GDP per capita 0.029∗∗∗ 0.030∗∗∗ 0.033∗∗∗ 0.033∗∗∗ 0.042∗∗∗
(0.005) (0.005) (0.007) (0.010) (0.009)
Initial FDI inflows to GDP −0.018∗ −0.008 0.063∗∗ 0.080∗∗ 0.044
(0.010) (0.007) (0.029) (0.035) (0.032)
Observations 132,481 68,764 63,717 38,895 24,822
R2 0.516 0.530 0.568 0.600 0.634
Adjusted R2 0.361 0.390 0.381 0.419 0.398
Country-Product FE Yes Yes Yes Yes Yes
Year-Product FE Yes Yes Yes Yes Yes
Robust standard errors clustered at the country level
∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
TABLE 13. Financial Reform and Quality Upgrading
by Income Group: Baseline
Dependent variable:
Quality Growth at SITC 4-digit
All H (UM+LM+L) UM (LM+L)
Initial Quality −0.577∗∗∗ −0.640∗∗∗ −0.560∗∗∗ −0.610∗∗∗ −0.526∗∗∗
(0.015) (0.023) (0.019) (0.027) (0.032)
Initial (log) GDP per capita 0.028∗∗∗ 0.031∗∗∗ 0.034∗∗∗ 0.034∗∗∗ 0.040∗∗∗
(0.004) (0.005) (0.007) (0.010) (0.009)
Financial reform index 0.014∗∗∗ 0.010∗∗∗ 0.013 0.001 0.031∗∗∗
(0.004) (0.004) (0.009) (0.011) (0.010)
Observations 132,481 68,764 63,717 38,895 24,822
R2 0.517 0.530 0.568 0.600 0.635
Adjusted R2 0.361 0.390 0.381 0.419 0.399
Country-Product FE Yes Yes Yes Yes Yes
Year-Product FE Yes Yes Yes Yes Yes
Robust standard errors clustered at the country level
∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
141
TABLE 14. Private Credit and Quality Upgrading by Income Group
Dependent variable:
Quality Growth at SITC 4-digit
All All All All H (UM+LM+L) UM (LM+L)
Init. Quality −0.577∗∗∗ −0.516∗∗∗ −0.577∗∗∗ −0.515∗∗∗ −0.562∗∗∗ −0.519∗∗∗ −0.571∗∗∗ −0.481∗∗∗
(0.015) (0.020) (0.015) (0.020) (0.030) (0.023) (0.034) (0.037)
Init. LGDP per capita 0.032∗∗∗ 0.028∗∗∗ 0.031∗∗∗ 0.028∗∗∗ 0.028∗∗∗ 0.039∗∗∗ 0.039∗∗∗ 0.044∗∗∗
(0.005) (0.004) (0.005) (0.004) (0.004) (0.007) (0.011) (0.010)
Init. Private credit to GDP −0.011∗∗∗ 0.169∗∗∗ −0.009∗ 0.177∗∗∗ 0.135∗∗∗ 0.212∗∗∗ 0.139∗∗ 0.420∗∗∗
(0.003) (0.029) (0.005) (0.029) (0.027) (0.058) (0.055) (0.130)
Init. Priv. cred. × Init. Quality −0.193∗∗∗ −0.195∗∗∗ −0.147∗∗∗ −0.240∗∗∗ −0.161∗∗∗ −0.481∗∗∗
(0.032) (0.031) (0.029) (0.064) (0.056) (0.158)
Init. Priv. cred. × External finance 0.005 0.007∗∗ 0.005 0.003 0.006 −0.003
(0.003) (0.003) (0.003) (0.008) (0.010) (0.014)
Init. Priv. cred. × Asset tangibility −0.014 −0.028∗∗ −0.004 −0.073∗∗∗ −0.042∗ −0.199∗∗∗
(0.009) (0.011) (0.011) (0.019) (0.021) (0.046)
Observations 132,481 132,481 132,481 132,481 68,764 63,717 38,895 24,822
R2 0.517 0.523 0.517 0.523 0.533 0.572 0.602 0.639
Adjusted R2 0.361 0.369 0.361 0.369 0.395 0.385 0.421 0.405
Country-Product FE Yes Yes Yes Yes Yes Yes Yes Yes
Year-Product FE Yes Yes Yes Yes Yes Yes Yes Yes
Robust standard errors clustered at the country level
∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
142
TABLE 15. FDI Net Inflows and Quality Upgrading by Income Group
Dependent variable:
Quality Growth at SITC 4-digit
All All All All H (UM+LM+L) UM (LM+L)
Init. Quality −0.577∗∗∗ −0.565∗∗∗ −0.577∗∗∗ −0.565∗∗∗ −0.631∗∗∗ −0.545∗∗∗ −0.584∗∗∗ −0.522∗∗∗
(0.015) (0.016) (0.015) (0.016) (0.024) (0.020) (0.029) (0.032)
Init. LGDP per capita 0.029∗∗∗ 0.029∗∗∗ 0.029∗∗∗ 0.029∗∗∗ 0.030∗∗∗ 0.034∗∗∗ 0.033∗∗∗ 0.043∗∗∗
(0.005) (0.004) (0.005) (0.004) (0.005) (0.007) (0.010) (0.009)
Init. FDI inflows to GDP −0.018∗ 0.644∗∗∗ −0.018 0.644∗∗∗ 0.247∗∗ 1.429∗∗∗ 1.792∗∗∗ 0.796
(0.010) (0.169) (0.013) (0.168) (0.118) (0.370) (0.441) (0.540)
Init. FDI × Init. Quality −0.699∗∗∗ −0.699∗∗∗ −0.264∗∗ −1.677∗∗∗ −2.099∗∗∗ −0.969
(0.183) (0.182) (0.127) (0.436) (0.513) (0.640)
Init. FDI × External finance 0.018 0.018 0.010 0.044 0.018 0.141∗∗
(0.013) (0.018) (0.017) (0.041) (0.055) (0.059)
Init. FDI × Asset tangibility −0.019 −0.017 −0.012 −0.032 0.063 −0.122
(0.024) (0.035) (0.026) (0.089) (0.086) (0.235)
Observations 132,481 132,481 132,481 132,481 68,764 63,717 38,895 24,822
R2 0.516 0.519 0.516 0.519 0.530 0.571 0.604 0.635
Adjusted R2 0.361 0.364 0.361 0.364 0.391 0.384 0.424 0.398
Country-Product FE Yes Yes Yes Yes Yes Yes Yes Yes
Year-Product FE Yes Yes Yes Yes Yes Yes Yes Yes
Robust standard errors clustered at the country level
∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
143
TABLE 16. Financial Reform and Quality Upgrading by Income Group
Dependent variable:
Quality Growth at SITC 4-digit
All All All All H (UM+LM+L) UM (LM+L)
Init. Quality −0.577∗∗∗ −0.488∗∗∗ −0.577∗∗∗ −0.488∗∗∗ −0.501∗∗∗ −0.466∗∗∗ −0.472∗∗∗ −0.436∗∗∗
(0.015) (0.018) (0.015) (0.018) (0.036) (0.022) (0.033) (0.032)
Init. LGDP per capita 0.028∗∗∗ 0.032∗∗∗ 0.028∗∗∗ 0.032∗∗∗ 0.025∗∗∗ 0.034∗∗∗ 0.032∗∗∗ 0.040∗∗∗
(0.004) (0.004) (0.004) (0.004) (0.004) (0.006) (0.008) (0.008)
Init. Reform index 0.014∗∗∗ 0.238∗∗∗ 0.007 0.234∗∗∗ 0.241∗∗∗ 0.331∗∗∗ 0.406∗∗∗ 0.399∗∗∗
(0.004) (0.023) (0.006) (0.023) (0.040) (0.033) (0.050) (0.052)
Init. Reform × Init. Quality −0.260∗∗∗ −0.260∗∗∗ −0.267∗∗∗ −0.392∗∗∗ −0.474∗∗∗ −0.456∗∗∗
(0.027) (0.026) (0.045) (0.038) (0.056) (0.060)
Init. Reform × External finance 0.004 0.012∗∗ 0.020∗ 0.012 0.004 0.025∗
(0.006) (0.006) (0.011) (0.008) (0.009) (0.015)
Init. Reform × Asset Tangibility 0.019 0.001 0.024∗ −0.004 −0.010 −0.058
(0.014) (0.012) (0.013) (0.021) (0.024) (0.037)
Observations 132,481 132,481 132,481 132,481 68,764 63,717 38,895 24,822
R2 0.517 0.530 0.517 0.530 0.538 0.583 0.615 0.648
Adjusted R2 0.361 0.379 0.361 0.379 0.401 0.401 0.440 0.420
Country-Product FE Yes Yes Yes Yes Yes Yes Yes Yes
Year-Product FE Yes Yes Yes Yes Yes Yes Yes Yes
Robust standard errors clustered at the country level
∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
144
CHAPTER V
CONCLUSION
This dissertation employs a variety of modeling techniques and databases
to shed more light on the role of financial and labor frictions on macroeconomic
outcomes. In Chapter II, the outcomes are employment status and labor income.
In Chapters III and IV, the outcomes are three margins of exports: firm counts
(extensive margin), firm size (intensive margin), and product quality.
Chapter II takes a unique view to exploring informality in the labor market,
using an equilibrium search model with heterogeneous labor productivity. In
the model, high-skilled employers can voluntarily leave formal firms to become
entrepreneurs in the informal sector. Working for high-skilled entrepreneurs is the
only occupational choice of low-skilled labor. Given this setting, dynamics in the
informal labor sector can distort the formal sector’s signal of efficiency, which is
reflected in the Beveridge curve. The model is solved numerically and calibrated
to Latin American countries. It has a single, unique equilibrium if formal firms
react more strongly to future loss from a job match rather than future gain.
The government can reduce informal employment with a budget neutral policy
where it taxes firms’ profits to subsidize unemployment benefits and a vacancy
maintenance fee. The policy comes at the cost of worsening labor and national
income. Labor informality will eventually shrink if labor skills improve. Therefore,
145
the government should focus on education and training programs rather than
intervening in the formal sector with policies.
Chapter III investigates the differential effect of financial development on
export performance empirically. The analysis uses a detailed data set from the
World Bank, the Exporter Dynamics Database. Financial development is measured
by the amount of credit that depository institutions provide to the private sector
as a proportion of GDP. I find that 67% of the positive impact of financial
development on exports comes from more firms able to survive or start exporting.
This result supports a popular view set by the heterogeneous firm trade theory
that the main channel through which finance influences exports is the extensive
margin. I do not find evidence regarding the role of financial development on the
intensive margin measured by average sales per firm. However, when I separate the
exporting firms by status, I find that the intensive margin of the more productive
groups (incumbents and surviving entrants) reacts positively to improvements in
credit markets. While the extensive margin remains the more important channel,
its relative role to the intensive margin is not the same between incumbents and
surviving entrants. The contribution of extensive margin to the total effect of
financial development is roughly 56% for incumbents, and 72% for surviving
entrants. This result highlights the challenge from fixed cost financing to first-
time exporters. Continuing exporters still incur fixed costs, e.g., from exploring
new destinations or exporting new products. However, these fixed costs are less of
146
a financial obstacle to continuing exporters. The results in this chapter should be
taken as strong associations instead of causation. There are several threats to the
causal inference, which can only control with more detailed data.
Chapter IV explores the roles of finance as drivers of export quality
upgrading, i.e., the speed at which export quality of a product in a specific
country approaches the world frontier. I propose that finance has a positive direct
effect: it speeds up quality upgrading through promoting innovative technologies.
Finance is more influential for products with specific characteristics: (i) further
away from the world frontier of quality, or (ii) requiring more external finance to
produce. These hypotheses are based on an established understanding of financial
development as a source of comparative advantage and an engine of economic
convergence. Finance as the main explanatory variable is measured in three
ways: private credit to GDP ratio, FDI net inflows to GDP ratio, and a financial
reform index. I use quality data from the IMF at the most disaggregated level
(SITC 4-digit). I confirm that all measures positively affects quality upgrading
and disproportionately more for products with initially low quality. Products of
more financially vulnerable sectors would upgrade their quality faster as a result
of improvements in private credit. A simple exercise shows that the importance
of three financial measures varies across countries of different income levels.
Similar exercises that explore the heterogeneity across countries and product
characteristics are necessary to enrich the analysis in this chapter.
147
APPENDIX A
COMPARATIVE STATICS GRAPHS FOR CHAPTER II
FIGURE 13. Comparative Statics: Productivity of high-skilled workers
FIGURE 14. Comparative Statics: Unemployment benefit
148
FIGURE 15. Comparative Statics: Efficiency of matching function
FIGURE 16. Comparative Statics: Adverse shock arrival rate
149
FIGURE 17. Comparative Statics: Formation rate of informal entrepreneurs
FIGURE 18. Comparative Statics: Sunk cost of formal firms
150
FIGURE 19. Comparative Statics: Real interest rate
FIGURE 20. Comparative Statics: Proportion of high-skilled worker in the population
151
TABLE A1. Comparative Statics: Percentage changes of equilibrium values
Parameters Range Change (%) θ∗ w∗ w∗ e∗ i∗ u∗HF HI h h h
y 8.00:12.00 50 97.1 52.2 53.0 33.7 -34.7 -4.8
b 2.00:5.00 150 -41.9 2.1 4.6 -46.2 139.6 -29.4
K 1.00:3.00 200 26.8 7.3 16.4 23.6 -14.1 -63.4
δF 0.10:1.00 900 -18.3 -5.5 -11.8 -17.6 13.4 82.4
α 0.490:1.50 200 -20.2 -6.3 -13.3 -19.1 15.3 120.0
κ 0.490:5.00 900 -92.0 -7.6 -13.3 -15.6 15.3 197.4
r 0.01:0.10 900 -92.7 -10.49 -18.0 -21.1 22.0 191.2
φ 0.30:0.490 100 -0.49 0.0 -0.4 -0.0 -0.0 0.3
z 3.00:5.00 66.7 0.0 -0.0 0.0 -0.0 0.0 -0.0
t 0.40:0.70 75 72.3 -0.8 1.0 3.5 -1.8 -21.1
Parameters Range Change (%) w ∗ ∗ ∗gap v GNIF GNII GNI
∗ ∗
T wl
y 8.00:12.00 50 -0.49 87.7 96.8 0.0 36.7 0.00
b 2.00:5.00 150 -2.3 -59.0 -42.0 150.8 32.6 150.8
K 1.00:3.00 200 -7.6 -53.6 26.0 0.0 11.6 0.00
δF 0.10:1.00 900 6.2 48.9 -18.4 0.0 -8.8 0.00
α 0.490:1.50 200 7.0 75.5 -20.2 -0.0 -9.9 -0.00
κ 0.490:5.00 900 6.0 -76.1 -19.1 0.0 -9.9 0.00
r 0.01:0.10 900 8.0 -78.6 -25.2 -0.0 -12.9 -0.00
φ 0.30:0.490 100 0.4 -0.3 100.0 38.9 71.2 86.7
z 3.00:5.00 66.7 -0.0 -0.0 -0.0 0.0 0.0 0.00
t 0.40:0.70 75 -1.7 35.9 0.0 -0.8 -0.4 -0.85
Unit: The range of all parameters (Column 2) are measured in level. The changes of
parameters (Column 3) and, except for wage gaps, changes of endogeneous variables
(Column 4 to 9) are measured in percent. The change in wage gaps is measured in
percentage point, since the wage gap is the difference between formal wage and informal
wage as a percentage of formal wage.
Notation: y: Productivity of high-skilled workers, b: Unemployment benefit of high-
skilled workers, K: Scale parameter of matching function, δF : Arrival rate of adverse
productivity shock to formal matches, α: Formation rate of informal firms, κ: Sunk cost
for formal jobs, r: Real interest rate, φ: Ratio of high-skilled workers in the population,
z: Productivity of low-skilled workers, t: Tax on profits of formal firms with filled jobs,
Endogenous variables: θ∗: Market tightness, v∗: Vacancy ratio, w∗HF : Formal wage
and w∗HI : Informal wage of high-skilled workers, w
∗
l : Wage of low-skilled workers, e
∗
h:
Formal employment rate and i∗h: Informal employment rate among high-skilled workers,
u∗h: Unemployment rate, GNI
∗
F : GNI of formal sector, GNI
∗
I : GNI of informal sector,
GNI∗T : total GNI.
152
APPENDIX B
SUPPLEMENTAL MATERIALS FOR CHAPTER III
A Model of Credit Constraints
Manova (2013) builds upon Melitz (2003) a static, partial equilibrium model
with credit constraints. There is a continuum of firms producing differentiated
products in J origin countries and S sectors. Destination country is denoted
i. Country i has a Cobb-Douglas utility Ui = Π C
θs
s is aggregated over sector
specific CES consumption indices Cis where Ωis is the set of available products,
 = 1/(1 − α) > 1 is the elasticity o∑f substitution, and θs ∈ (0, 1) is the share of
each sector in total expenditure Yi ( s θs = 1). Country i’s demand for a variety
with price pis(ω) is qis(ω) = (p
− 1−
is(ω) Yi)/(Pis ) where Pis is the ideal price index.
Consumers in country j have CES preferences over product varieties ω.
Denote  = 1/(1 − α) as the elasticity of substitution, Yj as total expenditure of
destination country, and θs ∈ (0, 1) as the share of each sector s in country j’s total
expenditure, and Pjs as ideal price index. Every period, a firm with a productivity
of 1/aijs in country i and sector s chooses its exporting level q, exporting price p,
and the repayment amount F to its investors. The firm maximizes profit subject
to three constraints: the demand for variety from foreign consumers, liquidity
153
constraint, and the promise of a non-negative return from the loan to the investors.
maxπijs(a) = pijs(a)qijs(a)−(1−ds)qijs(a)τijcisa−(1−ds)cisfij−λiF (a)−(1−λi)tscisfei
p,q,F
subject to
p (a)−ijs θsYj
(i)qijs(a) =
P 1−js
(ii)Aijs(a) = pijs(a)qijs(a)− (1− ds)qijs(a)τijcisa− (1− ds)cisfij ≥ F (a)
(iii)Bijs(a) = −dsqijs(a)τijcisa− dscisfij + λjF (a) + (1− λj)tscisfei ≥ 0
Here Aijs(a) denotes the revenue of the firms, which is also the highest repayment
that the firm could offer to the investors. Bijs(a) denotes their return from lending
the firms. Investors would only make loans to firms if the return from the loans
Bijs(a) exceeds their outside option, here normalized to 0.
Solving this optimization problem yields the optimal revenue rijs. The cost for
a firm to produce one unit of its variety is cisaijs, where cis is the cost of a cost-
minimizing bundle of inputs specific for each country and sector. Every firm draws
its productivity from a distribution G(a) after paying a fixed sunk cost of cisfei to
enter the domestic market. An exporting firm would incur extra expense: it pays a
fixed cost cisfij (where fij > 0 if i 6= j, fii =0) and an iceberg trade cost τij > 1
each period.
154
Contractual enforcement is denoted λi ∈ (0, 1). This exogenous parameter
reflects the probability that investors can be repaid in the case the firm defaults on
its debt. Contractual enforcement is greater in countries with better institutional
quality. Every period the exporting firm finances a fraction ds of its trade costs
with external funds from borrowing. The firm pledges the loan with tangible assets
that comes from its sunk entry cost. How much of the sunk entry cost goes into
tangible assets (ts ∈ (0, 1)), and how much of the trade costs is financed externally
(ds ∈ (0, 1)) are determined by the technological difference across industries. These
factors ts and ds, assumed to be exogenous to the firms, capture the heterogeneity
in sectoral financial vulnerability. A sector is more financially vulnerable if it needs
more external fund (higher ds) or has little tangible assets (lower ts).
The first testable hypothesis relates the extensive margin of trade with
financial development and sectoral financial vulnerability. The productivity cut-
off for exporting is higher in financial(ly more vulnerable sectors and lower in∂(1/aijs) ∂(1/aijs) ∂(1/aijs) )
financially more developed countries > 0, < 0, < 0 .
∂ds ∂ts ∂λi
Financial developm(ent lowers this cut-off relativel)y more in financially more∂2(1/a 2ijs) ∂(1/aijs)
vulnerable sectors < 0, > 0
∂λi∂ds ∂λi∂ts
The second testable hypothesis relates the intensive margin of trade with
financial development and sectoral financial vulnerability. Financial development
(weakly) increases the level of firms’ exports( from country i to countr)y j relatively∂2rijs ∂2rijs
more in financially more vulnerable sectors > 0, < 0 . The term
∂λi∂ds ∂λi∂ts
155
”weakly” is involved because financial development only increases the export level
of firms that have a productivity between aLijs and a
H
ijs. The liquidity constraint
(ii) is binding for firms with productivity lower than aLijs and not binding for firms
with productivity at least aHijs.
Data Exploration
TABLE A2. Variables description and source
Variable Description Source
Country-level variables
gdp GDP 2011 PPP constant USD WDI
gdppc ppp GDP per capita, 2011 PPP constant USD WDI
privcred Domestic credit to private sector to GDP (percent) WDI
legal Legal right index
rulelaw Rule of Law
so or Socialist origin before transition CEPII
fr or French origin before transition CEPII
uk or British origin before transition CEPII
sc or Scandinavian origin before transition CEPII
Industry-level variables - Author calculated
EXTFIN External financial dependence ratio Compustat
TANG Tangible assets to total assets ratio Compustat
Export data
A1 Number of Exporters EDD
A2 Number of Entrants EDD
A3 Number of Exiters EDD
A4 Number of Surviving Entrants EDD
A5 Number of Incumbents EDD
A6i Export Value per Exporter: Mean EDD
A7i Export Value per Entrant: Mean EDD
A9i Export Value per Surviving Entrant: Mean EDD
A10i Export Value per Incumbent: Mean EDD
Country-level data
Pool data highest and lowest country-year instances: The smallest countries
by GDP are Rwanda (2004, 2005) and Timor Leste (2006). Largest is Mexico
156
(2008, 2011, 2012). The one with the lowest GDP per capita is Niger (2009, 2010,
2008). Highest are Norway (2007), Kuwait (2009 and 2010). The one with the
lowest private credit, Timor Leste (2008, 2011, 2012). Highest is Denmark (2008,
2009, 2010). Pearson’s correlation: the pool data correlation coefficient between
GDPpc and Privcred is 0.68, significant at 10%, between GDP and GDPpc or
GDP and Privcred is 0.3 and significant at 10%
TABLE A3. Country level cross-section statistics
Variable Obs Mean SD Min Max
Privcred 52 36.92 37.55 2.96 182.58
GDP 52 2.21 E11 3.87 E11 1.02 E10 1.82 E12
GDP per capita 52 11394 14631.2 803.6 78563.4
TABLE A4. Countries by income ranking 2003
Income rank Description Obs Percent Countries
High income GNIpc > 9, 205 2,033 9.95 5
Upper-middle 2, 976 < GNIpc < 9, 205 3,552 17.39 6
Lower-middle 746 < GNIpc < 2, 975 7,259 35.54 18
Low income GNIpc < 745USD 7,582 37.12 23
Total 20,426 100 52
TABLE A5. Country level cross-section extremes
GDP per capita GDP Private credit
Top 5 countries
Portugal 26858.2 Pakistan 6.69 E11 Mauritius 77.5
Spain 32929.2 Turkey 1.25 E12 Norway 92.6
Denmark 44751 Iran 1.25 E12 Portugal 141.8
Norway 63140.1 Spain 1.50 E12 Spain 157.6
Kuwait 78563.3 Mexico 1.82 E12 Denmark 182.6
Bottom 5 countries
Niger 803.6 Timor-Leste 1.02 E10 Timor-Leste 2.96
Ethiopia 924.5 Rwanda 1.22 E10 Guinea 4.13
Malawi 975.8 Niger 1.27 E10 Yemen 6.31
Rwanda 1245.6 Malawi 1.44 E10 Zambia 8.71
Uganda 1370.6 Kyrgyz 1.48 E10 Uganda 8.78
157
Industry-level data
TABLE A6. List of 3-digit ISIC Rev. 3 industries
ISIC Description ISIC Description
151 Processing meat, fish, veggie, oils 281 Structural metal
152 Dairy 289 Other fabricated metal
153 Grain mill and starch 291 General purpose machinery
154 Other food 292 Special purpose machinery
155 Beverages 293 Domestic appliances n.e.c.
160 Tobacco 300 Office machinery
171 Spinning/finishing of textiles 311 Electric motors
172 Other textiles 312 Electricity apparatus
173 Knitted and crocheted fabrics 313 Insulated wire and cable
181 Wearing apparel, except fur apparel 314 Primary batteries
182 Dressing, dyeing, articles of fur 315 Electric lighting equipment
191 Tanning and dressing leather 319 Electrical equipment n.e.c.
192 Footwear 321 Electronic valves/tubes
201 Saw milling, planing of wood 322 Television/radio transmitters
202 Wood, cork, straw 323 Television/radio receivers
210 Paper 331 medical appliances
221 Publishing 332 Optical/photographic instruments
222 Printing and service activities 333 Watches and clocks
233 Processing of nuclear fuel 341 Motor vehicles
241 Basic chemicals 342 Bodies motor vehicles/trailers
242 Other chemicals 343 Motor accessories
243 Man-made fibres 351 Building, repairing ships/boats
251 Rubber 352 Railway/tramway locomotives
252 Plastics 353 Aircraft/spacecraft
261 Glass 359 Transport equipment n.e.c.
269 Non-metallic mineral n.e.c. 361 Furniture
271 Basic iron and steel 369 Manufacturing n.e.c.
272 Basic precious/non-ferrous metals
There are 55 manufacturing industries. Pearson’s correlation coefficient from
panel data between EXTFIN and TANG is -0.019, significant at 10%. Correlation
from cross section (industry-wise) data is similar at -0.013 and not significant.
158
TABLE A7. Two measures of financial vulnerability by industry
ISIC External finance Asset tangibility ISIC External finance Asset tangibility
151 -0.291 0.35 281 0.125 0.173
152 -0.088 0.382 289 -0.356 0.32
153 -0.779 0.328 291 -0.461 0.254
154 -0.408 0.382 292 0.188 0.166
155 -0.847 0.296 293 -0.587 0.185
160 -4.381 0.13 300 0.203 0.114
171 -0.152 0.402 311 0.215 0.159
172 -0.183 0.171 312 -1.002 0.28
173 -0.514 0.265 313 -0.1 0.248
181 -0.5 0.104 314 0.319 0.207
182 -0.044 0.216 315 -0.611 0.223
191 -0.33 0.223 319 -0.115 0.181
192 -0.364 0.208 321 0.104 0.196
201 -0.4 0.72 322 0.489 0.142
202 -0.355 0.289 323 1.544 0.094
210 -0.409 0.495 331 0.064 0.132
221 -1.099 0.152 332 -0.027 0.23
222 -0.781 0.326 333 -0.885 0.152
233 -0.375 0.376 341 -0.02 0.24
241 -0.079 0.373 342 -1.036 0.201
242 2.94 0.147 343 -0.818 0.239
243 -0.151 0.369 351 -0.315 0.291
251 -0.511 0.415 352 -2.737 0.147
252 -0.307 0.366 353 -1.083 0.2
261 0.205 0.354 359 1.441 0.243
269 -0.278 0.385 361 -0.835 0.257
271 0.037 0.374 369 -0.248 0.145
272 -0.154 0.346
EXTFIN TANG EXTFIN TANG
Mean -0.312 0.261 Min -4.381 0.094
SD 0.932 0.115 Max 2.940 0.720
159
TABLE A8. Industries with highest and lowest external finance
ISIC Description EXTFIN
Largest/most vunerable
242 Other chemical products 2.94
323 Television and radio receivers 1.544
359 Transport equipment n.e.c. 1.441
322 Television and radio transmitters 0.489
314 Primary cells and primary batteries 0.319
Smallest/least vunerable
160 Tobacco products -4.381
352 Railway and tramway locomotives -2.737
221 Publishing -1.099
353 Aircraft and spacecraft -1.083
342 Bodies for motor vehicles, trailers -1.036
TABLE A9. Top industries with highest lowest asset tangibility
ISIC Description TANG
Largest/least vulnerable
201 Saw milling and planing of wood 0.72
210 Paper 0.495
251 Rubber products 0.415
171 Spinning and finishing of textiles 0.402
269 Nonmetallic mineral products n.e.c. 0.385
Smallest/most vulnerable
323 Television and radio receivers 0.094
181 Wearing apparel, except fur apparel 0.104
300 Office, accounting and computing machinery 0.114
160 Tobacco products 0.13
331 Medical appliances 0.132
In the list above, Tobacco (ISIC 160) is among the industries with the least
tangible assets but is least dependent on external financial dependence. Tobacco is
a controversial industry because it is considered most vulnerable by one measure
(TANG) but least vunerable by the other measure (EXTFIN). Television and radio
receiver (ISIC 323), is considered more financially vunerable by both measures.
160
TABLE A10. Pearson Correlation of Export variables
GDP Private External Asset
Description GDP per capita credit finance tangibility
No. of Exporters 0.573* 0.320* 0.372* 0.069* -0.021*
No. of Entrants 0.571* 0.318* 0.356* 0.072* -0.030*
No. of Surviving Entrants 0.588* 0.296* 0.318* 0.067* -0.018*
No. of Incumbents 0.553* 0.317* 0.365* 0.067* -0.014*
Export Value per Exporter 0.010* 0.050* 0.034* -0.057* 0.036*
Export Value per Entrant 0.028* 0.013* 0.011 -0.045* 0.006
Export Value per Surviving 0.035* 0.010 0.011 -0.053* 0.006
Export Value per Incumbent 0.096* 0.038* 0.019* -0.036* 0.033*
TABLE A11. Nest regression with exporter firm type only
(1) (2) (3)
VARIABLES Total exports Extensive margin Intensive margin
Log GDP 1.030*** 0.572*** 0.457***
(0.095) (0.104) (0.073)
Private credit 1.694*** 0.972*** 0.722**
(0.380) (0.352) (0.291)
Private credit × Asset tangibility -2.394*** -0.970*** -1.423**
(0.758) (0.272) (0.685)
Private credit × External finance 0.103 0.094** 0.009
(0.067) (0.042) (0.062)
Observations 12,375 12,375 12,375
R-squared 0.681 0.766 0.455
Country FE No No No
Country controls Yes Yes Yes
Year FE Yes Yes Yes
Industry FE Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Exporters include all three firm types: incumbents, surviving entrants and non
surviving entrants. Effects on Surviving entrants are similar to the separate regression
161
TABLE A12. Nested regression with two firm types: Incumbents and Entrants
(1) (2) (3)
Total Extensive Intensive
VARIABLES exports margin margin
Log GDP 1.034*** 0.576*** 0.458***
(0.092) (0.105) (0.056)
Private credit 1.945*** 1.104*** 0.841***
(0.353) (0.352) (0.238)
Private credit × Asset tangibility -1.207* -0.574* -0.633
(0.689) (0.303) (0.596)
Private credit × External finance 0.061 0.081* -0.020
(0.062) (0.044) (0.053)
IEntrant × Private credit -0.626*** -0.174* -0.452***
(0.191) (0.087) (0.140)
IEntrant × Private credit × Asset tangibility -2.474*** -1.029*** -1.445**
(0.757) (0.243) (0.557)
IEntrant × Private credit × External finance 0.035 0.034** 0.001
(0.047) (0.016) (0.041)
Observations 24,750 24,750 24,750
R-squared 0.671 0.742 0.568
Country FE No No No
Country controls Yes Yes Yes
Year FE Yes Yes Yes
Industry FE Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
162
TABLE A13. Nested regression with three types of firms
(1) (2) (3)
Total Extensive Intensive
VARIABLES exports margin margin
Log GDP 1.034*** 0.572*** 0.462***
(0.095) (0.107) (0.046)
Private credit 1.918*** 1.130*** 0.788***
(0.344) (0.357) (0.218)
Private credit × Asset tangibility -0.705 -0.511* -0.193
(0.680) (0.302) (0.588)
Private credit × External finance 0.013 0.084* -0.071
(0.063) (0.045) (0.054)
INon−survivor × Private credit -0.561** -0.162 -0.399*
(0.266) (0.103) (0.210)
INon−survivor × Private credit × Asset tangibility -3.295*** -1.237*** -2.058**
(1.075) (0.306) (0.822)
INon−survivor × Private credit × External finance 0.106* 0.053*** 0.053
(0.060) (0.019) (0.049)
ISurvivor × Private credit -0.327* -0.170*** -0.157
(0.165) (0.056) (0.139)
ISurvivor × Private credit × Asset tangibility -1.728*** -0.504*** -1.223**
(0.559) (0.102) (0.490)
ISurvivor × Private credit × External finance 0.091* 0.009 0.082*
(0.052) (0.008) (0.048)
Observations 37,027 37,027 37,027
R-squared 0.681 0.747 0.561
Country FE No No No
Country controls Yes Yes Yes
Year FE Yes Yes Yes
Industry FE Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
163
TABLE A14. Marginal effects of Initial Private Credit
Firm types Exporter Entrant Incumbent Survivor
Total Export sales 1.095** 0.871** 1.253*** 1.203**
(0.473) (0.421) (0.483) (0.527)
Extensive margin 0.710 0.614 0.895* 0.725
(0.474) (0.497) (0.478) (0.542)
Intensive margin 0.385 0.257 0.358 0.478**
(0.327) (0.217) (0.318) (0.221)
The initial level of private credit to GDP ratio is set in 2003.
The values in this table are calculated from based on estimation
results reported in three tables following.
TABLE A15. Regression with Initial Value of Private Credit: Total Export Sales
(1) (2) (3) (4)
VARIABLES Exporter Entrant Incumbent Survivor
Log GDP 1.127*** 1.054*** 1.140*** 1.138***
(0.108) (0.093) (0.109) (0.110)
Initial Private credit 1.637*** 1.365*** 2.070*** 2.036***
(0.576) (0.438) (0.576) (0.550)
Init. Private credit × Asset tangibility -1.886* -1.843*** -2.866** -3.037***
(1.105) (0.573) (1.133) (0.711)
Init. Private credit × External finance dep. 0.165** 0.041 0.235*** 0.139**
(0.068) (0.062) (0.083) (0.069)
Observations 17,885 16,044 15,626 12,590
R-squared 0.684 0.639 0.642 0.585
Country controls Yes Yes Yes Yes
Country FE No No No No
Year FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1
164
TABLE A16. Regression with Initial Value of Private Credit: Extensive Margin
(1) (2) (3) (4)
VARIABLES Exporter Entrant Incumbent Survivor
lgdp 0.616*** 0.591*** 0.640*** 0.599***
(0.103) (0.104) (0.105) (0.113)
Init. Private credit 1.004** 0.900* 1.294** 1.153**
(0.475) (0.495) (0.485) (0.536)
Init. Private credit × Asset tangibility -0.973*** -0.995*** -1.332*** -1.498***
(0.324) (0.294) (0.392) (0.361)
Init. Private credit × External finance dep. 0.137*** 0.092 0.178*** 0.137**
(0.049) (0.055) (0.051) (0.061)
Observations 17,885 16,044 15,626 12,590
R-squared 0.768 0.752 0.750 0.730
Country controls Yes Yes Yes Yes
Country FE No No No No
Year FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1
TABLE A17. Regression with Initial Value of Private Credit: Intensive Margin
(1) (2) (3) (4)
VARIABLES Exporter Entrant Incumbent Survivor
lgdp 0.511*** 0.462*** 0.501*** 0.539***
(0.076) (0.042) (0.076) (0.046)
Init. Private credit 0.632 0.465* 0.776* 0.882***
(0.446) (0.243) (0.419) (0.253)
Init. Private credit × Asset tangibility -0.913 -0.848* -1.534 -1.539***
(0.962) (0.450) (0.945) (0.511)
Init. Private credit × External finance dep. 0.028 -0.051 0.057 0.002
(0.062) (0.054) (0.069) (0.047)
Observations 17,885 16,044 15,626 12,590
R-squared 0.450 0.348 0.394 0.316
Country controls Yes Yes Yes Yes
Country FE No No No No
Year FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1
165
TABLE A18. Robustness check: Excluding countries and sectors with highest shares of SOEs
(1) (2) (3) (4) (5) (6)
Full sample Edited sample
VARIABLES Total Export sales Firm count Avg. sales Total Export sales Firm count Avg. sales
Log GDP 1.030*** 0.572*** 0.457*** 0.948*** 0.596*** 0.352***
(0.095) (0.104) (0.073) (0.085) (0.102) (0.087)
Private credit 1.694*** 0.972*** 0.722** 1.565*** 1.079** 0.486
(0.380) (0.352) (0.291) (0.391) (0.410) (0.344)
Private credit × Asset tangibility -2.394*** -0.970*** -1.423** -2.016** -1.021*** -0.995
(0.758) (0.272) (0.685) (0.871) (0.235) (0.759)
Private credit × External finance 0.103 0.094** 0.009 0.088 -0.017 0.105
(0.067) (0.042) (0.062) (0.076) (0.033) (0.066)
Constant -17.013*** -17.058*** 0.044 -16.511*** -17.719*** 1.208
(2.090) (2.032) (1.550) (1.881) (2.067) (1.711)
Observations 12,375 12,375 12,375 7,316 7,316 7,316
R-squared 0.681 0.766 0.455 0.670 0.762 0.425
Country FE No No No No No No
Country controls Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes
Robust standard errors of regression clustered at the country level
*** p<0.01, ** p<0.05, * p<0.1
166
APPENDIX C
SUPPLEMENTAL MATERIALS FOR CHAPTER IV
TABLE A19. Summary statistics of main variables
Variable Obs Mean Std. Dev. Min Max
Log growth of export quality SITC-4 135185 0.0048 0.1078 -2.9902 3.1832
Log growth of export quality SITC-2 221035 0 0.0557 -1.7461 1.9019
Initial quality SITC-4 148177 0.8612 0.1494 0 1.8538
Initial quality SITC-2 39106 0.8676 0.1338 0.0793 1.2845
Initial log GDP per capita 231850 9.1519 1.0344 5.8914 11.3031
Initial Private credit to GDP ratio 231850 0.4179 0.3514 0.0131 1.8651
Initial Domestic financial reform index 231850 0.5367 0.2992 0 1
Initial FDI net inflows to GDP ratio 231850 0.0263 0.0484 -0.0409 0.4107
Asset Tangibility 231850 0.3103 0.1235 0.0906 0.6708
External Financial dependence 231850 0.2956 0.2411 -0.4512 1.1401
The country-level variables are ’initial’ values, meaning they are measured at the starting
year of the five-year non-overlapping period (seven years include 1975, 1980, 1985, 1990,
1995, 2000, and 2005). GDP per capita, Private credit to GDP ratio, and FDI net
inflows to GDP ratio are from World Development Indicator database, by The World
Bank. Domestic Financial reform index is from the Financial Reform Database by the
IMF (Prati et al., 2013; Abiad et al., 2010). Asset tangibility and External financial
dependence are from Braun (2003)
TABLE A20. Example of 4 levels of SITC Rev.1
Level Code Description
SITC-1 digit 1 Beverages and tobacco
SITC-2 digit 11 Beverages
SITC-3 digit 111 Non alcoholic beverages, nes.
112 Alcoholic beverages
SITC-4 digit 1121 Wine of fresh grapes including grape must
1122 Cider and fermented beverages, nes.
1123 Beer including ale,stout,porter
1124 Distilled alcoholic beverages
167
TABLE A21. Summary statistics by country groups
Country Countries Growth Growth Initial quality Initial quality
group no. SITC4 SITC2 SITC4 SITC2
High income 28 0.0032 0.0003 0.931 0.9621
Upper middle 27 0.0067 0.0002 0.8304 0.852
Lower middle 21 0.0062 -0.001 0.7544 0.7789
Low income 11 0.0048 0.0012 0.6822 0.6991
Country Log GDP Private Financial FDI net Obs.
group per capita credit reform inflow
High income 10.1646 0.6785 0.7008 0.04 88947
Upper middle 9.0875 0.3249 0.4802 0.0232 71157
Lower middle 8.2199 0.2025 0.3878 0.0135 53345
Low income 7.2072 0.1427 0.3935 0.0096 18401
Classification of countries is based on the World Bank’s methodology (using
GNI per capita for fiscal year 2011)
TABLE A22. Correlation matrix of country-level initial conditions
LGDP per cap. Private credit FDI inflows Financial reform
LGDP per cap. 1
Private credit 0.6074 1
FDI inflows 0.2835 0.312 1
Financial reform 0.5644 0.5053 0.4174 1
168
REFERENCES CITED
Abiad, A., Detragiache, E., and Tressel, T. (2010). A new database of financial
reforms. IMF Staff Papers, 57(2):281–302.
Acemoglu, D. (2001). Good jobs versus bad jobs. Journal of Labor Economics,
19(1):1–21.
Acemoglu, D., Johnson, S., and Robinson, J. A. (2001). The colonial origins of
comparative development: An empirical investigation. American economic
review, 91(5):1369–1401.
Aghion, P., Bloom, N., Blundell, R., Griffith, R., and Howitt, P. (2005a).
Competition and innovation: An inverted-u relationship. The Quarterly
Journal of Economics, 120(2):701–728.
Aghion, P., Blundell, R., Griffith, R., Howitt, P., and Prantl, S. (2009). The effects
of entry on incumbent innovation and productivity. The Review of
Economics and Statistics, 91(1):20–32.
Aghion, P. and Durlauf, S. (2005). Handbook of economic growth. Elsevier.
Aghion, P., Fally, T., and Scarpetta, S. (2007). Credit constraints as a barrier to
the entry and post-entry growth of firms. Economic policy, 22(52):732–779.
Aghion, P., Howitt, P., and Mayer-Foulkes, D. (2005b). The effect of financial
development on convergence: Theory and evidence. The quarterly journal of
economics, 120(1):173–222.
Aghion, P. and Howitt, P. W. (2008). The economics of growth. MIT press.
Aghion, P., Ljungqvist, L., Howitt, P., Howitt, P. W., Brant-Collett, M.,
Garćıa-Peñalosa, C., et al. (1998). Endogenous growth theory. MIT press.
Albrecht, J., Navarro, L., and Vroman, S. (2009). The effects of labour market
policies in an economy with an informal sector. The Economic Journal,
119(539):1105–1129.
Amiti, M. and Khandelwal, A. K. (2013). Import competition and quality
upgrading. Review of Economics and Statistics, 95(2):476–490.
Amiti, M. and Weinstein, D. E. (2011). Exports and financial shocks. The
Quarterly Journal of Economics, 126(4):1841–1877.
169
Anderson, J. E. and Van Wincoop, E. (2003). Gravity with gravitas: a solution to
the border puzzle. American economic review, 93(1):170–192.
Beck, T. (2002). Financial development and international trade: Is there a link?
Journal of international Economics, 57(1):107–131.
Beck, T., Demirguc-Kunt, A., Laeven, L., and Levine, R. (2008). Finance, firm
size, and growth. Journal of Money, Credit and Banking, 40(7):1379–1405.
Beck, T. and Levine, R. (2018). Handbook of finance and development. Edward
Elgar Publishing.
Benhabib, J. and Spiegel, M. M. (2005). Human capital and technology diffusion.
Handbook of economic growth, 1:935–966.
Berman, N. and Martin, P. (2012). The vulnerability of sub-saharan africa to
financial crises: the case of trade. IMF Economic Review, 60(3):329–364.
Braun, M. (2005). Financial contractability and asset hardness. Available at SSRN
2522890.
Cadot, O., Carrère, C., and Strauss-Kahn, V. (2011). Export diversification:
what’s behind the hump? Review of Economics and Statistics, 93(2):590–605.
Cebeci, T., Fernandes, A. M., Freund, C., and Pierola, M. D. (2012). Exporter
dynamics database. The World Bank.
Chaney, T. (2016). Liquidity constrained exporters. Journal of Economic
Dynamics and Control, 72:141–154.
Chang, A. C. and Li, P. (2018). Measurement error in macroeconomic data and
economics research: Data revisions, gross domestic product, and gross
domestic income. Economic Inquiry, 56(3):1846–1869.
Charmes, J., Gault, F., and Wunsch-Vincent, S. (2016). Formulating an agenda for
the measurement of innovation in the informal economy. The Informal
Economy in Developing Nations–Hidden Engine of Innovation, pages
332–362.
Chen, M., Vanek, J., and Heintz, J. (2006). Informality, gender and poverty: A
global picture. Economic and Political Weekly, pages 2131–2139.
Chen, M. A. (2012). The informal economy: Definitions, theories and policies.
Technical report, WIEGO working paper.
Choi, B. and Lugovskyy, V. (2019). Asymmetric effects of financial development
on export price and quality across countries. Review of International
Economics, 27(2):594–642.
170
Chor, D. and Manova, K. (2012). Off the cliff and back? credit conditions and
international trade during the global financial crisis. Journal of international
economics, 87(1):117–133.
Contessi, S. and De Nicola, F. (2012). What do we know about the relationship
between access to finance and international trade? FRB of St. Louis
Working Paper No.
Cuñat, A. and Melitz, M. J. (2012). Volatility, labor market flexibility, and the
pattern of comparative advantage. Journal of the European Economic
Association, 10(2):225–254.
De Soto, H. et al. (1989). The Other Path. Harper & Row New York.
Diamond, P. A. (1982). Wage determination and efficiency in search equilibrium.
The Review of Economic Studies, 49(2):217–227.
Do, Q.-T. and Levchenko, A. A. (2007). Comparative advantage, demand for
external finance, and financial development. Journal of Financial Economics,
86(3):796–834.
Durlauf, S. N., Johnson, P. A., and Temple, J. R. (2005). Growth econometrics.
Handbook of economic growth, 1:555–677.
Fan, H., Li, Y. A., and Yeaple, S. R. (2015). Trade liberalization, quality, and
export prices. Review of Economics and Statistics, 97(5):1033–1051.
Feenstra, R. C., Li, Z., and Yu, M. (2014). Exports and credit constraints under
incomplete information: Theory and evidence from china. Review of
Economics and Statistics, 96(4):729–744.
Fernandes, A., Freund, C., and Pierola, M. D. (2016). Exporter behavior, country
size and stage of development: Evidence from the exporter dynamics
database. Journal of Development Economics, 119(C):121–137.
Fields, G. S. (1990). Labour market modelling and the urban informal sector:
theory and evidence.
Fields, G. S. (2004). A guide to multisector labor market models. Working paper,
page 86.
Fisman, R. and Love, I. (2003). Trade credit, financial intermediary development,
and industry growth. The Journal of finance, 58(1):353–374.
Foley, C. F. and Manova, K. (2015). International trade, multinational activity,
and corporate finance. economics, 7(1):119–146.
171
Gautier, P. A. (2002). Unemployment and search externalities in a model with
heterogeneous jobs and workers. Economica, 69(273):21–40.
Greenaway, D., Guariglia, A., and Kneller, R. (2007). Financial factors and
exporting decisions. Journal of international economics, 73(2):377–395.
Griliches, Z. and Mairesse, J. (1995). Production functions: the search for
identification. Technical report, National Bureau of Economic Research.
Hallak, J. C. (2006). Product quality and the direction of trade. Journal of
international Economics, 68(1):238–265.
Harding, T. and Javorcik, B. (2011). Fdi and export upgrading. Technical report,
University of Oxford, Department of Economics.
Harris, J. R. and Todaro, M. P. (1970). Migration, unemployment and
development: a two-sector analysis. The American Economic Review,
60(1):126–142.
Hausmann, R., Hwang, J., and Rodrik, D. (2007). What you export matters.
Journal of economic growth, 12(1):1–25.
Henn, C., Papageorgiou, C., Romero, J. M., and Spatafora, N. (2017). Export
quality in advanced and developing economies: evidence from a new data set.
The World Bank.
Hosios, A. J. (1990). On the efficiency of matching and related models of search
and unemployment. The Review of Economic Studies, 57(2):279–298.
Iacovone, L. and Zavacka, V. (2009). Banking crises and exports: lessons from the
past. The World Bank.
Kapoor, M., Ranjan, P., and Raychaudhuri, J. (2012). The impact of credit
constraints on exporting firms: empirical evidence from india. Available at
SSRN 2023068.
Khandelwal, A. (2010). The long and short (of) quality ladders. The Review of
Economic Studies, 77(4):1450–1476.
Kiendrebeogo, Y. (2013). How do banking crises affect bilateral exports?
Empirical Economics, pages 1–29.
Kowalski, P., Büge, M., Sztajerowska, M., and Egeland, M. (2013). State-owned
enterprises: Trade effects and policy implications. Technical report, OECD
Publishing.
Kremer, M. (1993). The o-ring theory of economic development. The Quarterly
Journal of Economics, 108(3):551–575.
172
Krueger, A. B. and Lindahl, M. (2001). Education for growth: Why and for
whom? Journal of economic literature, 39(4):1101–1136.
Maloney, W. F. (2004). Informality revisited. World Development,
32(7):1159–1178.
Manova, K. (2012). Credit constraints, heterogeneous firms, and international
trade. Review of Economic Studies, 80(2):711–744.
Melitz, M. J. (2003). The impact of trade on intra-industry reallocations and
aggregate industry productivity. econometrica, 71(6):1695–1725.
Mortensen, D. T. (1982). The matching process as a noncooperative bargaining
game. In The Economics of Information and Uncertainty, pages 233–258.
University of Chicago Press.
Mortensen, D. T. and Pissarides, C. A. (1994). Job creation and job destruction in
the theory of unemployment. The Review of Economic Studies,
61(3):397–415.
Mummolo, J. and Peterson, E. (2018). Improving the interpretation of fixed effects
regression results. Political Science Research and Methods, 6(4):829–835.
Nunn, N. and Trefler, D. (2014). Domestic institutions as a source of comparative
advantage. In Handbook of international economics, volume 4, pages 263–315.
Elsevier.
Paravisini, D., Rappoport, V., Schnabl, P., and Wolfenzon, D. (2014). Dissecting
the effect of credit supply on trade: Evidence from matched credit-export
data. The Review of Economic Studies, 82(1):333–359.
Pissarides, C. A. (2000). Equilibrium unemployment theory. MIT press.
Porta, R. L., Lopez-de Silanes, F., Shleifer, A., and Vishny, R. W. (1998). Law and
finance. Journal of political economy, 106(6):1113–1155.
Prati, A., Onorato, M. G., and Papageorgiou, C. (2013). Which reforms work and
under what institutional environment? evidence from a new data set on
structural reforms. Review of Economics and Statistics, 95(3):946–968.
Rajan, R. and Zingales, L. (1998). Financial development and growth. American
Economic Review, 88(3):559–586.
Rauch, J. E. (1999). Networks versus markets in international trade. Journal of
international Economics, 48(1):7–35.
Schott, P. K. (2004). Across-product versus within-product specialization in
international trade. The Quarterly Journal of Economics, 119(2):647–678.
173
Sutton, J. and Trefler, D. (2011). Deductions from the export basket: Capabilities,
wealth and trade. Technical report, National Bureau of Economic Research.
Temple, J. (1999). The new growth evidence. Journal of economic Literature,
37(1):112–156.
Ulyssea, G. (2010). Regulation of entry, labor market institutions and the informal
sector. Journal of Development Economics, 91(1):87–99.
Verhoogen, E. A. (2008). Trade, quality upgrading, and wage inequality in the
mexican manufacturing sector. The Quarterly Journal of Economics,
123(2):489–530.
Wang, M. J.-Y. and Ronci, M. M. V. (2006). Access to trade finance in times of
crisis. International Monetary Fund.
Zenou, Y. (2008). Job search and mobility in developing countries. theory and
policy implications. Journal of Development Economics, 86(2):336–355.
174