Capital market imperfections and trade liberalization in general equilibrium

Accepted Manuscript Title: Capital market imperfections and trade liberalization in general equilibrium Author: Michael Irlacher. Florian Unger. PII: DOI: Reference:

S0167-2681(17)30339-6 https://doi.org/doi:10.1016/j.jebo.2017.11.029 JEBO 4211

To appear in:

Journal

Received date: Revised date: Accepted date:

18-1-2017 22-9-2017 27-11-2017

of

Economic

Behavior

&

Organization

Please cite this article as: Michael Irlacher, Florian Unger, Capital market imperfections and trade liberalization in general equilibrium, (2017), https://doi.org/10.1016/j.jebo.2017.11.029 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ip t

Capital Market Imperfections and Trade Liberalization in General Equilibrium Florian Unger

University of Munichy

University of Munichz

an

us

November 29, 2017

cr

Michael Irlacher

Abstract

M

This paper develops a new international trade model with rm-speci c credit frictions and endogenous borrowing costs in general equilibrium. We highlight new implications of globalization when general equilibrium e ects on capital markets are present.

d

In particular, we show that globalization increases the share of nancially constrained rms and a ects producers very di erently depending on their exposure to credit fric-

te

tions. While the positive e ect of globalization dominates for unconstrained

rms,

higher borrowing costs and tougher competition especially hurt credit-rationed pro-

Ac ce p

ducers. We show that these new adjustments increase the heterogeneity among rms and reduce welfare gains from trade. Our theory is consistent with new empirical patterns from World Bank rm-level data. We show that credit frictions are positively related to the degree of product competition and to the variance of sales across rms. Keywords: Credit constraints, General equilibrium, Globalization, Imperfect capi-

tal markets, Welfare.

JEL Classi cation: F10, F36, F61, L11

We thank Daniel Baumgarten, Ron Davies, Swati Dhingra, Carsten Eckel, Hartmut Egger, Gabriel Felbermayr, Lisandra Flach, Udo Kreickemeier, Eduardo Morales, Monika Schnitzer, Jens Suedekum, Alexander Tarasov, and Jens Wrona, as well as participants of the Munich \IO and Trade seminar", Workshop \Internationale Wirtschaftsbeziehungen" in Goettingen, BGPE Research Workshop in Passau, Annual Conference of the European Trade Study Group in Paris, FIW Research Conference in Vienna, HSE Workshop \International Trade and Spatial Economics" in Saint Petersburg, Annual Congress of the European Economic Association in Geneva, and the Annual Conference of the German Economic Association in Augsburg for helpful comments and suggestions. Financial support from the Deutsche Forschungsgemeinschaft through SFB/TR15 is gratefully acknowledged. y Corresponding author: Department of Economics, D-80539 Munich, Germany; e-mail: [email protected] z Department of Economics, D-80539 Munich, Germany; e-mail: [email protected]

Page 1 of 40

1

Introduction

Ac ce p

te

d

M

an

us

cr

ip t

Export activity of rms crucially depends on nancial development and access to external capital (Beck, 2003; Manova, 2008, 2013). However, producers face di erent obstacles to obtain credit, especially in developing countries.1 According to the World Bank Enterprise Surveys, 24 percent of rms report that access to nancing is a severe obstacle to their current operations.2 Related to this, empirical studies show that credit frictions at the rm level are important determinants of exports, even after controlling for other characteristics, such as size and productivity (Berman & Hericourt, 2010; Minetti & Zhu, 2011; Mu^ uls, 2015). International trade models mainly focus on nancial frictions at the industry- or country level without considering rm-speci c credit constraints (Manova, 2013; Chaney, 2016).3 However, introducing credit frictions at the rm level in general equilibrium helps to understand the relationship between international trade and imperfect capital markets. As we show in this paper, this leads to new insights how globalization a ects rm behavior and welfare. To motivate our theoretical model, we exploit World Bank enterprise survey data for developing countries and highlight four novel empirical patterns. First, we show that the majority of variation in credit constraints is across rms within industries rather than between industries. Second, a larger fraction of rms is nancially constrained in industries with a higher degree of product market competition. Third, countries with a lower nancial development show a higher share of credit-rationed producers. Fourth, lower nancial development is associated with a larger within-industry variance of rm sales. Importantly, all results hold after controlling for rm characteristics such as productivity and size. Consistent with these patterns, we build a theoretical model with rm-speci c credit frictions and endogenous borrowing costs in general equilibrium. A novel feature of our model is that rm heterogeneity results from the interaction between credit constraints at the rm level and capital market imperfections at the country level. Credit frictions arise from a rm-speci c moral hazard problem that reduces the pledgeability of sales. Producers require external capital to cover production costs and di er in their incentives to divert external funds, while being homogenous in other respects. If nancial institutions are imperfect, our model features rm heterogeneity and an endogenous share of credit-rationed rms that depends on industry characteristics such as the degree of competition and the interest rate. While unconstrained rms can produce optimally, 1

The link between credit frictions and international trade is particularly relevant in developing countries where the quality of nancial institutions is low (Banerjee & Du o, 2005, 2014). 2 The result is based on the authors' own computations from a cross-section of 21,067 rms in 2013. 3 See Foley & Manova (2015) for a review of the trade and nance literature.

1 Page 2 of 40

Ac ce p

te

d

M

an

us

cr

ip t

producers with high incentives to misbehave cannot overcome credit frictions and thus have to restrict production. As a further departure from previous theoretical work, we explicitly model a capital market equilibrium and endogenize borrowing costs in general equilibrium.4 Having characterized the equilibrium properties of our model, we analyze the e ects of globalization on producers and welfare. We show that globalization increases the share of constrained rms and a ects producers very di erently depending on their exposure to credit frictions. The intuition is as follows: globalization leads to additional export opportunities, which increases capital demand and thus the borrowing rate. Stronger competition from foreign producers reduces the pledgeability of sales which aggravates access to external credit. While the positive market size e ect of globalization dominates for unconstrained rms, higher borrowing costs and tougher competition especially hurt credit-rationed producers. Moreover, some initially unconstrained rms face credit-rationing, reduce output, and have to set higher prices. We show that these new adjustments increase the variance of sales and thus generate rm heterogeneity. We consider the indirect utility associated with quadratic preferences as a welfare measure. As consumers dislike price heterogeneity, a higher variance represents a negative welfare channel of globalization. This paper contributes to the literature on credit constraints in international trade. While previous work introduces nancial frictions in trade models with rm heterogeneity in productivity5 , we show an alternative mechanism: heterogeneity arises from the interaction of credit constraints at the rm level and capital market imperfections at the country level. A further notable feature of our model is the coexistence of constrained and unconstrained producers in equilibrium. Since the share of constrained rms is endogenous, the size of both groups responds to shocks such as globalization. Another di erence to the related literature is the way how the variance of rm sales is a ected by globalization. Previous work stresses rm selection whereas adjustments work through changes in the cuto productivity and thus the extensive margin. In our model, new results arise as the variance of sales is a ected through changes in the share of constrained rms and reallocations between constrained and unconstrained producers. Our empirical results regarding the variance of sales and the share of constrained rms also contribute to the existing literature which mainly focuses on the e ects of credit frictions on export decisions (Berman & Hericourt, 2010; Minetti & Zhu, 2011; Manova, 2013). As we analyze the e ects of credit frictions on product markets in general equilibrium, our 4

Foellmi & Oechslin (2010) consider a capital market equilibrium as well, see the discussion below. See e.g. Mu^ uls et al. (2008), Manova (2013), Chaney (2016), and Feenstra et al. (2014) for extensions of the Melitz (2003) model by nancial frictions. Gross & Verani (2013), Besedes et al. (2014), as well as Brooks & Dovis (2015) introduce credit constraints in dynamic trade models with heterogeneous rms. Peters & Schnitzer (2015) analyze borrowing constraints in the framework of Melitz & Ottaviano (2008). 5

2 Page 3 of 40

Ac ce p

te

d

M

an

us

cr

ip t

paper is related to Foellmi & Oechslin (2010). They also consider endogenous adjustments of borrowing costs, however, the focus of their approach is a di erent one. In a model with CES preferences and heterogeneity in wealth, they analyze the distributive impact of trade liberalization in less-developed countries. The authors show that globalization impedes access to external nance, especially for poor entrepreneurs, resulting in an increase of income inequality. In our setting with linear demand, we can disentangle the market size from the competition e ect and separately analyze their impacts on equilibrium outcomes. In contrast to a model with CES preferences, markups are endogenous and thus a ected by procompetitive e ects of globalization.6 The advantage of our framework is its high tractability, which allows us to explicitly solve for all endogenous variables, and to conduct comparative static analysis with respect to nancial development and globalization. We derive welfare and show how capital market adjustments alter the gains from trade. Another paper that analyzes the welfare implications of credit frictions is Formai (2013). In a general equilibrium framework based on Melitz (2003), the author shows how credit frictions distort the entry decision of producers, whereas trade liberalization can lead to negative welfare e ects. There is another related literature that models credit frictions with moral hazard a la Holmstrom & Tirole (1997). These papers consider several aspects of nancial market imperfections but the main di erence to our work is that we introduce rm-speci c credit frictions. Von Ehrlich & Seidel (2015) focus on the e ects of nancial development on labor mobility and regional income inequality. Egger & Keuschnigg (2015) analyze how nancial institutions a ect entry and innovation decision of heterogeneous rms. In our framework, the crucial mechanism in general equilibrium is the endogenous adjustment of the interest rate after globalization. Therefore, our analysis is related to models that study how credit frictions a ect international capital and trade ows. In a Heckscher-Ohlin model with heterogeneous nancial frictions across countries and sectors, Antras & Caballero (2009) show that trade integration increases the interest rate in nancially underdeveloped countries. Whereas this result is driven by specialization and across-sector reallocation of inputs, in our model interest rate adjustments after globalization lead to within-industry reallocation of market shares between constrained and unconstrained rms. While our focus is on adjustments on the capital market, we abstract from the fact that trade endogenously a ects a country's nancial development (Do & Levchenko, 2004; Hodler, 2011; Khandelwal et al. , 2013). The paper is structured as follows. The next section provides empirical motivation for our theoretical setup. Section 3 presents the theoretical model and comparative statics in 6

Foellmi & Oechslin (2015) focus on how trade a ects the distribution of markups. The combination of wealth inequality and nancial imperfections leads to endogenous markups in a CES framework.

3 Page 4 of 40

partial equilibrium. Section 4 introduces the capital market and shows general equilibrium e ects of globalization. In Section 5, we discuss the welfare e ects from globalization in partial and general equilibrium and nally, Section 6 concludes.

Empirical motivation

ip t

2

Ac ce p

te

d

M

an

us

cr

We present new empirical patterns which relate rm-level credit constraints to product market competition and variation in rm performance. The empirical analysis is entirely descriptive and aims to motivate our theoretical framework. In this section, we describe the main empirical facts, whereas a detailed description of the data and further robustness checks are provided in Appendix 7. We use cross-sectional rm-level data from the World Bank Enterprise Surveys (WBES). The analysis is based on a sample of developing countries for which credit frictions represent important barriers to international trade. To capture credit frictions at the rm level, we use a self-reported measure of access to credit.7 A survey question asks rms to state whether access to nancing is an obstacle to the current operations of the establishment.8 We introduce a dummy variable for nancially constrained producers if rms perceive access to nancing as a major or very severe obstacle. To measure nancial development, we use a broad index developed by the International Monetary Fund (Svirydzenka, 2016). The main advantage is that it combines several characteristics of nancial systems related to depth, access and e ciency of nancial institutions.9 We relate our measure for credit access to the degree of product market competition, as well as to the within-industry variance of rm sales. To obtain a proxy for competition, we exploit a survey question which asks rms to assess the impact of a hypothetical price increase by 10% for their main product on own demand. We use variation of this categorical variable at the rm level and compute the industry mean. Variables are reported in local currency units, which we convert to 2005 U.S. dollars. In a rst step, we decompose the total variation in credit constraints into within- and between-industry variation. The literature on international trade stresses the importance of rm heterogeneity. Hence, one concern is that the within-industry variation is mainly driven by di erences in rm characteristics such as size or productivity. To address this, we include a set of rm-level controls related to productivity, size, legal status, and ownership structure. 7

In a previous version of the paper Irlacher & Unger (2016), we obtain similar results when using tangible over total assets as an alternative measure of credit constraints. 8 This self-reported information on credit access is used in Gorodnichenko & Schnitzer (2013) and Foellmi & Oechslin (2015) as well. For a survey of empirical studies using rm-level data see Wagner (2014). 9 A detailed description of the measure can be found in Svirydzenka (2016).

4 Page 5 of 40

ip t cr us an

M

Figure 1: Within- and between-industry variation of nancial constraints

te

d

Figure 1 shows results for ve countries at three levels of industry aggregation and reveals that a substantial part of the variation is within industries.10 The observed pattern suggests that producers within the same industry are a ected very di erently by credit constraints, even after controlling for other rm characteristics.

Ac ce p

Empirical pattern 1 The majority of variation in nancial constraints is across rms within industries rather than between industries. In a next step, we show a positive relationship between credit constraints and the degree of product market competition. Table 1 shows the correlations both at the rm level, as well as the industry level. Firms that report more price-sensitive consumers face stronger creditrationing. This pattern holds after controlling for other rm characteristics and industry as well as country xed e ects (see Appendix 7.3). The positive relationship holds at the industry level as well, whereas in industries with a higher degree of competition a larger fraction of producers is nancially constrained.11 Empirical pattern 2 Industries with a higher degree of product market competition show a larger fraction of nancially constrained rms. We further relate the e ect of nancial development to the share of nancially constrained 10

The pattern holds for all countries with available data in our sample (see Table 4 in the Appendix 7.2) Empirical patterns 1 and 2 are based on the cross section 2002-2005. Due to data availability, pattern 2 is based on the following countries: El Salvador, Guatemala, Honduras, Nicaragua, Vietnam, and Zimbabwe. 11

5 Page 6 of 40

Table 1: Correlation credit constraints and product market competition

YEM UGA

TZA

KHM TZA GHA

6

.4

COD

.2

KEN

.1

BLR

ARM BGD MKD GEO UKR SRB BIHLTU

BGD

MNG LVA

HRV BGR

ROU

ZMB

POL

TJK

CZE

MDA KAZ EST

HUN

TUR

LTU GEO SRB ALB ARM MKD

BLR

AZE

2

.1

.2

.3 .4 Financial development

Share of financially constrained firms

.5 Fitted values

.6

0

M

0

.1

JOR

NPL

MDG

SVN

0

ALB

TUR

KEN

an

AZE MDG KHM

us

ROU

YEM UGA TJK

JOR

NPL

4

.3

ZMB

cr

8

COD

.5

.6

GHA

ip t

Degree of competition Access to nance Share constrained rms Firm level 0.0832*** Industry level 0.1530*** Obs. 27,474 300 Notes: *** indicates signi cance at the 1% level.

.2

UKR BIH

POL

LVA MDA BGR CZE

HUN SVN

EST KAZMNG

.3 .4 Financial development

Within-country variance of firm sales

HRV

.5

.6

Fitted values

d

Figure 2: Financial development and share constrained rms (lhs) / variance of sales (rhs)

Ac ce p

te

rms within an industry. The left panel of Figure 2 depicts that higher nancial development is associated with a lower share of nancially constrained rms. For further robustness, we show that the negative e ect of nancial development on credit access holds at the rm level as well (see Appendix 7.4). Empirical pattern 3 Countries with lower nancial development are characterized by a higher share of credit-rationed producers. Finally, we analyze the relationship between nancial development and the within-industry variance of rm sales. The right panel of Figure 2 shows a signi cantly negative relationship. A major concern is that the variance in sales is driven by di erences in size and productivity. Hence, in Appendix 7.5 we show that the pattern is robust when we include industry means and variances of both rm size and labor productivity.12 Empirical pattern 4 Countries with lower nancial development are characterized by a larger within-industry variance of rm sales. 12

Empirical patterns 3 and 4 are based on cross sections for the years 2009 and 2013 with 58 developing countries (see Sections 7.4 and 7.5).

6 Page 7 of 40

The model

us

3

cr

ip t

Our patterns indicate that the variation in credit constraints is only partly explained by di erences in size and productivity. Motivated by the rst empirical pattern, the next section introduces a new international trade model with heterogeneity in credit frictions at the rm level. Financial frictions are positively related to product market competition and heterogeneity in rm performance after controlling for other rm characteristics such as size and productivity (see patterns 2-4). Our theoretical framework provides a rationale for these patterns as rm heterogeneity results from the interaction between credit constraints at the rm level and capital market imperfections at the country level.

3.1

Ac ce p

te

d

M

an

This section develops a model of international trade with heterogeneity in credit frictions at the rm level. The world economy consists of k identical countries, each of which is populated by a number of L consumers and an exogenous mass of m producers. We motivate nancial frictions by a simple moral hazard problem between borrowing rms and external investors. The following subsection presents the demand side of the model, whereas we assume a quadratic speci cation of preferences and derive market demand by aggregating over the number of consumers in the economy. Section 3.2 shows how rms optimally behave in the presence of capital market imperfections depending on their exposure to nancial frictions. The industry equilibrium, outlined in section 3.3, is determined by total industry output and an endogenous share of credit-rationed producers. Finally, in section 3.4, we analyze the e ects of globalization and of an interest rate shock in partial equilibrium.

Consumer side

The representative consumer's utility is de ned over per variety consumption q(i) and total R consumption Q q(i)di, where the index i represents one variety and is the set of i2 horizontally di erentiated products: U = aQ

1 b (1 2

e)

Z

q(i)2 di + eQ2 .

(1)

i2

The quadratic utility function depends on the non-negative preference parameters a, b and on an inverse measure of product di erentiation e which lies between 0 and 1. Lower values of e imply that products are more di erentiated and hence less substitutable. If e = 1, consumers have no taste for diversity in products and demand depends on aggregate output Q only. Thus, the parameter e determines the degree of product market competition and is closely

7 Page 8 of 40

ip t

related to the competition variable in our empirical motivation. Consumers maximize utility R in equation (1) subject to the budget constraint i2 p(i)q(i)di I, where p (i) denotes the price for variety i and I is individual income.13 The maximization problem yields the linear inverse demand function: p(i) = a b [(1 e)q(i) + eQ] , (2)

b0 [(1

e)x(i) + eX] ,

(3)

an

p(i) = a

us

cr

where is the marginal utility of income, the Lagrange multiplier attached to the budget constraint. As rms are in nitesimally small in the economy, they take as given. In the following, we set the marginal utility of income as the numeraire equal to one.14 To ensure market-clearing, total output of each rm equals the aggregate demand of all consumers in the world economy: x(i) = kLq(i). Hence, the inverse world market demand is given by:

Firm's maximization problem

d

3.2

M

b is an inverse measure for where a is the consumers' maximum willingness to pay and b0 kL R the market size. Finally, X x (i) di represents the total volume of varieties produced i2 and consumed in the world economy.

Ac ce p

te

The industry consists of an exogenous mass of m rms. Each rm is run by an entrepreneur with zero assets and produces a horizontally di erentiated variety i. Firms receive revenues p(i)x(i) and have to nance total variable production costs cx(i) by external capital from a perfectly competitive credit market. There are no xed costs of production. Motivated by empirical pattern 1, we assume that rms di er in their exposure to credit constraints. While producers are homogenous in marginal production costs c, the interaction of rmlevel credit frictions and capital market imperfections creates rm heterogeneity. If nancial institutions are imperfect, only a fraction of producers can overcome credit frictions, receives the required capital amount and is able to produce the optimal output. In contrast, rms with high exposure to credit constraints su er from underprovision of external capital and cannot behave optimally. In equilibrium, the share of nancially unconstrained rms is endogenously determined and a ected by trade shocks. As we are interested in the e ects of globalization on producers with di erent exposure to credit constraints, we do not consider 13 In general equilibrium, aggregate income consists of rm pro ts and factor income. We assume that capital is the only factor of production. Section 4 discusses the general equilibrium of the model. 14 Using the marginal utility of income as a numeraire ( = 1) is standard in the literature of oligopoly in general equilibrium (GOLE). See Neary (2016) for further discussion.

8 Page 9 of 40

an

us

cr

ip t

endogenous entry and exit decisions.15 In the following, we describe the rm's maximization problem and introduce credit frictions at the rm- as well as the country level. The decision problem of a producer consists of two stages. At date t = 0, the rm borrows the credit amount d(i) from an outside investor at the rate of return r.16 In partial equilibrium, the interest rate is treated as exogenous, whereas we endogenize it in general equilibrium as discussed in section 4. To motivate credit frictions at the rm level, we introduce a project choice of the entrepreneur which is non-veri able for outside investors and hence prone to moral hazard.17 After credit provision, the entrepreneur can choose whether to use the external funds for production or divert the total credit amount and reap a private bene t. In case of diligent behavior, the pro ts realize with success probability 0 < < 1. At date t = 1, production yields expected pro ts which consist of revenues net of loan repayment: (i) = [p(i)x(i) rd(i)] , (4) whereas the rm faces the following budget constraint: cx(i).

M

d(i)

(5)

15

Ac ce p

te

d

Alternatively, the entrepreneur can choose to divert the total loan without using the provided capital in the production process. Hence, no revenues are realized and the loan cannot be repaid.18 Instead, in period t = 1 the entrepreneur receives a non-veri able private bene t proportional to the repayment: (i) (1 ) rd(i).19 This private bene t consists of a country-speci c (1 ) and a rm-level component (i). We follow Antras et al. (2009) and assume that private bene ts are negatively related to the quality of nancial institutions captured by the parameter 2 [0; 1] : Countries with more e cient nancial institutions (larger ) tend to enforce laws that limit the scope for misbehavior which implies easier access to external credit.20 Hence, this corresponds to our empirical measure for nancial In the Web Appendix, we endogenize the number of rms by allowing for free entry and show that the qualitative implications of our model remain robust. 16 The rate of return r is de ned as 1 + interest rate. For simplicity, we refer to r as the interest rate throughout our analysis. 17 See Holmstrom & Tirole (1997) as well as Tirole (2006) for moral hazard in corporate nance. Recent papers that introduce credit constraints motivated by moral hazard in a trade context are von Ehrlich & Seidel (2015) and Egger & Keuschnigg (2015). 18 For tractability, we assume that in case of misbehavior all funds are lost and cannot be seized by investors. 19 This assumption is related to Holmstrom & Tirole (1997) and Egger & Keuschnigg (2015) where the private bene t is proportional to the size of the investment project. 20 See Tirole (2006) as well as Antras et al. (2009) for a similar notion of nancial contract enforcement in models with moral hazard. Related to Do & Levchenko (2004) and Hodler (2011), we model nancial development as the quality of institutions. Compared to this literature, we abstract from an endogenous relationship of trade and nancial development.

9 Page 10 of 40

(i)

M

an

us

cr

ip t

development which captures the e ciency of nancial institutions and the access to credit markets. In contrast to standard moral hazard approaches, we assume that producers are uniformly distributed at the unit interval and are heterogeneous in (i) 2 [0; 1], which we denote the agency costs of a rm i. A higher (i) increases the private bene t and thus the incentive for misbehavior. This assumption introduces heterogeneity in credit constraints at the rm level. The agency costs (i) can be interpreted in two ways. First, the parameter may capture di erences in incentives to divert external funds. This could be the case if entrepreneurs attach di erent values to the misuse of loans. Second, a high (i) might re ect a larger scope for misbehavior of the entrepreneur as investment projects are opaque or corporate control is weak. As the success probability in case of diligent behavior is lower than one, lenders cannot observe whether the entrepreneur has misbehaved or whether the outcome is due to a bad shock. Hence, it is impossible to sign a contract contingent on the project choice.21 To prevent misbehavior of agents and thus losses from lending, investors have to ensure that the following incentive constraint holds: (i) (1

) rd(i):

(6)

te

d

At period t = 1, expected pro ts in case of production and loan repayment have to be (weakly) higher than private bene ts in case of misbehavior. Rearranging equation (6) shows that moral hazard restricts the borrowing capacity:

Ac ce p

d(i)

h

p(i)x(i) 1 + (i) 1

i : r

(7)

Firms with high agency costs (i) derive large private bene ts from diverting the loan. Hence, investors restrict credit provision to prevent misbehavior. If nancial institutions are perfect ( = 1), entrepreneurs have no incentives to misbehave and equation (6) collapses to a zero-pro t condition. In this case, di erences in agency costs (i) play no role and rms are homogenous. In contrast, if nancial institutions are imperfect ( < 1), rm-speci c moral hazard divides producers into two groups.22 Unconstrained rms For producers with relatively low (i) the nancial constraint (7) is not binding. Hence, these rms maximize pro ts (4) subject to the budget constraint (5). 21

In the Web Appendix, we extend the model by monitoring and show under which conditions our main results remain robust. 22 See the Web Appendix for a derivation of the rm's maximization problem.

10 Page 11 of 40

The optimal output of unconstrained rms is independent of (i) and given by: a

xU =

b0 eX rc : 2b0 (1 e)

(8)

cr

ip t

By inserting equation (8) into the inverse demand function (3), we derive the optimal price of unconstrained rms: a b0 eX + rc pU = : (9) 2

1

:

(10)

an

pC ( ) = cr 1 + (i)

us

Constrained rms Firms with high agency costs (i) face credit rationing and cannot behave optimally. As the nancial constraint (7) is binding, the constrained price has to be equal to the e ective marginal production costs:

te

d

M

Producing one unit of the good yields the price pC ( ) which has to compensate for the marginal production costs cr and the opportunity costs of diligent behavior cr (i) 1 . Intuitively, higher agency costs (i) lead to a larger incentive to misbehave and reduce the pledgeability of sales in the nancial constraint (7). Hence, constrained rms have to set a higher price above the marginal production costs cr to obtain external credit. A higher price of credit-rationed producers result in a lower output which is given by: a

b0 eX

Ac ce p

xC ( ) =

h cr 1 + (i) 1

b0 (1

e)

i

:

(11)

In our model, the only source of rm heterogeneity occurs in . As optimal output (8) and prices (9) do not depend on , all unconstrained producers behave in the same way. It can be shown that unconstrained rms charge lower prices and o er larger quantities compared to credit-rationed producers.

3.3

Industry equilibrium

In equilibrium, we derive a critical value of agency costs e above which rms are nancially constrained. We exploit that the nancial constraint (6) is just binding for the marginal unconstrained producer and insert the optimal output from equation (8), which leads to: e=a

b0 eX 2cr 1

rc

:

(12)

11 Page 12 of 40

xU , xC ( β )

xU

ip t

xC ( β )

cr

xC ( β = 1) ~ β

β

1

us

0

an

Figure 3: Output pro le of constrained and unconstrained rms

Condition 1 e < 1 if

te

d

M

The share of nancially constrained rms is given by 1 e, which corresponds to the fraction in empirical patterns 2 and 3. In a particular industry, a fraction e of rms is unconstrained and chooses the identical optimal output as shown in Figure 3. Following equation (11), output of constrained rms decreases in agency costs . To arrive at an output pro le as depicted in Figure 3, we impose two conditions. First, to ensure that both groups of rms occur, the threshold value e has to be smaller than one. a b0 eX cr

< 1 + 21

Ac ce p

Second, the output of the rm with the highest agency costs, Otherwise it would not be active in the market. Condition 2 xC ( = 1) > 0 if

a b0 eX cr

>1+

(i) = 1, has to be positive.

1

Inserting Condition 2 in equation (12) leads to a lower limit value for the share of unconstrained rms el = 12 . To determine the industry equilibrium, average output x e in the economy can be expressed as: x e=

Z

0

e

xU di +

Z

1 e

xC ( ) di:

(13)

Inserting the optimal outputs (8) and (11) in equation (13) and aggregating leads to: a 2 x e=

e h

b0 2 (1

h

2

e+2 1

e) + 2 12

e

01 c

e e km

i

i

cr ;

(14)

Page 13 of 40

~ x L

~ Scale: ~x (β )

ip t

m

m

~

us

Cutoff: β ( ~x )

cr

L

~ β

an

Figure 4: Industry equilibrium and trade liberalization

Ac ce p

te

d

M

R1 1 with 0c (i) di being the average agency costs within the group of constrained e e 1 producers. Figure 4 depicts the industry equilibrium. As the world economy consists of m producers in k countries and only a fraction of rms is successful, the aggregate output is given by: X = kme x. Equations (12) and (14) represent two relationships between the e endogenous variables and x e. The curve Cutof f : e (e x) illustrates equation (12) and determines the fraction of nancially constrained rms dependent on average industry output. Intuitively, the negative slope captures the fact that higher industry scale increases competition and forces more rms into the constrained status. The curve Scale: x e e is derived from equation (14) and re ects that with a higher critical value e more rms are unconstrained and thus choose optimal output levels. Hence, average industry scale increases. The intersection of the two curves in Figure 4 characterizes the industry equilibrium.

3.4

Comparative statics in partial equilibrium

The previous section has characterized the partial equilibrium in the economy. In a next step, we investigate how globalization, an exogenous change in the interest rate, and a change in the degree of competition a ect the industry. All results are derived by total di erentiation of the two equilibrium conditions (12) and (14). See Appendix 8.1 for a detailed derivation. Furthermore, we extend the model by free entry and endogenize the number of producers in the Web Appendix.

13 Page 14 of 40

2

1 |{z}

2 (1 |

Market size e ect

e e km

e) + 2 {z

M

d ln x e = d ln k

an

us

cr

ip t

Globalization Following Eckel & Neary (2010), we interpret globalization as an increase in the number of countries k in the integrated world economy. This shock a ects optimal rm behavior through two channels. On the one hand, producers face a market size e ect which corresponds to an increase in the number of consumers L. On the other hand, globalization is associated with increased competition from foreign rms. Therefore, this competition e ect works like a rise in the number of producers m. To gain intuition for the e ects of globalization, we analyze the two channels separately. From equation (3), we observe that a larger market rotates the inverse world demand outwards without a ecting the intercept. Thus, rms face a larger demand and raise output levels resulting in a one-to-one increase in industry scale. This market size e ect is counteracted but not outweighed by tougher competition. Consequently, globalization increases average industry scale:

e e km }

> 0:

(15)

Competition e ect

d

The positive market size e ects shifts the curve Scale: x e e upwards and the curve Cutof f : e (e x) outwards in Figure 4. A larger market increases the pledgeable income and thus relaxes

Ac ce p

te

the nancial constraint (6). As Figure 4 shows, the change in market size does not a ect the share of credit-rationed producers in equilibrium. However, the competition e ect leads to a partial backward shift of the two curves. A greater number of competitors producing at a larger average scale x e aggravates nancial constraints and increases the share of creditrationed rms: d ln e = d ln k

(1 |

h

e) eb0 X

(1

) cr e 2 (1

e) + 2 {z

Competition e ect

e e km

i < 0:

(16)

}

Tougher competition reduces rm revenues and therefore pledgeable income as shown by equation (7). If goods are perfectly di erentiated (e = 0), the competition e ect disappears and globalization leads to a one-to-one increase in output without a ecting the share of nancially constrained producers. Proposition 1 In partial equilibrium, globalization increases industry scale as the positive market size e ect dominates the counteracting competition e ect. The latter increases the share of nancially constrained producers (lower e). 14

Page 15 of 40

Borrowing costs In this section, we analyze the e ects of an exogenous change in the interest rate r. An increase in the borrowing costs reduces average industry scale x e and forces more producers into the constrained status: d ln x e d ln e <0; < 0: d ln r d ln r

ip t

(17)

2b0 (1

3

0 6 e7 61 + eb X d ln x 7 7 0: e)xU 4 cr |d {z ln r}5

M

cr

2

(18)

( )

d

d ln xU = d ln r

an

us

cr

In partial equilibrium, an exogenous increase in the borrowing rate leads to a higher share of nancially constrained rms and reduces industry scale (proof see Appendix 8.1). For both groups, an increase in the borrowing rate has a direct negative impact on rm outputs, whereby the e ect is stronger for credit-rationed rms. By comparing equations (8) and (11), this can be explained by the agency problem which leads to higher e ective marginal production costs for nancially constrained producers. Whereas credit-rationed rms experience strong contraction, total di erentiation of equation (8) shows a counteracting competition e ect for unconstrained rms:23

Ac ce p

te

Besides the direct negative impact of an increase in the interest rate, unconstrained producers optimally react to the reduction in industry scale by an increase of individual output. If varieties are perfectly di erentiated (e = 0), the latter e ect vanishes and unconstrained rms clearly reduce sales. However, the larger is the substitutability of goods, the more unconstrained rms bene t from reductions of rival rms' outputs. Degree of competition Consistent with our empirical motivation, the share of nancially constrained rms in equation (12) depends on nancial development and the degree of competition. We show that a higher degree of product market competition (larger e) reduces e, conditional on industry characteristics:24 @e = @e

b0 X < 0: 2cr (1 )

(19)

Proposition 2 The share of unconstrained rms e decreases in the degree of competition e (see empirical pattern 2).

x e See Appendix 8.1 for an explicit derivation of the expression dd ln ln r < 0. In Appendix 8.1 we compute the total derivative and show that the negative relationship holds when taking into account the e ect on industry output. 23

24

15 Page 16 of 40

General equilibrium

cr

4

ip t

The negative relationship between e and e corresponds to our empirical motivation. The survey question exploited in empirical pattern 2 captures the price sensitivity that a producer faces within an industry. A larger substitutability increases the degree of competition as consumers react more sensitive to an increase in prices. This is captured by the parameter e in our model.

4.1

te

d

M

an

us

The partial equilibrium analysis is based on the assumption that the interest rate is exogenously given. This implies that capital supply is completely elastic. In the next subsection, we endogenize the interest rate by introducing a simple capital market with xed supply. Our results can be interpreted as a short-run equilibrium as we abstract from endogenous entry and exit decisions of rms (see the Web Appendix for an extension with free entry). Furthermore, we do not allow for adjustments of capital supply. After trade liberalization, the borrowing rate increases caused by higher capital demand. In the long-run, this e ect might be counteracted by an increase in capital supply or capital market liberalization. In the following, we analyze how endogenous adjustments of borrowing costs a ect the implications of globalization. Furthermore, we show the impact of nancial development in general equilibrium.

Capital market clearing

Ac ce p

Each rm has to cover variable production costs by external nance and hence demands cxj (i) units of capital, with j 2 C; U . We assume that the economy is endowed with a xed amount of capital KS . In equilibrium, the inelastic supply of capital has to be equal to total capital demand KD of m rms in a country: KS = KD = cm

Z

0

e

xU di +

Z

1 e

!

xC ( ) di

= cme x:

(20)

By evaluating the equilibrium condition (20), we can explicitly solve for the interest rate: a 2 r=

e

h

2

h b0 2 (1

e) + 2

e+2 1

e

e e km i 01 c c

i

Ks cm

:

(21)

We add equation (20) to the system of equations from the partial equilibrium analysis (12) and (14). In general equilibrium, pro ts and capital income determine the aggregate income 16 Page 17 of 40

r

~ CME: r(β )

ip t

k

~

cr

CUT: β (r )

us

~ β

an

Figure 5: Globalization in general equilibrium

4.2

M

of consumers I, which is spent on the consumption of goods.25

Comparative statics in general equilibrium

Ac ce p

te

d

This section analyzes the e ects of globalization and changes in nancial development in general equilibrium. As capital market clearing pins down average industry scale x e, we express our equilibrium by two equations in the endogenous variables r and e. The curve CU T : e (r) in Figure 5 combines capital market clearing (20) with the nancial condition (12). Intuitively, the curve is downward sloping as a higher interest rate increases the share of nancially constrained rms and thus reduces the cuto value e. The curve CM E: r e is derived by inserting equation (20) into (14), and illustrates the relationship between r and e such that the capital market is in equilibrium. A higher share of unconstrained producers leads to an increase of average output and thus to higher capital demand. To ensure capital market clearing, the interest rate has to rise.26 Globalization In general equilibrium, the xed capital amount determines average industry output. Therefore, in contrast to section 3.4, globalization (an increase in k) has no e ect on industry scale: d ln x e = 0: (22) d ln k 25

A rise in the interest rate r has no e ect on aggregate income as the resulting increase in capital income is exactly o set by a decrease in rm pro ts. 26 In the Web Appendix, we show that capital demand still increases after globalization with free entry.

17 Page 18 of 40

Globalization leads to an upward shift of the curve CM E: r e in Figure 5. For a given share of nancially constrained rms, the dominating market size e ect increases capital demand resulting in a higher interest rate: d ln r > 0: d ln k

ip t

(23)

M

an

us

cr

This result is based on the assumption of xed capital supply. An increase in the interest rate occurs as long as trade liberalization is not accompanied by large capital in ows. We discuss the role of nancial policy on our results at the end of section 5.2. The curve CU T : e (r) is una ected such that the new equilibrium is characterized by the intersection point with the new capital market clearing condition. Consequently, the share of nancially constrained producers increases as higher borrowing costs impose stronger restrictions on the nancial constraint: d ln e < 0: (24) d ln k Proposition 3 In general equilibrium, globalization increases the interest rate and the share of nancially constrained rms. Proof. See Appendix 8.2.

Ac ce p

te

d

In line with our result, Foellmi & Oechslin (2015) nd that the share of nancially constrained rms increases in industries which are strongly exposed to trade liberalization. Comparing equations (16) and (24) shows that globalization leads to a stronger increase in the share of nancially constrained producers in general equilibrium (see Appendix 8.2 for a formal proof). This result is driven by the endogenous increase in borrowing costs which forces more rms into the constrained status. In contrast to partial equilibrium, the increase in the interest rate leads to di erent rm responses after globalization: d ln xU =1 d ln k

d ln xC ( ) =1 d ln k

a

cr

d ln r > 0; a cr d ln k h i 1 cr 1 + (i) d ln r h i < 0: d ln k b0 eX cr 1 + (i) 1 b0 eX

(25)

(26)

The increase in the number of countries k a ects optimal rm behavior in two opposing ways. As shown in partial equilibrium, the market size e ect dominates the competition e ect which induces rms to increase outputs. The endogenous adjustment of the interest rate in general equilibrium counteracts the positive impact of globalization. This e ect especially hurts nancially constrained producers with high agency costs (i) shown by the larger weight of the interest rate in equation (26) compared to unconstrained rms (25). 18 Page 19 of 40

xU , xC ( β )

xU A

ip t

xC ( β )

B

1

β

us

~ β

0

cr

xC ( β = 1)

an

Figure 6: Output pro les and globalization

d

M

Proposition 4 In general equilibrium, globalization leads to an output expansion among unconstrained rms, whereas nancially constrained producers have to reduce output due to increased capital costs. Proof. See Appendix 8.2.

Ac ce p

te

The expansion among unconstrained rms is illustrated in Figure 6 by an upward shift of the output pro le. In contrast, credit-rationed producers su er from increased capital costs and thus decrease output depending on their agency costs. As the most constrained rm with = 1 faces the strongest output reduction, the constrained output pro le rotates clockwise. ) The slope is given by bcr(1 0 (1 e) (compare equation (11)) and thus increases in the interest rate and the market size. The di erential responses across the two groups of producers increase the variance of output and prices within the industry. This result will be crucial for the welfare consequences which we discuss in more detail in section 5. As average industry scale is una ected due to xed capital supply, the output gain of unconstrained rms (region A in Figure 6) o sets the contraction of nancially constrained producers (region B). Financial development An increase in reduces the incentives to reap private bene ts and thus enhances the pledgeability of revenues. This shock can be interpreted as an improvement of nancial contract enforcement. Comparable to trade liberalization, there is no e ect on aggregate output due to xed capital supply. Consistent with empirical pattern 3, a higher quality of nancial institutions (larger ) relaxes the nancial constraint (6) and

19 Page 20 of 40

increases the share of unconstrained producers in the economy: d ln e > 0: d ln

(27)

cr

ip t

Furthermore, the increase in pledgeable income translates into higher capital demand and thus a higher borrowing rate: d ln r > 0: (28) d ln

us

Note that this result holds under the assumption of xed capital supply. Hence, a higher quality of nancial institutions only a ects capital demand.

an

Proposition 5 In general equilibrium, higher nancial development decreases the share of nancially constrained rms (see empirical pattern 3). Proof. See Appendix 8.2.

d

M

An improvement in the quality of nancial institutions increases the borrowing capacity of credit-rationed rms. This direct positive e ect is counteracted by an increase in capital costs. Whereas nancially constrained rms expand output, unconstrained producers do not bene t from higher nancial development, but face a higher interest rate:

cr e) xC ( )

Ac ce p

d ln xC = 0 d ln b (1

rc

2b0 (1

te

d ln xU = d ln

(i)

d ln r < 0; e)xU d ln 1+

1

(i)

(29) d ln r > 0: d ln

(30)

Consequently, an increase in nancial development induces a reallocation of market shares towards credit-rationed producers. This e ect can be seen graphically by a downward shift of the unconstrained output pro le as well as an outward rotation of the output line for constrained rms in Figure 7. Hence, higher nancial development reduces the withinindustry variance of sales, which provides a rationale for empirical pattern 4. Proposition 6 In general equilibrium, higher nancial development reduces the variance of sales within an industry as nancially constrained rms expand outputs at the expense of unconstrained producers (see empirical pattern 4). Proof. See Appendix 8.2.

20 Page 21 of 40

xU , xC ( β )

xU A

xC ( β )

ip t

B

cr

xC ( β = 1)

~ β

1

β

us

0

5

an

Figure 7: Output pro les and nancial development

Welfare

5.1

Indirect utility

te

d

M

This section analyzes how globalization a ects consumer welfare. In a rst step, we derive a welfare measure for a representative consumer. We disentangle the e ects of trade liberalization on consumer welfare. In particular, we highlight general equilibrium e ects on the variance of prices as a new welfare channel.

Ac ce p

As an appropriate measure for consumer welfare, we derive the indirect utility function for a representative consumer associated with the preference structure in equation (1). As we choose the marginal utility of income as numeraire ( = 1), indirect utility can be expressed as follows: U = km

a2 (1

e) + e km (pU + pC )2 [1 + e ( km 2b(1 e) [1 + e ( km 1)]

1)] (

2 C

+

2 U)

:

(31)

The welfare measure increases in the rst moments of prices for unconstrained and conRe R1 strained rms respectively, pU = 0 pU di, pC = e pC ( )di, and decreases in the second Re R1 moments of prices for both groups, 2U = 0 (pU )2 di and 2C = e (pC ( ))2 di. The structure of the utility function is comparable to welfare measures in general oligopolistic equilibrium models.27 In these papers, consumer welfare decreases in the variance of prices which in our 2 pj for j 2 C; U . Two important properties of the case would be de ned as 2j = 2j welfare function will be crucial for the subsequent analysis. Following from the preference 27

Compare e.g. Neary (2016), among others.

21 Page 22 of 40

structure in equation (1), consumers love variety and dislike heterogeneity in consumption levels and prices.

5.2

Welfare e ects of trade liberalization

a2

(

2 C

2b

+

2 U)

.

(32)

us

U = km

cr

ip t

We start our analysis by focusing on the new welfare channel in our model. By setting e = 0 (i.e. perfect product di erentiation), we shut down pro-competitive e ects of globalization. In this case the welfare function (31) simpli es to:

M

an

Globalization measured as an increase in the number of countries k a ects welfare in two ways. Consumers' utility increases in the number of available varieties ( km) and decreases in the second moments of prices ( 2C + 2U ). As shown in Figure 6, globalization increases heterogeneity among constrained producers. For the latter, the second moment of prices is given by: # " 2 2 3 (1 ) (1 ) 2 2 2 1 e+ 1 e + 1 e . (33) C = c r 3 2

Ac ce p

te

d

In Proposition 4, we have shown that globalization increases the interest rate (r) and reduces the share of unconstrained rms (lower e). Both e ects clearly increase the second moment of prices in equation (33) and hence reduce welfare in (32). This new channel is also present when we return to the general utility function (31) and consider pro-competitive e ects (e > 0). In the following, we simulate the changes of consumer welfare to globalization and compare results in partial and general equilibrium.28 Similar to our previous analysis, we rst consider only the market size e ect of globalization (change in the number of consumers L). Subsequently, we take into account that trade liberalization increases competition and the number of varieties available to consumers (change in k).

Market size e ect The market size e ect re ects increased export opportunities after globalization. The left panel of Figure 8 shows that a larger market has no e ect on consumer welfare in partial equilibrium (PE), but leads to welfare losses in general equilibrium (GE). This di erence is driven by the endogenous adjustment of the borrowing rate when the capital market equilibrium is taken into account. 28

We simulate the model in general equilibrium with MATLAB. The simulation code is available from the authors upon request. The following parameter values are chosen: a = 100, b = 1, m = 2, e = 0:25, c = 25, = 0:25, = 0:9, KS = 1500.

22 Page 23 of 40

4600

5400

4500

5300 5200

4400

5100 4300

5000

U

4200

4900 4800

4100

ip t

U

4700

4000 PE GE

3900

4600

1.05

1.1

1.15

1.2

1.25

1

1.05

L

cr

4500 1

1.1

1.15

1.2

1.25

us

k

an

Figure 8: Welfare e ects of market size (L) and globalization (k)

Ac ce p

te

d

M

As equation (31) shows, consumer welfare depends on the rst and second moments of prices for both groups. In partial equilibrium, an increase in the market size L leads to a proportional expansion of output among all rms without a ecting optimal price setting and the share of unconstrained rms e (compare section 3.4). Therefore, consumer welfare does not respond to changes in the market size as the rst and second moments of prices remain constant. In contrast, increased capital demand raises the interest rate in general equilibrium which leads to a higher variance of prices and thus to welfare losses. As discussed in section 4.2, higher borrowing costs increase the within-industry variance of prices in two ways. First, a larger fraction of rms becomes nancially constrained (lower e). Second, unconstrained producers expand output at the expense of credit-rationed rms. Globalization By considering the e ect of an increase in the number of countries k; we introduce two additional channels how globalization a ects consumer welfare (31). In contrast to the left graph, the right panel of Figure 8 shows that globalization leads to welfare gains both in partial and general equilibrium resulting from (i) lower prices due to increased competition and (ii) larger consumption variety. Importantly, the positive welfare e ects are considerably lower in general equilibrium. Whereas the partial equilibrium analysis re ects well-known gains from trade through competition and larger variety, our model stresses an additional negative welfare channel of globalization driven by an increase in capital costs. Whereas unconstrained rms bene t from trade liberalization due to the market size e ect, the higher interest rate especially hurts the most constrained producers (with high values of ). Compared to existing work, the negative welfare channel of a larger market is driven by two components of our model. First, the introduction of heterogeneity in nancial frictions at 23 Page 24 of 40

ip t

the rm level induces endogenous selection of producers into unconstrained and constrained groups. Second, by considering capital market clearing in general equilibrium, the interest rate is endogenized and increases with globalization. In the presence of rm-speci c credit frictions and endogenous capital costs, trade liberalization leads to a larger variance of prices and reduces positive welfare e ects.

te

d

M

an

us

cr

Policy implications The additional negative welfare channel of globalization is especially relevant if nancial development is low and credit frictions are signi cant. Thus, from a policy perspective, our model implies that trade liberalization should be accompanied by nancial reforms that aim to mitigate negative e ects. To do so, our theoretical framework suggests two potential policy measures: an improvement in the quality of nancial institutions or an increase in capital supply KS . Both measures reduce price heterogeneity and hence dampen potential welfare losses, but work through di erent channels. An increase in alleviates credit frictions and induces a reallocation of market shares towards nancially constrained producers (see the discussion in section 4.2). As a second measure, globalization should be accompanied by an increase in capital supply KS to weaken the increase in borrowing costs which bene ts all rms. Indeed, globalization is often accompanied with nancial policies that may increase capital supply. In our framework an increase in KS would reduce the interest rate and hence, counteract the general equilibrium e ect. We show to which extent capital supply would have to be adjusted to o set the e ect of globalization on the interest rate:

Ac ce p

d ln KS d ln k

r=0

=h

2 (1

2 (1

e) i > 0: e e) + 2 em k

(34)

Equation (34) shows that the required adjustment in capital supply is less than proportional to the increase in the number of countries k. By switching o the competition e ect of globalization (e = 0), the relationship becomes proportional. Hence, an increase in the market size due to globalization can only be o set by a one to one increase in capital supply.

6

Conclusion

This paper has developed a new international trade model with rm-speci c credit frictions and endogenous adjustments of capital costs in general equilibrium. A key element of our model is that credit constraints at the rm level interact with capital market imperfections at the country level. Credit frictions arise from a simple moral hazard problem, whereas rms

24 Page 25 of 40

Ac ce p

te

d

M

an

us

cr

ip t

di er in their exposure to nancial constraints. Our model is consistent with new empirical patterns from Enterprise Surveys data of the World Bank. We show that the majority of variation in exposure to nancial constraints is across rms within an industry. Furthermore, this paper highlights a positive relationship between the degree of product market competition and the share of nancially constrained rms. Additionally, our framework rationalizes a positive relationship between nancial development and the share of nancially constrained rms, as well as the variance of sales. We use this model to analyze the e ects of globalization on rm performance and consumer welfare. The main idea is that aggregate implications of trade liberalization are very di erent if general equilibrium e ects on capital costs are taken into account. In general equilibrium, we show that endogenous adjustments of capital costs represent an additional channel which reduces gains from trade. Trade liberalization increases capital demand which pushes the borrowing rate upwards. This general equilibrium e ect induces a within-sector reallocation of pro ts towards unconstrained rms at the expense of nancially constrained producers, and increases the share of credit-rationed rms. We show that these adjustments increase the variance of prices and reduce consumer welfare. A key insight of our model is that globalization leads to a reallocation of market shares from constrained to unconstrained rms. Importantly, rm heterogeneity arises from credit frictions and not from di erences in productivity. To identify this new channel empirically, further research is needed. Besides an appropriate exogenous trade shock, it is crucial to disentangle the role of credit frictions from alternative channels. Previous literature has shown that ex ante productivity di erences as well as technology adoption can lead to similar reallocations across rms. Hence, a robust identi cation of our channel requires a good measure of nancial constraints at the rm level while controlling for other determinants of rm performance. From a policy perspective, our model implies that trade liberalization could lead to negative welfare e ects and should be accompanied by nancial reforms to counteract an increase in within-industry heterogeneity across rms. This implication is especially relevant in developing countries where credit frictions are signi cant and nancial development is low.

25 Page 26 of 40

References Antras, Pol, & Caballero, Ricardo J. 2009. Trade and Capital Flows: A Financial Frictions Perspective. Journal of Political Economy, 117(4), 701{744.

ip t

Antras, Pol, Desai, Mihir A., & Foley, C. Fritz. 2009. Multinational Firms, FDI Flows, and Imperfect Capital Markets. The Quarterly Journal of Economics, 124(3), 1171{1219.

cr

Banerjee, Abhijit V, & Du o, Esther. 2005. Growth Theory Through The Lens of Development Economics. Handbook of economic growth, 1, 473{552.

an

us

Banerjee, Abhijit V, & Du o, Esther. 2014. Do Firms Want to Borrow More? Testing Credit Constraints Using a Directed Lending Program. The Review of Economic Studies, 81(2), 572{607. Beck, Thorsten. 2003. Financial Dependence and International Trade. Review of International Economics, 11(2), 296{316.

d

M

Berman, Nicolas, & Hericourt, Jer^ome. 2010. Financial Factors and The Margins of Trade: Evidence from Cross-Country Firm-Level Data. Journal of Development Economics, 93(2), 206{217.

te

Besedes, Tibor, Kim, Byung-Cheol, & Lugovskyy, Volodymyr. 2014. Export Growth and Credit Constraints. European Economic Review, 70(C), 350{370.

Ac ce p

Brooks, Wyatt, & Dovis, Alessandro. 2015. Credit Market Frictions and Trade Liberalizations. Working Paper. University of Notre Dame. Chaney, Thomas. 2016. Liquidity Constrained Exporters. Journal of Economic Dynamics and Control, 72(C), 141{154. Do, Quy-Toan, & Levchenko, Andrei A. 2004 (June). Trade and Policy Research Working Paper Series 3347. The World Bank.

nancial development.

Eckel, Carsten, & Neary, J. Peter. 2010. Multi-Product Firms and Flexible Manufacturing in the Global Economy. Review of Economic Studies, 77(1), 188{217. Egger, Peter, & Keuschnigg, Christian. 2015. Innovation, Trade, and Finance. American Economic Journal: Microeconomics, 7(2), 121{57. Feenstra, Robert C., Li, Zhiyuan, & Yu, Miaojie. 2014. Exports and Credit Constraints under Incomplete Information: Theory and Evidence from China. The Review of Economics and Statistics, 96(3), 729{744. 26 Page 27 of 40

Foellmi, Reto, & Oechslin, Manuel. 2010. Market Imperfections, Wealth inequality, and the Distribution of Trade Gains. Journal of International Economics, 81(1), 15{25.

ip t

Foellmi, Reto, & Oechslin, Manuel. 2015. Harmful Pro-Competitive E ects of Trade in Presence of Credit Market Frictions. Working Paper. University of St. Gallen. Foley, C. Fritz, & Manova, Kalina. 2015. International Trade, Multinational Activity, and Corporate Finance. Annual Review of Economics, 7(1), 119{146.

us

cr

Formai, Sara. 2013 (Nov.). Heterogenous rms and credit frictions: a general equilibrium analysis of market entry decisions. Temi di discussione (Economic working papers) 940. Bank of Italy, Economic Research and International Relations Area.

an

Gorodnichenko, Yuriy, & Schnitzer, Monika. 2013. Financial Constraints And Innovation: Why Poor Countries Don't Catch Up. Journal of the European Economic Association, 11(5), 1115{1152.

d

M

Gross, Till, & Verani, Stephane. 2013. Financing constraints, rm dynamics, and international trade. Finance and Economics Discussion Series 2013-02. Board of Governors of the Federal Reserve System (U.S.).

te

Hodler, Roland. 2011. Institutions, Trade, and the Political Economy of Financial Development. Journal of Institutional and Theoretical Economics (JITE), 167(3), 445{464.

Ac ce p

Holmstrom, Bengt, & Tirole, Jean. 1997. Financial Intermediation, Loanable Funds, and the Real Sector. The Quarterly Journal of Economics, 112(3), 663{91. Irlacher, Michael, & Unger, Florian. 2016 (Feb.). Capital Market Imperfections and Trade Liberalization in General Equilibrium. FIW Working Paper series 162. FIW. Khandelwal, Amit K., Schott, Peter K., & Wei, Shang-Jin. 2013. Trade Liberalization and Embedded Institutional Reform: Evidence from Chinese Exporters. American Economic Review, 103(6), 2169{2195. Manova, Kalina. 2008. Credit constraints, Equity Market Liberalizations and International Trade. Journal of International Economics, 76(1), 33{47. Manova, Kalina. 2013. Credit Constraints, Heterogeneous Firms, and International Trade. Review of Economic Studies, 80(2), 711{744. Melitz, Marc J. 2003. The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity. Econometrica, 71(6), 1695{1725. 27 Page 28 of 40

Melitz, Marc J., & Ottaviano, Gianmarco I. P. 2008. Market Size, Trade, and Productivity. Review of Economic Studies, 75(1), 295{316.

ip t

Minetti, Raoul, & Zhu, Susan Chun. 2011. Credit Constraints and Firm Export: Microeconomic Evidence from Italy. Journal of International Economics, 83(2), 109{125. Mu^ uls, Mirabelle. 2015. Exporters, Importers and Credit Constraints. Journal of International Economics, 95(2), 333{343.

us

cr

Mu^ uls, Mirabelle, et al. . 2008. Exporters and Credit Constraints: A Firm-Level Approach. Tech. rept. National Bank of Belgium Brussels.

an

Neary, J. Peter. 2016. International Trade in General Oligopolistic Equilibrium. Review of International Economics, 24(4), 669{698.

M

Peters, Katrin, & Schnitzer, Monika. 2015. Trade liberalization and Credit Constraints: Why Opening Up May Fail to Promote Convergence. Canadian Journal of Economics, 48(3), 1099{1119.

d

Svirydzenka, Katsiaryna. 2016 (Jan.). Introducing a New Broad-based Index of Financial Development. IMF Working Papers 16/5. International Monetary Fund.

te

Tirole, Jean. 2006. The Theory of Corporate Finance. Princeton: Princeton University Press.

Ac ce p

von Ehrlich, Maximilian, & Seidel, Tobias. 2015. Regional implications of nancial market development: Industry location and income inequality. European Economic Review, 73(C), 85{102. Wagner, Joachim. 2014. Credit constraints and exports: a survey of empirical studies using rm-level data. Industrial and Corporate Change, 23(6), 1477{1492.

28 Page 29 of 40

7

Empirical Appendix

In this section, we provide a detailed description of the data and robustness checks for patterns 1-4 as presented in the main text.

Data

ip t

7.1

an

us

cr

We use cross-sectional rm-level data from the World Bank Enterprise Surveys (WBES) for two di erent time periods. As information on competition is not available for all countries, we restrict the rst part of our analysis to a subsample from 2002-2005. To analyze the relationship between credit constraints and the variance of rm sales at the country level, we use more recent cross-section data for the years 2009 and 2013, which is available for a larger set of countries. Table 2 describes the variables used, and Table 3 shows summary statistics.

M

Table 2: Description of variables Variable Financial variables: Access to nance (Access)

Variable description

Access to nance is an obstacle to business: 0=no obstacle, 1=minor obstacle,

Financial development (FinDev )

Dummy = 1 rm answered Access to nance with 3 or 4.

te

Share constrained rms (Constrained)

d

2=moderate, 3=major, 4=very severe

A combination of indicators that capture size and liquidity of nancial markets, access to nancial services and e ciency of nancial institutions. Source: IMF.

Expected e ect of hypothetical 10% price increase of main product on demand:

Ac ce p

Degree of competition (Comp)

1=no e ect, 2=small decrease, 3=large decrease, 4=customers stop buying.

Firm-level controls: Size

Log number of workers

Labor productivity

Log sales / number of workers

Legal status

1=publicly listed, 2=private, 3=cooperative, 4=sole proprietorship, 5=partnership

Age

Number of years in business

Exporter

=1 if rm exports

Domestic private ownership

Percentage of rm owned by domestic private sector

Foreign private ownership

Percentage of rm owned by foreign private sector

Government ownership

Percentage of rm owned by government / state

Data source: WBES 2002-2005, 2009, 2013. Financial development: IMF, Svirydzenka (2016). Variables reported in local currency units are converted to 2005 U.S. dollars. Database available at: http://www.enterprisesurveys.org.

29 Page 30 of 40

Table 3: Summary statistics Variable

Obs.

Mean

Median

S.D.

Min

Max

Share of constrained rms

69,377

0.21

0.18

0.19

0

1

Degree of competition

28,620

2.63

3.00

1.08

1

14.05

13.77

2.89

-2.16

18,307

10.28

9.62

3.92

-6.91

Size

68,831

3.35

1.63

3.14

Labor productivity

54,299

10.61

10.20

3.02

Legal status

46,942

3.39

4.00

1.56

Age

69,606

17.33

11.50

Exporter

70,350

1.80

2.00

Domestic private ownership

70,818

84.33

100

Foreign private ownership

70,246

9.50

0

Government ownership

69,982

5.14

0

28.79

28.23

11.12

-6.62

26.80

1

6

18.56

0

1002.00

0.40

1

2

34.22

0

100

27.12

0

100

21.05

0

100

us

0

an

Cross section 2009

ip t

13,175

Log total assets

4

cr

Log sales

M

Cross section 2002-2005

Share of constrained rms

18,911

0.30

0

0.46

0

1

Log sales

16,903

12.84

12.82

2.56

0.27

22.65

49

0.24

0.20

0.16

0.02

0.66

3.46

3.22

1.46

0

11.07

Financial development

20,359

Legal status Age

9.37

9.41

1.89

-1.92

19.01

2.62

2.00

1.08

1

6

20,478

2.60

2.56

1.02

0

7.61

20,404

0.18

0

0.38

0

1

Domestic private ownership

19,773

85.99

100

32.46

0

100

Foreign private ownership

19,768

10.83

0

29.28

0

100

Government ownership

19,766

0.90

0

7.20

0

100

Share of constrained rms

21,067

0.24

0

0.42

0

1

Log sales

16,737

12.28

12.20

2.38

-0.81

28.35

Ac ce p

Exporter

16,854

19,353

te

Labor productivity

d

Size

Cross section 2013

Financial development

36

0.25

0.24

0.13

0.07

0.54

Size

21,384

3.15

2.89

1.33

0.00

9.39

Labor productivity

16,660

9.11

0.10

1.90

-4.79

24.43

Legal status

21,509

2.69

2

0.99

1

6

Age

21,583

2.63

2,64

0.98

0

7.61

Exporter

20,967

0,19

0

0.39

0

1

Domestic private ownership

21,364

90.59

100

27.08

0

100

Foreign private ownership

21,303

7.65

0

24.81

0

100

Government ownership

21,308

0.69

0

6.73

0

100

Source: Authors' own computations from the WBES.

30 Page 31 of 40

Table 4: Within-industry and between-industry variation of credit constraints 2-digit

3-digit

4-digit

Obs.

within

between

within

between

within

between

Chile

666

91.91

8.09

88.38

11.62

79.79

20.21

El Salvador

405

93.67

6.33

91.39

8.61

85.62

14.38

Guatemala

204

96.73

3.27

90.27

9.73

81.35

Honduras

313

92.24

7.76

90.13

9.87

84.88

Madagascar

219

77.70

22.30

67.92

32.08

58.15

Nicaragua

302

88.94

11.06

83.46

16.54

75.35

24.65

South Africa

567

89.64

10.36

82.97

17.03

74.20

25.80

Thailand

1280

95.50

4.50

95.07

4.93

94.18

Vietnam

1512

89.11

10.89

83.62

16.38

ip t

Country

18.65 15.12

us

cr

41.85

71.94

5.82

28.06

Controls: labor productivity; size; legal status; age; exporter; domestic private owner-

an

ship; foreign private ownership; government ownership. Source: Authors' own computations from the WBES, 2002-2005.

Empirical pattern 1

M

7.2

te

d

We regress a measure for credit constraints (Accessif ) on industry dummies i at three di erent levels of aggregation. Accessif is generated from a survey question that asks rms to state whether access to nancing is an obstacle to the current operations (0=no obstacle, 1=minor obstacle, 2=moderate, 3=major, 4=very severe). We include a set of rm-level controls Xif (see notes in Table 4).

Ac ce p

Accessif =

+ Xif +

i

+ "if ,

(35)

Table 4 shows the within- and between industry variation of our credit measure for a subsample of countries for which data is available.29

7.3

Empirical pattern 2

In a next step, we construct a dummy variable Constrained which takes the value of 1 if a rm reports nancial constraints as a major or very severe obstacle (values 3 or 4 of variable Access). We compute an industry mean of this dummy variable: Constrainedci . This corresponds to the share of nancially constrained rms within an industry i in country c. We relate this measure to the degree of product market competition. Empirical pattern 29

In the working paper version Irlacher & Unger (2016), we provide results for a di erent measure of credit constraints. We use tangible over total assets as a proxy and show that the majority of variation in this measure is across rms within industries rather than between industries.

31 Page 32 of 40

2 shows that industries with a higher degree of product competition are characterized by a larger fraction of nancially constrained rms. We estimate the following equation: Constrainedci =

+ Compci +

c

+

i

+ "ci .

(36)

+ Compcif + Xcif +

+

an

Accesscif =

us

cr

ip t

The variable Compci denotes the industry mean of the degree of competition (see Table 2). We control for country xed e ects c and industry dummies i .30 Column (1) of Table 5 shows results for this speci cation and highlights the positive relationship between competition and the share of nancially constrained rms. Moreover, we use the rm-level variable for access to external nance instead of the industry share in regression (36) and regress it on the rm-level degree of competition Compcif . We additionally control for rm-level variables. c

i

+ "cif .

(37)

7.4

Ac ce p

te

d

M

Column (2) shows that credit-rationing is positively associated with tougher competition. Firms that report more price-sensitive consumers face stronger credit-rationing. The advantage of the rm-level regression is that we further control for industry- xed e ects at the 4-digit level. Consistent with the existing literature (Berman & Hericourt, 2010; Minetti & Zhu, 2011; Manova, 2013), rms that are larger, more productive, and older face less obstacles to external nance. However, Table 6 shows that the correlation coe cient between productivity and access to nance is rather small ( 0:14). This indicates that productivity is not the only source of rm heterogeneity in nancial constraints.

Empirical pattern 3

We relate the e ect of nancial development to the share of nancially constrained rms. We use a measure for nancial development F inDevc that is developed by the International Monetary Fund. It captures indices of size and liquidity of nancial markets, access to nancial services, and e ciency of institutions to provide nancial services. F inDevc is highly correlated with other measures of nancial development which are typically used in the literature (e.g. private credit to GDP). However, the main advantage of this proxy lies in the multiple indicators that allow a broader assessment of the quality of nancial institutions across countries. A detailed description can be found in Svirydzenka (2016). We estimate 30

As the WBES contain cross sectional data for countries surveyed in di erent years (2002-2005), we also include year dummies in our regression.

32 Page 33 of 40

Table 5: Regression analysis credit constraints, competition and nancial development Share constrained

Access to nance

2002-2005 Degree of competition (industry)

0.065**

Degree of competition ( rm level)

ip t

(0.026) 0.060***

0.094***

(0.000)

(0.000)

cr

Firm-level controls:

-0.023**

-0.023

(0.016)

(0.460)

us

Size

Labor productivity

-0.395*

(0.045)

(0.036)

0.003

-0.029

(0.793)

(0.144)

-0.003***

-0.004**

(0.000)

(0.023)

-0.004

-0.053

(0.901)

(0.402)

an

Legal status

-0.023**

Age

Domestic private ownership

M

Exporter

0.000 (0.829)

-0.005***

-0.007***

(0.000)

(0.003)

-0.001

-0.002

(0.614)

(0.442)

te

d

Foreign private ownership

0.000 (0.726)

Government ownership

0.018

Ac ce p

Log of total assets

(0.333)

Year dummies

Yes

Yes

Yes

Country xed e ects

Yes

Yes

Yes

Industry xed e ects

No

Yes

Yes

300

15,350

4,260

0.865

0.193

0.147

Observations R-squared

Note: Industry xed e ects at 4-digit level; p-values are reported in parentheses. Robust standard errors, column (1): clustered at country level; columns (2)-(3): clustered at the country-industry level.

the following equation: Constrainedci =

+ F inDevc +

i

+ "ci ,

(38)

and present results in column (1) of Table 7. For further robustness, we analyze the e ect of nancial development on the self-reported 33 Page 34 of 40

Table 6: Correlation credit constraints and productivity

ip t

Access to nance Labor productivity -0.1374*** Sales -0.025*** Obs. 51,076 Notes: *** indicates signi cance at the 1% level.

+ F inDevc +

i

+ Xcif + "cif .

(39)

us

Accesscif =

cr

measure of credit access at the rm level Accesscif , as shown by the following regression:

Ac ce p

te

d

M

an

The advantage of this speci cation is that we control for rm-level characteristics and industry xed e ects. Column (2) of Table 7 shows that the negative relation between nancial development and the access to nance remains robust.

34 Page 35 of 40

Table 7: Regression analysis credit constraints, competition and nancial development Share constrained

Access to nance

-0.309**

-0.831***

(0.037)

(0.000)

Financial development

Size

-0.027***

Labor productivity

-0.042***

(0.003)

cr

(0.000)

ip t

Firm-level controls:

Legal status

-0.048***

us

(0.001) Age

0.016

(0.167)

Exporter

-0.107***

an

(0.000)

Domestic private ownership

0.000

(0.715)

-0.003***

M

Foreign private ownership

Industry xed e ects

R-squared

te

Observations

0.000 (0.782)

Yes

Yes

1,611

27,724

0.08

0.04

d

Government ownership

(0.001)

Note: Cross-section for years 2009 and 2013. Industry xed e ects at 4-digit level;

Ac ce p

p-values are reported in parentheses. Robust standard errors, column (1) clustered at country level; column (2) clustered at country-industry level.

7.5

Empirical pattern 4

In a last step, we analyze the e ect of nancial development F inDevc on the within-industry variance of rm sales V arianceci , whereas we restrict our analysis to industries with more than 25 rm observations. We estimate the following regression: V arianceci =

+ F inDevc + Xci + "ci .

(40)

A major concern is that the variance in sales is driven by di erences in size and productivity. Hence, we include industry means and variances of both rm size and labor productivity. Results in Table 8 show a negative e ect of nancial development on the variance of sales even after controlling for variation in size and productivity. The results remain robust when including industry xed e ects in column (2) and using the coe cients of variation instead 35 Page 36 of 40

of the variances in column (3). Table 8: Regression analysis credit constraints, competition and nancial development Within-industry variance of sales

Variance productivity

-1.771**

-1.856**

-1.819**

(0.012)

(0.013)

(0.010)

0.957***

0.954***

(0.000)

(0.000)

30.627***

cr

Coe cient of variation productivity

ip t

Financial development

(0.000)

0.116

0.102

(0.185) Variance size

(0.304)

1.170*** (0.000)

(0.000)

Industry xed e ects Observations R-squared

9.886*** (0.000)

-0.045

-0.190

0.896

(0.682)

(0.708)

(0.147)

No

Yes

Yes

1609

1609

1609

0.66

0.68

0.59

M

Industry mean size

(0.000)

1.178***

an

Coe cient of variation size

0.629***

us

Mean productivity

d

Note: Cross-section for years 2009 and 2013. Industry xed e ects at

te

4-digit level. P-values are reported in parentheses. Robust standard

8 8.1

Ac ce p

errors clustered at country level.

Mathematical Appendix Comparative statics in partial equilibrium

Proposition 2 (Competition and share of unconstrained rms) Inserting equation (14) into the share of unconstrained rms (12) and totally di erentiating yields: d ln e d ln e

=

h

h

2 cr(1eb xe ) + 4e

4 (1

0

e) + 4 e

To derive the latter expression, note that 1 Proof. To show that

d ln e d ln e

i 1 e km i < 0: 2e e km e

e2

0 c

=

R1 e

(i) di =

h < 0, it is su cient to proof that 4e

e2

2 1 e . 2

i 1 > 0. As the latter

expression increases in e, we insert the lowest possible cuto value el =

1 2

(see Condition 2

36

Page 37 of 40

+

e m

e) + 2 e m

"

2 (1 0

# e) x eb

# e e km

d ln k +

"

e e km

e

2 1

0 ck

2e

2cr 1

" h k 2

x ebd ln L #

#

0

L

"

# be xd ln x e = e d ln e

e+2 1 e 1 + 2e 1

crL d ln

"

i#

0 1 c

crLd ln r

cr

2 (1

e) + 2

#

us

+

"

2 (1

e kb

2

b

e me xd ln m

an

"

ip t

in the main body). To prove the following propositions, we totally di erentiate the two equilibrium conditions (12) and (14) and write the results in matrix notation:

The determinant of the coe cient matrix is given by: L 2 (1

e) + 2

M

= 2cr

h

1

i e e mk > 0:

d

Interest rate e ect In partial equilibrium, we analyze the e ects of an exogenous change in the interest rate r. The e ect on average industry scale x e is given by: e+2 1

te

h

h

2 (1

Ac ce p

d ln x e = d ln r

2

e) + 2

e

d ln r

2 (1

=

(41)

h i h i 2 e) 1 + 2e 1 + (1 ) 4e e 1 emk h i < 0. 2 1 e 2 (1 e) + 2 e e mk

whereby the proof of Proposition 2 ensures that

8.2

i

cr i < 0. e e mk b0 x e

The e ect on the cuto e is given by: d ln e

01 c

d ln e d ln r

(42)

< 0.

Comparative statics in general equilibrium

Totally di erentiating the system of equations results in the following matrix equation: 2h 6 6 4

2 (1

e) + 2 be m cm

i e e km b

0 21

crL 0

h

2

e+2 1 h 2e 1

e

0 1 c

i + 1 cL

0

i

3 kcL 7 7 5

37 Page 38 of 40

2 3 3 2 k 1 e 2 (1 e) x eb x ed ln x e 6 6 7 6e 7 6 e 0 5 d ln k + 4 4 d ln e 5 = 4 0 rd ln r 0 2

0 c

3

2 3 0 7 7 7 2crL d ln + 6 05 Ks d ln Ks , 4 5 1

whereas the determinant of the coe cient matrix is given by: =

2

1

e+2 1

2

01 c

e

c3 rL2 km < 0:

ip t

GE

us

cr

Proposition 3 (Globalization) The general equilibrium e ects of globalization are given by: d ln r 2 (1 e) x eb0 i >0 h = (43) (1 ) d ln k 0 e e cr 2 +2 1 c

h

1

(1

e+2 1

2

i

e) 1 + 2e 1

b0 x e

i

an

d ln e = d ln k

h

e

01 c

cr e

< 0:

(44)

2

i

e+2 1

e

01 c

d ln e d ln k

+ (1 ih 2 (1

te

2 (1

e) 1 + 2e 1

Ac ce p

h

h

GE

d

d ln e d ln k

M

Comparing the e ects on e in partial and general equilibrium leads to: =

PE

h

i 1

e2

) ekm 4e

e e km

e) + 2

i > 0;

(45)

whereas the proof in Proposition 1 ensures that the last term is positive.

Proposition 4 (Firm-level e ects of globalization) Inserting the interest rate e ect of globalization (43) into equations (25) and (26) leads to the following expressions: d ln xU =1 d ln k

d ln xC ( ) =1 d ln k

x e h xU 2

1 e+2 1

x e xC ( ) 2

h

e

01 c

2 1 + (i)

1

e+2 1

e

i

i > 0; 01 c

< 0:

(46)

(47)

1 < 1, the e ect in (46) is clearly positive. [2 e+2(1 e) 0c 1 ] Proof. In the case of constrained rms, note that xC ( ) < x e. A su cient condition for a negative e ect of globalization on constrained output is that the last fraction of expression

As xU > x e and

38 Page 39 of 40

e2

(47) is larger than one. This is the case if (i) > 1 2 . Evaluating this condition for the marginal rm with (i) = e and inserting the lower bound el leads to: (i) > 83 , which is always satis ed. Thus, the e ect of globalization is negative for all rms with (i) e.

1

e+2 1

e

0 c

e

> 0:

us

2

01 c

(49)

an

d ln r = d ln 2

cr

ip t

Proposition 5 (Financial development) The e ects of nancial development are given by: h i 2 e e 4 1 d ln e h i >0 (48) = 1 d ln 0 e e e 2 (1 ) 2 +2 1 c

cr 4 (1 e) xC ( ) 2

d

b0

2

te

d ln xC ( ) = d ln

M

Proposition 6 (Firm-level e ects of nancial development) To show that the e ect of nancial development on constrained output (30) is unambiguously positive, we insert expression (49) resulting in: (i) 2

e

e+2 1

1 e

e2

01 c

3

5 > 0:

(50)

Ac ce p

Proof. As the numerator of the term in brackets increases in e, we insert the lower bound el = 1 to show that this term is always positive. 2

39 Page 40 of 40