How Does Foreign Direct Investment Promote Economic Growth? Exploring the Effects of Financial Markets on Linkages

Transcription

1 How Does Foreign Direct Investment Promote Economic Growth? Exploring the Eects o Financial Markets on Linkages Laura Alaro Areendam Chanda Sebnem Kalemli-Ozcan Selin Sayek Copyright 2006 by Laura Alaro, Areendam Chanda, Sebnem Kalemli-Ozcan, and Selin Sayek Working papers are in drat orm. This working paper is distributed or purposes o comment and discussion only. It may not be reproduced without permission o the copyright holder. Copies o working papers are available rom the author.

2 How Does Foreign Direct Investment Promote Economic Growth? Exploring the Eects o Financial Markets on Linkages Laura Alaro Harvard Business School Sebnem Kalemli-Ozcan University o Houston and NBER August 2006 Areendam Chanda Louisiana State University Selin Sayek Bilkent University Abstract The empirical literature inds mixed evidence on the existence o positive productivity externalities in the host country generated by oreign multinational companies. We propose a mechanism that emphasizes the role o local inancial markets in enabling oreign direct investment (FDI) to promote growth through backward linkages, shedding light on this empirical ambiguity. In a small open economy, inal goods production is carried out by oreign and domestic irms, which compete or skilled labor, unskilled labor, and intermediate products. To operate a irm in the intermediate goods sector, entrepreneurs must develop a new variety o intermediate good, a task that requires upront capital investments. The more developed the local inancial markets, the easier it is or credit constrained entrepreneurs to start their own irms. The increase in the number o varieties o intermediate goods leads to positive spillovers to the inal goods sector. As a result inancial markets allow the backward linkages between oreign and domestic irms to turn into FDI spillovers. Our calibration exercises indicate that a) holding the extent o oreign presence constant, inancially well-developed economies experience growth rates that are almost twice those o economies with poor inancial markets, b) increases in the share o FDI or the relative productivity o the oreign irm leads to higher additional growth in inancially developed economies compared to those observed in inancially under-developed ones, and c) other local conditions such as market structure and human capital are also important or the eect o FDI on economic growth. JEL Classiication: F23, F36, F43, O40 Keywords: FDI spillovers, backward linkages, inancial development, economic growth. An earlier version o this paper circulated under the title FDI Spillovers, Financial Markets and Economic Development. Corresponding author: Laura Alaro, Harvard Business School, 263 Morgan Hall, Boston, MA 02163, lalaro@hbs.edu. Areendam Chanda, Department o Economics, 2107 CEBA, Louisiana State University, Baton Rouge, LA 70803, achanda@lsu.edu. Sebnem Kalemli-Ozcan, Department o Economics, University o Houston, Houston, Texas, 77204, Sebnem.Kalemli-Ozcan@mail.uh.edu. Selin Sayek, Department o Economics, Bilkent University, Bilkent Ankara Turkey, sayek@bilkent.edu.tr. We are grateul to Pol Antras, Bruce Blonigen, Santanu Chatterjee, Ron Davis, Peter Howitt, Michael Klein, Katheryn Niles Russ, Pietro Peretto, our discussants, Yan Bai and Ahmed Mobarak, and to participants at Stanord Institute or Theoretical Economics Summer Workshop, 2006 NEUDC, 2006 Midwest Macroeconomic Meetings, seminar at University o Oregon, 2006 Society o Economic Dynamics Meetings, and 2006 DEGIT Conerence at Hebrew University, or useul comments and suggestions.

3 1 Introduction There is a widespread belie among policymakers that oreign direct investment (FDI) generates positive productivity eects or host countries. The main mechanisms or these externalities are the adoption o oreign technology and know-how, which can happen via licensing agreements, imitation, employee training, and the introduction o new processes, and products by oreign irms; and the creation o linkages between oreign and domestic irms. These beneits, together with the direct capital inancing it provides, suggest that FDI can play an important role in modernizing a national economy and promoting economic development. Yet, the empirical evidence on the existence o such positive productivity externalities is sobering. 1 The macro empirical literature inds weak support or an exogenous positive eect o FDI on economic growth. 2 Findings in this literature indicate that a country s capacity to take advantage o FDI externalities might be limited by local conditions, such as the development o the local inancial markets or the educational level o the country, i.e., absorptive capacities. Borensztein, De Gregorio, and Lee (1998) and Xu (2000) show that FDI brings technology, which translates into higher growth only when the host country has a minimum threshold o stock o human capital. Alaro, Chanda, Kalemli-Ozcan and Sayek (2004), Durham (2004), and Hermes and Lensink (2003) provide evidence that only countries with well-developed inancial markets gain signiicantly rom FDI in terms o their growth rates. The micro empirical literature inds ambiguous results or the eect o FDI on irm s productivity. This literature comes in three waves. Starting with the pioneering work o Caves (1974), the irst generation papers ocus on country case studies and industry level cross sectional studies. 3 These studies ind a positive correlation between the productivity o a multinational enterprise (MNE) and average value added per worker o the domestic irms within the same sector. 4 Then comes the second generation 1 See Blomstrom and Kokko (1998), Gorg and Greenway (2004), Lipsey (2002), Barba Navaretti and Venables (2004), and Alaro and Rodriguez-Clare (2004) or surveys o spillover channels and empirical indings. 2 See Carkovic and Levine (2002). 3 See also Blomstorm (1986). 4 A multinational enterprise (MNE) is a irm that owns and controls production acilities or other income-generating assets in at least two countries. When a oreign investor begins a green-ield operation (i.e., constructs new production acilities) or acquires control o an existing local irm, that investment is regarded as a direct investment in the balance o payments statistics. An investment tends to be classiied as direct i a oreign investor holds at least 10 percent o a local 1

4 studies, which use irm level panel data. However, most o these studies ind no eect o oreign presence or ind negative productivity spillover eects rom the MNEs to the developing country irms. 5 The positive spillover eects are ound only or developed countries. 6 Based on these negative results, a third generation o studies argues that since multinationals would like to prevent inormation leakage to potential local competitors, but would beneit rom knowledge spillovers to their local suppliers, FDI spillovers ought to be between dierent industries. Hence, one must look or vertical (inter-industry) externalities instead o horizontal (intra-industry) externalities. This means the externalities rom FDI will maniest themselves through orward or backward linkages, i.e., contacts between domestic suppliers o intermediate inputs and their multinational clients in downstream sectors (backward linkage) or between oreign suppliers o intermediate inputs and their domestic clients in upstream sectors (orward linkage). 7 Javorcik (2004) and Alaro and Rodriguez-Clare (2004) ind evidence or the existence o backward linkages between the downstream suppliers and the MNE in Lithuania and in Venezuela, Chile, and Brazil respectively. 8 These results are consistent with FDI spillovers between dierent industries. The purpose o this study is twoold. First, in a theoretical ramework, we ormalize the mechanism through which FDI leads to a higher growth rate in the host country via backward linkages, which is consistent with the micro evidence ound by the third-generation studies described above. The mechanism depends on the extent o the development o the local inancial sector. Financial markets act as a channel or the linkage eect to be realized and create positive spillovers, which is consistent with the macro literature cited above that shows the importance o absorptive capacities. We are not irm s equity. This arbitrary threshold is meant to relect the notion that large stockholders, even i they do not hold a majority stake, will have a strong say in a company s decisions and participate in and inluence its management. Hence, to create, acquire or expand a oreign subsidiary, MNEs undertake FDI. In this paper, we oten reer to the MNE and FDI interchangeably. 5 See Aitken and Harrison (1999). 6 Haskel, Pereira and Slaughter (2002), or example, ind positive spillovers rom oreign to local irms in a panel data set o irms in the U.K.; Gorg and Strobl (2002) ind that oreign presence reduces exit and encourages entry by domestically owned irms in the high-tech sector in Ireland. 7 Hirschman (1958) argues that the linkage eects are realized when one industry may acilitate the development o another by easing conditions o production, thereby setting the pace or urther rapid industrialization. He also argues that in the absence o linkages, oreign investments could have limited or even negative eects in an economy (the so-called enclave economies). 8 See also Kugler (2006). 2

5 aware o any other study that is consistent with both micro and macro empirical evidence. In a small open economy, inal goods production is carried out by oreign and domestic irms, which compete or skilled labor, unskilled labor, and intermediate products. To operate a irm in the intermediate goods sector, entrepreneurs must develop a new variety o intermediate good, a task that requires upront capital investments. The more developed the local inancial markets, the easier it is or credit constrained entrepreneurs to start their own irms. The increase in the number o varieties o intermediate goods leads to positive spillovers to the inal goods sector. As a result, inancial markets allow the backward linkages between oreign and domestic irms to turn into FDI spillovers. 9 Our model also implies the existence o horizontal spillovers in the inal goods sector since the greater availability o intermediate inputs not only beneits the oreign irms but also raises the total actor productivity o the domestic irms in the inal goods sector, thus creating a horizontal spillover as an indirect result o the backward linkage. In the second hal o the paper, we use the model to quantitatively gauge how the response o growth to FDI varies with the level o development o the inancial markets. To the best o our knowledge, this paper is unique in this respect. We ind that a) holding the extent o oreign presence constant, inancially well-developed economies experience growth rates that are almost twice those o economies with poor inancial markets, b) increases in the share o FDI or the relative productivity o the oreign irm leads to higher additional growth in inancially developed economies compared to those observed in inancially under-developed economies. The calibration section, additionally, highlights the importance o local conditions such as market structure and human capital, the so-called absorptive capacities, or the eect o FDI on economic growth. For example, we ind larger growth eects when goods produced by domestic irms and MNEs are substitutes rather than complements. By varying the relative skill endowments while assuming that MNEs use skilled labor more intensively we obtain results consistent with Borensztein, De Gregorio, and Lee (1998) who highlight the critical role o 9 In our model, linkages are associated with pecuniary externalities in the production o inputs. In contrast to knowledge spillovers, pecuniary externalities take place through market transactions, see Hirschman (1958). Hobday (1995), in a case study o developing East Asia, inds many situations in which MNEs investment created backward linkages eects to local suppliers. 3

6 human capital. Theoretical models o FDI spillovers via backward linkages include Rodriguez-Clare (1996), Markusen and Venables (1999), and Lin and Saggi (2006). None o these models investigate the critical role played by local inancial markets and neither do they ocus on the dynamic eects o FDI spillovers. Instead, these are static models. Our model closely ollows Grossman and Helpman s (1990, 1991) small open economy setup o endogenous technological progress resulting rom product innovation via increasing intermediate product diversity. We modiy their basic ramework to incorporate oreign- owned irms and inancial intermediation. The standard Grossman-Helpman setting is preerred since it provides the most transparent solution. Further, models o FDI such as the ones mentioned above also use the intermediate product variety structure in a static setting, thus making it a natural choice when moving to a dynamic ramework. 10 Recently, Aghion, Howitt, and Mayer-Foulkes (2005) have modeled technology transers with imperect inancial markets in a Schumpeterian growth model. Their model is dierent than ours in the sense that they ocus on credit constraints impeding international technology transers (and hence international convergence), while we ocus on the role o inancial markets easing the credit constraints and allowing linkages between multinational irms and local suppliers in the host country to materialize. Thus, we are concerned with linkages within an economy once FDI has taken place. In a related paper that is closer to the spirit o our paper, Aghion, Comin, and Howitt (2006) develop a model that highlights the role o local savings in attracting and complementing oreign investment which spurs innovation and growth. The importance o well-unctioning inancial institutions in augmenting technological innovation and capital accumulation, ostering entrepreneurial activity and hence economic development has been recognized and extensively discussed in the literature. 11 Furthermore, as McKinnon (1973) stated, the development o capital markets is necessary and suicient to oster the adoption o best-practice technologies and learning by doing. In other words, limited access to credit markets restricts entrepreneurial development. In this paper, we extend this view and argue that the lack o development o the local 10 Gao (2005) also incorporates FDI into a growth model that closely ollows Grossman and Helpman (1991). The author neither models the role o domestic inancial markets nor relates the model to empirical evidence. 11 See Goldsmith (1969), Greenwood and Jovanovic (1990), and King and Levine (1993a,b), among others. 4

7 inancial markets can limit the economy s ability to take advantage o potential FDI spillovers in a theoretical ramework, a premise which is already supported by empirical evidence. Our results on the importance o the inancial markets thus contributes to an emerging literature that emphasizes the importance o the local policies and institutions in limiting the potential beneits that FDI can provide to the host country. 12 The rest o the paper is organized as ollows. Section 2 presents the model. Section 3 perorms a calibration exercise using values or the parameters rom the empirical literature. Section 4 concludes. 2 The Model 2.1 Households Consider a small open economy. The economy is populated by a continuum o ininitely lived agents o total mass 1. Households maximize utility over their consumption o the inal good, U t = t e ρ(τ t) log u(c τ )dτ, (1) where u(.) is a continuously dierentiable strictly concave utility unction, ρ is the time preerence parameter, and C τ denotes consumption o the inal good at time τ. The inal good, denoted by Y t, is a numeraire and is reely traded in world markets at a price p t which we normalize to 1. The total expenditure on consumption is thus given by E τ = p τ C τ = C τ. Households maximize utility subject to the ollowing intertemporal budget constraint, t e r(τ t) E τ dτ t e r(τ t) w τ dτ + A t, (2) where A t denotes the value o the assets held by the household at time t, and w τ is the wage income. The intertemporal budget constraint requires that the present value o the expenditures, E τ, not exceed the present value o labor income plus the value o asset holdings in the initial period. The solution 12 For example, lack o adequate contract and property rights enorcement can limit the interaction between oreign and local irms. A oreign irm can decide instead o buying inputs in the host country to produce them within the boundaries o the irm or import them, restricting their local activities to hiring labor. See Antras (2003), and Lin and Saggi (2006). 5

8 o this standard problem implies that the value o the expenditures must grow at a rate equal to the dierence between the interest rate and the discount rate. However, i this rate o growth o expenditure is dierent rom the endogenous rate o growth o the economy then either the transversality condition is violated or the economy no longer remains a small open economy. To rule out these possibilities, we assume that households are credit constrained and can borrow at most a ixed raction o their current income. Further, we assume that this constraint is binding, and thereore the actual rate o growth o expenditures is proportional to the rate o growth o income: Production The Final Goods Sector. E E Ẏ Y Final good production combines the production processes o domestic and oreign irms denoted respectively by Y t,d and Y t,, which are not traded. Let p t,d and p t, denote their respective prices. The aggregate production unction or this composite inal good is given by, Y t = [Y ρ t,d + µy ρ t, ]1/ρ, (3) where ρ 1 and ε = 1/ (1 ρ) represents the elasticity o substitution between Y t,d and Y t,. We do not model the decision o oreign irms to enter the market. Thereore, the aggregator o oreign and domestic irms production serves as an artiact that allows us to capture the interaction o oreign and domestic irms in an economy. 14 We can exogenously vary µ to capture realistic shares o oreign and domestic irms in the inal output. I ε =, oreign and domestic irms produce perect substitutes; ε =, they produce complements. I ε = 1, the production unction or the inal good becomes Cobb 13 This is only an assumption o convenience since, as we will see later, the entrepreneurs are also credit-constrained and we would rather treat both groups the same to rule out any gains rom arbitrage. This assumption also ensures that the consumption side o the economy has no implications or the production side. Hence, there is no dierence between assuming a household cannot borrow orever and a household cannot borrow over a certain raction. 14 For a similar setup, see Markusen and Venables (1999). 6

9 Douglas. Proit maximization yields, p t, p t,d = µ [ Yt,d Y t, ] 1 ρ. (4) The cost unction is given by, C (Y t, p t,, p t,d ) = Y t [ p 1 ε t,d ] 1 + µε p 1 ε 1 ε t,. Setting the price equal to marginal cost, 1 = [ p 1 ε t,d ] 1 + µε p 1 ε 1 ε t,, which allows us to derive an expression between the price o the domestic irm and oreign irm goods, ( p d = 1 µ ε p 1 ε ) 1 1 ε. (5) Foreign and Domestic Firms Production Processes Both oreign and domestic irms production processes combine unskilled labor, skilled labor, and a composite intermediate good. The intermediate good is assembled rom a continuum o horizontally dierentiated goods. Unskilled and skilled labor are not traded and available in ixed quantities L and H, correspondingly. Competition in the labor market ensures that unskilled and skilled wages, w t,u and w t,s, are equal to their respective marginal products. To capture the importance o proximity between suppliers and users o inputs, we assume that all varieties o intermediate goods are non-traded. 15 Domestic production is characterized by, Y t,d = A d L β d t,d Hγ d t,d Iλ t,d, (6) 15 This is a common assumption used to capture transportation costs or local content requirements; see Grossman and Helpman (1990), Markusen and Venables (1999) and Rodriguez-Clare (1996). Alternatively, one could assume that there are some intermediate goods that are tradable and others that are non-tradable. Our results will hold as long as each intermediate good enters both domestic and oreign production unctions with the same intensity. 7

10 with 0 < β d < 1, 0 < γ d < 1, 0 < λ < 1 and β d + γ d + λ = 1. L t,d, H t,d, and I t,d denote, respectively, the amount o unskilled labor, skilled labor, and the composite intermediate good used in domestic production at any instant in time, and A d represents the time invariant productivity parameter. Foreigners directly produce in the country rather than license the technology. The industrial organization literature suggests that irms engage in FDI not because o dierences in the cost o capital but because certain assets are worth more under oreign than local control. I lower cost o capital were the only advantage a oreign irm had over domestic irms, it would still remain unexplained why a oreign investor would endure the troubles o operating a irm in a dierent political, legal, and cultural environment instead o simply making a portolio investment. Graham and Krugman (1991), Kindleberger (1969), and Lipsey (2003) show that investors oten ail to bring all the capital with them when they take control o a oreign company; instead, they tend to inance an important share o their investment in the local market. An investor s decision to acquire a oreign company or build a plant instead o simply exporting or engaging in other orms o contractual arrangements with oreign irms involves two interrelated aspects: ownership o an asset and the location to produce. 16 First, a irm can possess some ownership advantage a irm-speciic asset such as a patent, technology, process, or managerial or organizational know-how that enables it to outperorm local irms. And this is one o the reasons why researchers ail to ind evidence o horizontal spillovers since this means that a oreign irm will seek to use this special asset to its advantage and prevent leakages o its technology. Hence, we model potential beneits rom FDI as occurring via linkages and not through technology spillovers. Second, domestic actors, such as opportunities to tap into local resources, access to low-cost inputs or low-wage labor, or bypass taris that protect a market rom imported goods can also lead to the decision to invest in a country rather than serve the oreign market through exports. 17 Since our objective in this paper is to understand the eects o oreign production on local output and the role o inancial markets, and not the decision to invest abroad, we model the rictions o doing 16 This approach to the theory o the multinational irm is also known as the OLI ramework ownership advantage, localization, internalization. See Dunning (1981). 17 For models that endogenize the FDI decision, see Helpman (1984), Markusen (1984), and Helpman, Melitz, and Yeaple (2004). 8

11 business in the domestic economy with the parameter φ. 18 Thus, oreign irms use the ollowing Cobb Douglas production unction, Y t, = A φ Lβ t, Hγ t, Iλ t,, (7) with 0 < β < 1, 0 < γ < 1, and β +γ +λ = 1. Like beore, L t,, H t,,and I t, denote, respectively, the amount o unskilled labor, skilled labor, and the composite intermediate good used in oreign production at any instant in time, and A represents the time invariant productivity parameter. Unskilled and skilled labor have dierent shares within the domestic and oreign production, though the total labor share is assumed to be the same across both types o irms. This relects the common observation that the share o labor tends to be around two-thirds o total actor payments while at the same time permitting dierent skill intensities within domestic and oreign production. A corollary o assuming the same total labor share is, γ γ d = β d β. (8) The composite intermediate good is assembled rom dierentiated intermediate inputs. Following Ethier (1982), we assume that, or a given aggregate quantity o intermediate inputs used in the inal good production, output is higher when the diversity in the set o inputs used is greater. This speciication captures the productivity gains rom increasing degrees o specialization in the production o inal goods. [ n 1/α I t,d = I t, = I t = xt,idi] α, (9) where x t,i is the amount o each intermediate good i used in the production o the inal good at time t, and n is the number o varieties available. Let p i denote the price o a variety i o the intermediate good x. The CES speciication imposes constant and equal elasticity o substitution (1/(1 α)) between a pair 0 18 Burnstein and Monge-Naranjo (2005) assume taxes on oreign irms to be the barrier in each country. We allow a broader interpretation, as oreign irms need to bear a wide range o costs/risks o doing business abroad, including sovereign risk, taxes, and inrastructure and dealing with dierent institutions and cultures. We also considered an alternative scenario where MNEs receive a net price p /φ where φ > 1, relecting these disadvantages, obtaining similar results. 9

12 o goods. Each variety o intermediate good enters the production unction identically and the marginal product o each variety is ininite when x t,i = 0. This implies that the irm will use all the intermediate goods in the same quantity, thus x t,i = x t. Let X t = n t x t be the total input o intermediate goods employed in the production o the inal good at time t, then we can rewrite I t = n 1 α α t X t. Domestic production is given by, 19 Y d = A d L β d d Hγ d d Xλ d n λ(1 α) α, (10) and oreign production by, Y = A φ Lβ Hγ d Xλ n λ(1 α) α. (11) Thus, raising the varieties o intermediate inputs n, holding the quantity o intermediate goods constant, raises output productivity. Using the cost unction and the act that in a symmetric equilibrium all intermediate goods are priced similarly, p i = p x, we can write the equilibrium conditions or the domestic and oreign irms respectively as, p d = A 1 d β β d d λ λ γ γ d d w β d u w γ d s p λ xn λ(α 1) α, (12) p = φa The Intermediate Goods Sector β β λ λ γ γ w β u w γ s p λ xn λ(α 1) α. (13) The intermediate goods sector is characterized by monopolistic competition. There exists an ininite number o potential varieties o intermediate goods, but only a subset o varieties is produced at any point in time as entrepreneurs are required to develop a new variety. Since the set o potential intermediate goods is unbounded, an entrepreneur will never choose to develop an already existing variety. Thereore, variety i o x is produced by a single irm which then chooses the price p i to maximize proits. Firms 19 Since we will ocus exclusively on the balanced growth path, we omit the time subscript or the rest o the paper. 10

13 take as given the price o competing intermediate inputs, the price o the inal good, and the price o the actors o production. In a symmetric equilibrium all intermediate goods are priced similarly, p i = p x. Hence, proit maximization in every time period or each supplier o variety i implies, max π i = p x x i c x (w u, w s, x i )x i, (14) where c x (w u, w s, x i ) represents the cost unction and x i = x d + x is the sum o the demand or the intermediate product i by domestic and oreign irms respectively. Production o intermediate goods requires both skilled and unskilled labor according to the ollowing speciication, x i = L δ x i H 1 δ x i. (15) Hence, the cost unction or the monopolist is given by, c (w u, w s, x i ) = δ δ (1 δ) (1 δ) w δ uw (1 δ) s x i. (16) Proit maximization yields the result that each variety is priced at a constant markup (1/α) over the marginal cost. 20 Hence, the price o each intermediate good is given by, p x = δ δ (1 δ) (1 δ) wδ uw s (1 δ). (17) α The raction that domestic irms spend on all intermediate goods is given by the corresponding share in the production unction, λp d Y d. This implies that or each intermediate good, the amount spent by domestic irms is given by λp d Y d /n. Similarly, the amount that oreign irms spend on these goods is given by λp Y /n. The sum o amounts spent by oreign and domestic irms is the total revenue o the intermediate producer, 20 See Helpman and Krugman (1985), chapter 6. 11

14 Thereore, we can write the operating proits per irm as, p x x i = λp dy d + λp Y n n. (18) π i = (1 α) n [λp d Y d + λp Y ]. (19) What is the value o introducing new intermediate goods and thus the value o the monopolistic irm? Let v t denote the present discounted value o an ininite stream o proits or a irm that supplies intermediate goods at time t, v t = t e r(s t) π s ds. Equity holders o the irm are entitled to the stream o uture proits o the irm. They make an instantaneous return o (π t + v),. (proits and capital gain). They can also invest the same amount in a risk-ree bond and receive return rv t (the prevailing market interest rate). Arbitrage in capital markets ensures that, π + v = rv π +. v v = r. (20) Thus, the rate o return o holding ownership shares is equal to the interest rate Introduction o New Varieties and Financial Markets In order to operate a irm in the intermediate good sector, entrepreneurs must develop a new variety o intermediate goods. The introduction o each new variety requires some initial capital investment according to the ollowing speciication, ṅ = K a nθ. (21) 21 Note that the arbitrage condition does not contradict our assumption o credit-constrained households since they can choose to lend to irms or invest in a risk ree bond. 12

15 In contrast to Grossman and Helpman (1991), who assume that new varieties are developed with two inputs, labor and general knowledge, we opt or one input only, capital, to simpliy the analysis. 22 Our main results do not depend on this simpliying assumption. The main implication o this simpliication is that our results are less dependent on the production parameters o the innovation sector. This has important advantages or our calibration exercise, as the stylized acts o the innovation and imitation processes are not well documented. Our central argument is that entrepreneurs ace diiculties in obtaining, or example, loans to set up irms and this prevents the creation o backward linkages even under the presence o FDI. Assuming only capital is used or these setup costs then allows us to ocus better on this issue. The introduction o a new variety depends on the existing stock o varieties. We introduce the parameter θ since this allows a more general production structure. A value o θ < 0 suggests a ishing out eect (increasing complexity in introducing new varieties) while a value o θ > 0 implies positive externalities ( standing on the shoulder o giants ). At this stage, we do not postulate an exact value o θ. 23 This will be pinned down by conditions required to satisy balanced growth. Finally, a can be viewed as the level o eiciency in the innovation sector. The initial capital investment must be inanced by borrowing rom domestic inancial institutions. The domestic inancial system intermediates resources at an additional cost, as in Edwards and Vegh (1997). This cost relects the level o development o the domestic inancial markets where lower levels o development are associated with higher costs. These costs maniest themselves in a higher borrowing rate, i which is greater than the lending rate, r. As King and Levine (1993a) mention, this wedge could relect taxes, interest ceilings, required reserve policies, high intermediation costs due to labor regulation, or high administration costs, etc. This simpliication allows us to ocus on the main theme o the paper: the role o inancial markets in allowing FDI beneits to materialize. Thus, this assumption should 22 Grossman and Helpman (1991) assume that the greater the stock o general knowledge among the scientiic community, the smaller the input o human capital needed to invent a new product. They assume ṅ = KL/a, where K represents the stock o general knowledge capital and not physical capital like in our model. Hence, in our model, in incurring setup costs, intermediate irms do not compete or labor inputs against the inal good sector. 23 For more on the implications o these alternative assumptions, see Jones (1995). While we do not postulate an exact value, we will work under the assumption that θ 1. Also or convenience, the discussion will treat θ as positive. 13

16 be regarded as a shortcut to more complex modelling o the inancial sector. The reader is reerred to the appendix or a cost veriication approach ollowing King and Levine (1993b) that yields similar implications. Our qualitative results will be the same under this ramework. Given n θ and a, i an entrepreneur wants to introduce one variety at any instant in time, the amount o capital needed will be K = a/n θ, so that ṅ = 1. Thereore, the cost o introducing a new variety is ia n θ. (22) There is ree entry into the innovation sector. Entrepreneurs will have an incentive to enter i ia n θ < v. However, this condition implies that the demand or capital will be ininite, which cannot be an equilibrium solution. Hence, we can rule out this condition ex-ante. I, on the other hand, ia n θ > v, entrepreneurs will have no incentive to engage in innovation. This possibility cannot be ruled out exante but would lead to zero growth. Thereore, in equilibrium, i there is growth in the number o varieties it must be the case that, ia = v i ṅ > 0. (23) nθ This also implies, v v = θ ṅ n, i.e., more innovation reduces the value o each irm. Using this expression and the arbitrage condition in the capital markets, π v + v v = r, we can rewrite equation (20) as, π v θ ṅ = r. (24) n Using irm proit equation (19), equation (23), and equation (24) we obtain, In order to simpliy, we deine (1 α) λ ia [ pd Y d n 1 θ + p ] Y n 1 θ θ ṅ = r. (25) n Y d n 1 θ = Ỹd and Y n 1 θ = Ỹ as eiciency units o outputs and get, 14

17 ṅ (1 α) λ ] = [p d Ỹ d + p Ỹ r n θia θ (26) As it is standard in this class o models, the growth rate o varieties, ṅ/n, ultimately pins down the growth rate o both domestic output and oreign output and thus aggregate output as well. 2.3 General Equilibrium and the Balanced Growth Path Using the eiciency unit adjusted output levels, Ỹd and Ỹ, we can also rewrite equation (4) as, And we can rewrite equation (21) as, p p d = µ [ Ỹd Ỹ ] 1 ρ. (27) where K is the capital stock per eiciency unit. ṅ n = 1 K a n 1 θ = 1 a K, Substituting the price or intermediate inputs (17) into the equilibrium conditions or the domestic (12) and oreign production (13), and deining eiciency wages or both skilled and unskilled labor ) ) as w s = w s / (n (1 α)λ α and w u = w u / (n (1 α)λ α, or the domestic and oreign sector, respectively, we obtain the ollowing expressions, p d = A 1 d β β d d λ λ γ γd d ( ) λ δ δ (1 δ) (1 δ) w β d+δλ u α w γ d+(1 δ)λ s (28) p = φa 1 β β λ λ γ γ ( δ δ (1 δ) (1 δ) α )λ w β +δλ u w γ +(1 δ)λ s (29) Equilibrium conditions in the labor market imply that the labor employed by the domestic, the oreign, and the intermediate goods production processes add up to the total labor supply in the economy. This implies, or skilled and unskilled labor, respectively, 15

18 L d + L + nl x = L, (30) H d + H + nh x = H. (31) Using the cost unctions or the domestic, oreign and intermediate goods sector and Shephard s Lemma, we can rewrite these two constraints as, (β d + δαλ) p d Y d w u (γ d + (1 δ)αλ) p d Y d w s + (β + δαλ) p Y w u = L, + (γ + (1 δ)αλ) p Y w s = H. as, Using our output and wage equations in terms o eiciency units, we can rewrite both constraints (β d + δαλ) p d Ỹ d n 1 θ w u n (1 α)λ α + (β + δαλ) p Ỹ n 1 θ w u n (1 α)λ α (γ d + (1 δ)αλ) p d Ỹ d n 1 θ + (γ + (1 δ)αλ) p Ỹ n 1 θ w s n (1 α)λ α w s n (1 α)λ α = L, = H. ) Along the balanced growth path, labor shares must be constant. This means that (p i Ỹ i n 1 θ / w j n (1 α)λ α (i = d, ; j = u, s) should be constant. There are two ways to achieve this. The irst is to assume that prices, p i, grow at the rate ( (1 α)λ α (1 θ) ) ṅ n. This assumption still implies that p d/p would be constant since none o these parameters relect sectoral dierences and the relative price would thus be driven by other actors (mainly ρ and µ). An alternative would be to impose (1 α)λ α = 1 θ. Although this assumption has the disadvantage o being a knie-edge condition, on the other hand, it does oer a big advantage. It allows us to back out the value o θ, or which empirical measures are not available (measures or λ and α on the other hand are easily available rom the literature). This condition still implies that p d /p would be constant. Thereore, we assume that this condition holds. Thus, we can rewrite the above as, 16

19 (β d + δαλ) p d Ỹ d + (β + δαλ) p Ỹ = L, (32) w u w u (γ d + (1 δ)αλ) p d Ỹ d w s + (γ + (1 δ)αλ) p Ỹ w s = H. (33) Now we can solve or all the endogenous variables and derive the equilibrium balanced growth. In order to be able to solve the equilibrium growth rate o varieties, ṅ n, we need to solve the set o prices } {p d, p, w u, w s } and the outputs o the domestic and oreign sectors, {Ỹd, Ỹ. To solve or the prices and the outputs, we use equations (4), (5), (28), (29), (32) and (33). These equations can be solved in a sequential order. The details are presented in the appendix. While we can solve or the FOCs and derive implicit relationships, because o equation (5), we cannot derive explicit solutions or the endogenous variables in terms o the parameters. Our model exhibits some o the standard properties o product variety-based endogenous growth. Combining the above with equation (26), ṅ (1 α) λ ] = [p d Ỹ d + p Ỹ r n θia θ we can see that, the scale eect is very much present larger endowments (L and H) will lead to larger output and thus to higher growth rate. Furthermore, higher λ, which is the share o intermediate input costs in inal output production, also drives up the growth rate o n by the same reasoning. Similarly, lower substitutability among intermediate goods (α) increases the growth rate since it raises the proitability o new intermediate goods. An increase in either, A, or µ, will lead to a reallocation o resources away rom the domestic irm to the oreign irm. Thereore, the instantaneous eect will be a decline in domestic irms share in output. In the long run, both domestic and oreign irms will beneit rom the higher growth rate. However, in the short-run, the horizontal spillovers in the inal goods sector, which indirectly result rom the backward linkages between the oreign irm and the intermediate goods sector, exist only or the surviving domestic irms. This is an additional contribution o our setup, which can shed light on why empirical studies ail to ind evidence o positive horizontal 17

20 spillovers or developing countries and even ind negative spillovers in some cases. 24 Moving on to the role o the inancial markets, one can see that the lending rate and the borrowing rate have negative eects on the growth rate. The negative eect o the lending rate r is standard a higher r relects a greater opportunity cost o investing in a new variety and thus reduces the growth o varieties. The negative eect o the higher borrowing rate, i, is more novel. It relects the higher per unit cost o initial capital investment (because o ineiciencies in inancial markets) and thus also unambiguously reduces the growth rate. I we restrict ourselves to the special case o where the aggregator is a Cobb Douglas unction (or perect substitute case), we can solve explicitly or all the endogenous variables. We turn to these next to get a sense o the qualitative implication o the model. The Special Case o Cobb Douglas Production Function The Cobb Douglas case is CES with ρ 0. We can rewrite the aggregator or the domestic and oreign output as, Y = Y 1 1+µ d µ 1+µ Y, (34) Y = Y η d Y 1 η, (35) where µ = (1 η) /η or simpler notation. Note that proit maximization here implies that p d p = η Y. (36) 1 η Y d 24 While our model is very much in the spirit o traditional endogenous growth models, more recently there has been a trend to move towards models that suggest a long run exogenous growth rate with endogenous growth in transition (e.g. see Aghion et al (2005)). This class o models implies, that in the long run, dierences in inancial markets or extent o FDI would be relected in transition paths or dierences in relative income levels instead. We have not adopted this scheme a) because ocusing on relative income levels abstracts rom some o the dynamic spillovers that are interesting when one talks about FDI and growth, b) the product variety models do not easily lend themselves to endogenous growth in transition and exogenous growth in the steady state. 18

21 As or the cost unction and equilibrium conditions, C(y, p d, p ) = η η (1 η) (1 η) p η d p(1 η) Y = 1, p d = η (1 η) (1 η) η (1 η) η p. (37) Recalling the arbitrage condition, (1 α) λ ia Using (36), the previous expression as, [p d Ỹ d + p Ỹ ] θ ṅ n = r. (1 α) λ ia (1 η) p Ỹ θ ṅ = r. (38) n We can solve the model completely, using equations (28), (29), (32), (33), (36), and (37). The details are worked out in the appendix. The main equation o interest is (68) in the appendix, renumbered here as (39) ( ) ( (1 α) λ ia w s H η (γ d + (1 δ)αλ) + (1 η) (γ + (1 δ)αλ) ) θ ṅ = r, (39) n ( ) ( where w s H = ΥΛ η d Λ1 η (ΦL) Ψ β H Ψγ ; Υ = η η (1 η) (1 η) ; Λ d = A d β β d d γγd d ; Λ = A β β γγ ) /φ; Φ = η(γ d+(1 δ)αλ)+(1 η)(γ +(1 δ)αλ) ; Ψ η(β d +δαλ)+(1 η)(β +δαλ) β = η (β d + δλ) + (1 η) (β + δλ) ; Ψ γ = η (γ d + (1 δ)λ) + (1 η) (γ + (1 δ)λ). There are a couple o conclusions one can draw rom the Cobb Douglas case. First, higher productivity o either domestic or oreign irms raises the growth rate in the economy since, all else being equal, a higher A and a higher A d both tend to raise growth. The equation or w s H above, is also increasing in both L and H. Essentially, an increase in any o these our exogenous parameters increases 19

22 the proits rom introducing new intermediate goods and thus causes the growth rate to rise. However, note that even i A increases relative to A d (i.e. a higher productivity gap between the domestic and the oreign producer) this will not increase the share o the oreign producer in the market. This is the clear drawback o using a Cobb Douglas speciication. Second, the eect o a higher share o oreign production, (1 η), on aggregate growth is ambiguous. The ambiguity can be attributed to the term, Υ = η η (1 η) (1 η) which is a U-shaped unction o η that is minimized at η = 0.5. For most o the countries in the world η > 0.5 and or most developing economies it would be near to 1. Thereore, even i there is a productivity gap between the domestic and the oreign producer, the term Υ, which is independent o this gap, could drive down the growth rates. The Special Case o Perect Substitutes CES aggregators allow or inite elasticities o substitution. But what i the two outputs were perect substitutes? Perect substitutes is o course a special case o the CES with ρ 1 (and or simplicity assume that µ = 1 as well). In this situation, it is easier to bypass the aggregator (since both products will have the same price and are indistinguishable) and assume that they are traded in the international market with the world price normalized to 1. One might wonder i the domestic sector would survive at all given the technological superiority o the oreign irms. However, note that the production unction parameters or the domestic and oreign irms are dierent allowing or both irms to co-exist while setting the marginal costs equal to the price o the inal good, since the relative marginal costs are not completely driven by productivity dierences. The explicit solution or the perect substitute case is worked out in the appendix. Next, we turn to the calibration exercises, where by using empirical estimates o our parameters we quantitatively study the comparative static eects we have discussed so ar. 20

23 3 Calibration Exercise The purpose o the calibration exercise is to study the quantitative growth eects o FDI, ocusing on dierent levels o inancial market development. We begin with a description o the parameters used in the analysis. Financial Development: We group countries based on their inancial market development levels. Dierent measures have been used in the literature to proxy or inancial market development. The broader inancial market development measures, such as the monetary-aggregates as a share o GDP and the private sector credit extended by inancial institutions as a share o GDP, capture the extent o inancial intermediation; interest rate spreads, on the other hand, capture the cost o intermediation. Given that the spread between the lending and borrowing rates better captures the spirit o our model, we preer it as the measure or the development o the inancial markets. 25 We ind that the alternative measures o inancial market development, such as the size o the inancial market, the share o private sector credit in total banking activity, and the overhead costs are all highly correlated with interest rate spreads. Hence, dierent measures yield similar results. 26 The average spread or the low inancially developed (poor) countries, medium inancially developed (middle income) countries and the high inancially developed (rich) countries between 2000 and 2003 are 14.5%, 8.5%, and 4.5%, respectively. Elasticity o Substitution: In our model, ρ relates to the elasticity o substitution between goods produced by oreign and domestic irms. Evidence regarding the appropriate choice o the elasticity o substitution parameter ρ is sparse, given that such depiction o inal goods production is an artiact to capture the interaction between oreign and domestic irms. The evidence that is closest to the spirit o our model is rom the consumption literature that uses a constant elasticity o substitution utility unction between varieties o domestic and oreign goods, or between tradable and nontradable goods. Ruhl (2005) provides a detailed overview o the Armington elasticity, i.e., the elasticity o substitution between oreign and home goods, and inds that an appropriate value or ρ is around 25 Erosa (2001) deines the inancial intermediation cost as the resources used per unit o value that is intermediated, which is the total value o inancial assets owned by the inancial institutions. He measures the inancial intermediation cost as the spread between the lending and borrowing rates. 26 Alternative measures are rom Levine et al. (2000). 21

24 While our benchmark analysis is based on the CES production unction with ρ = 0.2, we also undertake robustness analysis by allowing ρ to vary between 0.9 and 0.9. In section 3.2, we present the quantitative characteristics o the model or the Cobb Douglas case (ρ = 1). The share o intermediate goods in the production o the inal good (λ) is assumed to be the same across the two production technologies. The ormulation o the production technology allows setting the share o the intermediate goods equal to the share o physical capital in inal goods production. Following Gollin (2002), we set this share equal to 1/3. The remaining 2/3 is accounted by skilled and unskilled labor. The remaining parameters used in the benchmark analysis are chosen such that those or the domestic irm capture the characteristics o the production technologies available in the developing countries; whereas, those or the oreign irm capture the characteristics o the production technologies available in the industrial countries. Domestic Firms: According to Weil (2004), the share o wages paid to skilled labor is 49% or the developing countries. We take this value to be that o domestic irms, suggesting that o labor s 2/3rd share in inal goods production, 49% is due to skilled labor. Thereore, we set the share o skilled labor in domestic irms, γ d, at 32%. In parallel, the share o unskilled labor in domestic irms, β d, is set at 35%. For the benchmark analysis, we set the total actor productivity A d equal to 1. Foreign Firms: The share o skilled and unskilled labor costs in output o the oreign irm is calculated in a similar ashion. Following Weil (2004), the share o wages paid to skilled labor is taken as 65% in industrial countries. Accordingly, the share o skilled labor in the oreign irm s production, γ, is set equal to 40%. Similarly, the share o unskilled labor, β, is set equal to 27%. Thus, γ > γ d. 28 As a benchmark, the productivity o the oreign irm, A, is initially set to be twice that o the domestic irm ollowing Hall and Jones (1999), who show the productivity parameter or a very large sample o non-industrial countries is around 45% o the productivity parameter o the U.S. With respect to the 27 A wide range o estimates are available rom trade and business cycles literatures ranging between 0 and 0.5. Ruhl (2005) argues that a model with temporary and permanent trade shocks can replicate both the low elasticity o substitution igures used by the international real business cycle studies and the high elasticity o substitution values ound by the empirical trade studies. Such an encompassing model justiies a value o ρ around As Barba Navaretti and Venables (2004) note, there is ample evidence that oreign irms employ more skilled personnel than domestic irms. They also tend to be larger, more eicient, and pay higher wages. 22

25 cost o doing business that the oreign irms ace, φ, our benchmark case is one where there is no such cost. However, note that an increase in the cost o doing business is equivalent to lower productivity o oreign irms. Thus, by considering variations in relative productivities, we can also iner implications or the variations in cost o doing business. Share o Foreign Production: The share o oreign production to total output is not exogenous in the CES production unction case and the choice o µ implicitly determines this share. As such, the benchmark value or µ is determined to allow or the matching o the relative output values to the real data. Lipsey (2002) estimates that in 1995 the share o world production due to FDI lows was at best 8%. 29 Keeping this in mind, we set µ = 0.1 as our benchmark value since, as we shall see later, it produces a share o approximately 6%. In the Cobb Douglas production unction case, we round o the share o oreign irms in total output to 5% (i.e. η = 0.95). Intermediate Goods Sector: Based on the work o Basu (1996), the mark-up is assumed to be 10%, and hence the value o the reciprocal o (1+mark-up) is given by α = Given the lack o any estimate, the share o unskilled labor in the production o the intermediate goods, δ, is taken as 0.5. The Stock o Skilled and Unskilled Labor: H and L, respectively, are set ollowing Duy, Papageorgiou, and Perez-Sebastian (2004). The authors argue that there is an aggregation bias caused by dierences in terms o eiciency units o the dierent types o labor. To overcome this bias, they weigh the length o education by the returns to schooling, and compute what they call weighted labor stock data. We calculate averages o their data or the low inancially developed (poor) countries, medium inancially developed (middle income) countries, and the high inancially developed (rich) countries. Accordingly, we set the ratio o unskilled labor to skilled labor equal to 12 or the poor countries, 9 or the middle income countries, and 5 or the rich countries. To rule out the possibility o scale eects driving dierences in growth rates, we assume that H + L = 1. The shares o the two actors are allocated according to these three ratios so that they sum to 1 (e.g., or poor countries H = and L = 0.923). 29 Mataloni (2005) inds that oreign owned companies were responsible or 12% o GDP in Australia, 5% in Italy, 7% in Finland, 19% in Hungary, and 22% in the Czech Republic. 23