Tradability and market penetration costs: Explaining foreign market servicing intensities

Tradability and market penetration costs: Explaining foreign market servicing intensities

International Review of Economics and Finance 22 (2012) 190–200 Contents lists available at SciVerse ScienceDirect International Review of Economics...

317KB Sizes 1 Downloads 116 Views

International Review of Economics and Finance 22 (2012) 190–200

Contents lists available at SciVerse ScienceDirect

International Review of Economics and Finance journal homepage: www.elsevier.com/locate/iref

Tradability and market penetration costs: Explaining foreign market servicing intensities Katherine N. Schmeiser a,⁎, Miguel F. Ricaurte b a b

Mount Holyoke College, Department of Economics, 50 College Street, 115 Skinner Hall, South Hadley, MA 01002, United States Central Bank of Chile, Research Department, Agustinas 1180, Santiago, Chile

a r t i c l e

i n f o

Article history: Received 15 April 2011 Received in revised form 6 October 2011 Accepted 10 October 2011 Available online 17 October 2011 JEL classifications: F12 F23

a b s t r a c t Industry level data shows striking differences among sectors in ratios of exports to FDI sales. We identify the elements behind the sectoral differences in the mode of foreign market servicing in the context of a general equilibrium model of monopolistic competition. Our calibration exercise shows that traditional margins such as transportation, fixed entry costs, utility weights, and dispersion of firm productivity are not enough to capture the observed sectoral differences, as is commonly assumed. We propose augmenting the model to allow for sectoral differences in intangible costs of operating in a foreign market in order to explain these observations. © 2011 Elsevier Inc. All rights reserved.

Keywords: Trade Gravity FDI Monopolistic competition Foreign market access

1. Introduction In this paper we identify an intersection of two strands of the international trade literature, tradability and gravity, that has the virtue of explaining the endogenous (and large) sectoral differences in firms' choices of serving a foreign market through either exports or a foreign affiliate (henceforth, FDI sales). In the tradability literature, Helpman, Melitz, and Yeaple (2004) and other authors argue that models of monopolistic competition can capture firms' choices of foreign market servicing. In particular, these authors use reduced-form regression analysis to show that sectoral differences in the form of servicing can be explained by product tradability (i.e., how easy it is to actually trade/ship a product), fixed setup costs, and firm productivity dispersion. Neverthe-less, while reduced form modeling of firms' behavior accurately matches the patterns in the data, structural, general equilibrium modeling fails to do so. This is where the “gravity” models of FDI, which Mitze, Björn, and Gerhard (2010) summarize, appear as a viable complement to the tradability literature. Very much like their trade counterparts, these models can be employed to explain overall geographic patterns of investment flows. However, these and other authors report that estimates of such equations fail to accurately match FDI data patterns when sectoral differences, empirically treated as sector-specific effects, are not taken into account. We take the next step, combining the two literatures to model the choice of exports and FDI including such sectoral differences as cross-sectoral variations in the difficulty of foreign market penetration — which embraces intangible costs beyond traditional transportation costs. We derive a general equilibrium model of monopolistic competition where firms choose between the

⁎ Corresponding author. E-mail addresses: [email protected] (K.N. Schmeiser), [email protected] (M.F. Ricaurte). 1059-0560/$ – see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.iref.2011.10.011

K.N. Schmeiser, M.F. Ricaurte / International Review of Economics and Finance 22 (2012) 190–200

191

options of exports and FDI. Calibrating the model, we find that a measure of tradability in conjunction with costs in the spirit of gravity can determine the sector-level ratios of exports to FDI sales. We call these new costs foreign market penetration costs. The study of this problem requires acknowledging the fact that firms' interactions with each other can be analyzed from different angles. Traditionally foreign direct investment (FDI) has been defined as either “horizontal” (when the entire production process is duplicated in order to sell goods abroad) or “vertical” (fragmenting the production process across national borders). This dichotomous choice has been shown to be empirically too restrictive (for example, see Feinberg and Keane, 2006 and Neary, 2009) as firms tend to participate in more hybrid forms of FDI hence, we abstract from horizontal or vertical FDI. Following this, for our purposes we suggest sales can be represented within a horizontal and intra-firm FDI framework. Fig. 1.1a shows the relationship between the transportation cost (which represents freight and tariff costs, as explained in Section 3) and the ratio of exports to FDI sales from the U.S. to Canada (omitting the outlier sector of Petroleum and Coal Products). It is evident that, while sectors with higher transport cost tend to be FDI sales intensive, no definite relationship between the form-of-service ratio and the transportation cost exists. Moreover, Fig. 1.1b shows there is no direct relationship between the transportation cost and total sector sales. It is evident that the explicit transportation cost of goods is not sufficient to explain the sectoral differences reported. Additionally, we find that alternative sectoral differences explored in the empirical literature (productivity dispersion, fixed costs, and consumer utility weights) perform less successfully in a theoretical framework. This suggests that alternative sectoral costs such as marketing costs, contracting concerns, and government incentives might be important. We show that with a broader definition of foreign market operation costs – which we refer to as the difficulty of foreign market penetration – the explanatory power of models of monopolistic competition is improved. Most similar to our theoretical approach, Helpman et al. (2004) extends the Melitz (2003) model to allow firms to choose between servicing the “local” market only and servicing local and foreign markets, the latter done through two alternative channels: exports and foreign affiliates (henceforward referred to as FDI). Within the FDI choice literature, however, there are many who acknowledge the importance of sectoral determinants. Most notably, Antràs (2003) constructs an incomplete contracting model where final goods firms decide whether to purchase components from a vertically integrated supplier or at arm's length. This is done to explain why the share of intra-firm imports in total U.S. imports is significantly higher the higher the capital intensity of the exporting industry. Yeaple (2006) further shows that a higher degree of sectoral productivity dispersion can be associated with more FDI within that sector. 1 Finally, sectoral differences have been shown to be a key component in explaining FDI patterns in the gravity literature. Empirical estimates of gravity-type equations fail to accurately match FDI data patterns when sectoral differences in foreign investment, treated as sector-specific effects, are not taken into account. For example, Toubal, Kleinert, and Buch (2003) use firm level German data for the 1990–2000 period to empirically analyze regional and sectoral patterns of FDI using this approach. They find that sectoral differences are key to enhancing the explanatory power of their models. In particular, while varying set up costs explain some of the difference in FDI reported across some sectors, other less tangible differences (such as cultural proxies, regulations, etc.) matter in some cases. By analyzing the export- and FDI-intensity differences across sectors employing elements from gravity and tradability, we are able to theoretically reconcile our model with the empirical facts. It has to be noted that we remove ourselves from the traditional sense of physical gravity (i.e., distance) by looking only at the U.S. and Canada. This allows us to focus on sectoral costs that are influential in servicing decisions. Nevertheless, we allow for intangible elements in the spirit of gravity, alongside differences in the transportation cost, to drive patterns in sectoral differences in servicing intensities. The rest of the paper is organized as follows. In Section 2, we develop the model. Section 3 describes the data used in the numerical experiments discussed in Section 4. Finally, we draw concluding remarks in Section 5. 2. Model We build on Helpman et al. (2004), which uses differences in fixed setup costs, as well as marginal and transportation costs to induce different choices among firms. 2 We opt for the interpretation that exports and FDI are mutually exclusive, based on evidence such as that presented by Blonigen (2001) using Japanese data. Moreover, due to recent evidence that the focus on type of FDI is less important than the distinction of intra-firm versus arm's length trade, we model the firm choice of FDI as an intra-firm interaction. Bernard, Jensen, and Schott (2005) document that 46% of U.S. firms in 1993 were involved in at least 25% of their trade as intra-firm, accounting for about 90% of multinational export value. Similarly, using data on U.S. multinational trade with their affiliates in Canada, Feinberg and Keane (2006) find that 69% of firms are “hybrids” which cannot be classified as purely horizontal or vertical multinationals. This hybrid result is further supported by Neary (2009). Hence, we diverge away from papers such as Markusen (2004) who analyzes the choice between horizontal and vertical FDI (relying on legal conditions such as copyright, patent, and imitation protection laws) and Antràs (2003) and focus on the decisions to keep trade within (FDI) or outside (exports) the firm. Our model is consistent with Helpman et al. (2004) in that: 1) among firms which choose to service the foreign market, the most efficient ones engage in FDI and the least efficient ones in exports; 2) Firm level heterogeneity in productivity adds an 1 At a more disaggregated level, Gleason, Lee, and Mathur (2002) find that even firm specific factors such as firm history and future intentions are important in determining FDI of U.S. firms into China. 2 Nocke and Yeaple (2007) extend the discussion to consider the choice between mergers and acquisitions (M&A) and greenfield FDI, which is outside the scope of this paper.

192

K.N. Schmeiser, M.F. Ricaurte / International Review of Economics and Finance 22 (2012) 190–200

Fig. 1.1. Relationship between transportation cost, exports, and FDI.

important dimension to the tradeoff between exports and FDI: ceteris paribus, sectors with higher dispersion in productivity have lower relative export sales (and higher FDI sales); 3) when the transportation cost varies between sectors, sectors with high transport costs have lower relative export sales. We show, however, that without extending the analysis to incorporate further notions of tradability and market penetration costs, the model cannot account for the large differences in intensities. This is an n-country model of differentiated firms making foreign market servicing decisions. We not only model the marginal production decisions, but the choice of whether to serve a destination through either exports or FDI. However, for now we abstract from contractual decisions, copyright laws, imitation risk, and other issues that might influence the setup decision. The choice of whether to act as a pure exporter or multinational is influenced by costs of trade, both observed and unobserved. Firms choose whether to produce domestically and whether and how to service a foreign country. Firms that service foreign countries may do so either through exporting or through a multinational subsidiary. As Feinberg and Keane (2006) point out, multinationals are neither purely horizontal nor vertical, but are hybrids that perform many tasks, including exporting back to parent companies. In our model, we do not explicitly model the decision of affiliates to re-export either intermediate or final goods back to their parents. This allows us to concentrate on identifying the driving forces behind the choice of trade versus FDI sales. Additionally, in our calibration exercise, we employ data that does not allow us to distinguish this type of exports from others. 3 This set up is equivalent to thinking of firms as participating in either arms-length or intra-firm foreign economic activity.   ρ 1ρ Consumers in country i in period t aggregate goods ω according to standard CES utility function where Q it ¼ ∫ω xit ðωÞ dω  ρ−1 ρ ρ with aggregate price P it ¼ ∫ω pi ðωÞρ−1 dω : In each country there is a continuum of firms producing differentiated goods. In our benchmark model, each firm draws two distinguishing qualities: 1) productivity level φ, where higher φ implies higher productivity and φ ∈ (1, ∞) is distributed Pareto; 2) iceberg transportation cost τ ∈ (1, ∞) which represents the measurable ease of transportation. The probability of observing τ is q(τ). As is common in the literature, firms only pay a τ > 1 when exporting their goods. A firm producing good ω is identified by the pair (φ, τ) and τ identifies the sector the firm participates in. Firms that enter the domestic market (d) must pay a fixed entry cost fe prior to observing their realizations of φ and τ. If they remain in the market, they pay a fixed operational cost fd, pay domestic labor wages, and transportation costs are normalized to 1. These firms may service foreign markets by choosing between exporting (x) and FDI (m). When firms choose to serve a foreign market, they have to pay an additional fixed entry cost which differs for each option of foreign market servicing (fx, to engage in exports; fm, to engage in hybrid- multinational operations). An exporting firm pays the iceberg transportation cost τ on all exported goods while a multinational firm pays no transportation cost but pays the foreign wage. Given there is no uncertainty in our model, once a firm enters the market, it will never choose to exit. There is, however, an exogenous firm death rate (δ ∈ (0, 1)) that bounds the value of entering the market. Labor is the only factor of production. The technology is determined by labor input functions that are linear in output



y τ þ f k ; for k ∈ fd; m; xg; τx > 0; τd;k ¼ 0: φ k

3 This also means that our export or FDI sales data may be over- or underestimating the actual export and FDI sales. However, we presume that this bias is randomly distributed in data collection across sectors and hence, does not systematically drive the differences in export/FDI sales ratios.

K.N. Schmeiser, M.F. Ricaurte / International Review of Economics and Finance 22 (2012) 190–200

193

For simplicity we assume international markets are segmented to preclude arbitrage of goods from occurring. Hence, we exclude the option to re-export. Firm (φ, τ) then maximizes profits in each market independently, resulting in an equilibrium parallel to that of Melitz (2003). For a draw of φ and τ, a firm chooses the best servicing option according to highest profit (produce only for the domestic market, export, or sell to affiliate). Given prices {P i}, wages {w i}, and our assumption on market segmentation, in each country i, good type (φ, τ), and servicing option k ∈ {d, x, m}, goods markets clear when: i

i

yk ¼ xk : Let Ωki be the set of (φ, τ)-firms in country i that are actively producing for market k ∈ {d, x, m}. We assume that all countries have the same distributions for τ and φ. Thus, the labor market clearing condition in country i, pinning down the number of firms, is: Li ¼ ∫ðφ;τÞ∈Ωi lid ðφ; τÞdGðφÞdF ðτ Þ þ ∫ðφ;τÞ∈Ωix lix ðφ; τÞdGðφÞdF ðτ Þ d

þ ∑j ∫ðφ;τÞ∈Ωj lim ðφ; τ ÞdGðφÞdF ðτÞ þ Me f e

ð2:1Þ

m

where Me is the mass of entrant firms. ⁎ The solution to this model can be represented by the aggregate price level {P} and three cutoff functions: {φd⁎(τ), φ⁎ x (τ), φm (τ)}. The cutoff functions φk⁎(τ) come from the zero profit conditions for each servicing scheme. Hereafter, for notational simplicity (and since firms belong to sectors identified by τ), we write all functions of τ as: zτ(⋅) ≡ z(τ ; ⋅).4 !1−ρ i ρ w ; ¼ Pρ !1−ρ i wi f x ρ wτ  ; φx;τ ¼ Pρ ð1−ρÞR !1−ρ j wj f m ρ w  : φm;τ ¼ ð1−ρÞR Pρ  φd;τ

wi f d ð1−ρÞR

ð2:2Þ

Notice we also assume that countries are symmetric, which implies prices and other aggregate variables (where R = P ⁎ Y, and Y is country income) are common across countries. These cutoff functions segment firms into the different options of production. The segmentation within each transportation cost type τ, discussed below, is as follows: the lowest productivity firms, upon paying the entry cost and realizing they are unprofitable, chooses not to produce. The firms with intermediate productivity will only serve the domestic market. The firms with highest productivity choose to serve the foreign market. This is a standard assumption in the literature (for example, see Helpman et al., 2004). We also assume that for each transportation cost τ, among the firms servicing foreign markets, the least efficient choose to export and pay this cost and the most efficient engage in FDI sales. This regularity has been empirically reported for a number of countries: Buch, Kleinert, Lipponer, and Toubal (2005) and Wagner (2006) find this is the case on German firms and Head and Ries (2003) find the same for Japanese exporting firms. We explore variations of this assumption. For these assumptions over the productivity cutoffs to follow, we must impose certain conditions on fixed costs, which can be trivially derived from (2.2):

fm

ρ ρ−1 f x > f d ðτÞ ⇔φd;τ bφx;τ ;∀τ ! ρ wi wi 1−ρ   ⇔φx;τ bφm;τ ;∀τ: > fx j τ w wj

ð2:3Þ

From 2.2, we notice that for τ2 higher than τ1 (which implies the good is more costly to trade), fewer firms will find it profitable to export and the cutoff φx,⁎ τ1 b φx,⁎ τ2. 4 This is especially useful when we switch to discrete values for τ later on. Then zτ(ω) = z(φ, τ) is not a function of τ, but identified according to each good ω and corresponding τ.

194

K.N. Schmeiser, M.F. Ricaurte / International Review of Economics and Finance 22 (2012) 190–200





~ k;τ φk;τ be the average productivity of firms with index of tradability τ, conditional on engaging in activity k (domestic Let φ production, exports, or FDI sales): 2 31−ρ ρ   ρ 1  ∞ 1−ρ 4 ~    φ k;τ φk;τ ¼ ∫φ φ g ðφÞdφ5 ; k;τ 1−G φk;τ

ð2:4Þ

  ~ k;τ . ~ k;t φk;τ as φ where φ*k, τ is the productivity of the marginal entrant with τ. For notational simplicity, we will refer to φ Productivity φ is drawn randomly from a Pareto distribution with:  a b aba CDF : GðφÞ ¼ 1− ; and PDF :g ðφÞ ¼ aþ1 ; φ φ where b is the lower support of the distribution and a determines the shape of the distribution. Given this distribution, for a firm engaged in k ∈ {d, x, m} with tradability index τ, (2.4) becomes: ~ k;τ ¼ φ

 φk;τ

!1−ρ ρ a : ρ a− 1−ρ

ð2:5Þ

It is trivial to show that average profit for firms with index τ engaged in market servicing k ∈ {d, x, m} calculated this way is equal to the profits of the average producer with that tradability index servicing market k:   ~ k;τ ; k∈fd; x; mg; π k;τ ¼ πk;τ φ In the following sections, we describe the data used in calibrating this model and then show that sectoral transportation costs and other common margins are not enough to explain variations in export/fdi intensities. Instead we will need to elaborate on a concept of market penetration costs which incorporates gravity-type attributes of trade. 3. Data Because bilateral FDI and exports have been extensively studied by the gravity literature, from this point on we will concern ourselves with explaining sectoral intensities. By focusing on one bilateral trade relationship, the United States and Canada, we are able to more clearly understand the underlying product specific costs that might otherwise be lost in a gravity framework. We test a simplified two-country version of the model presented above. To do so, we require multi-sector data in four categories: (1) bilateral trade; (2) bilateral FDI sales; (3) total employment and non-production labor, and (4) indices of transportation costs. All data is in SIC (rev. 1987) format for the United States and Canada in 1997. There are two reasons we choose this year. The first one is the discontinuity in industrial classification systems: the FDI sales data was recorded in SIC prior to 1999, and in NAICS format thereafter. The second reason is that 1997 provided data for the largest number of sectors. Conveniently, since the model can be treated as static, there is no need for multiple years. We organize the industry-level data into the 20 sectors presented in Table 3.1. We describe the data below. 5 We use import and export data from Feenstra, Romalis, and Schott (2002) for the U.S. and Canada for 1997. The import data used is the custom value of imports (millions of U.S. dollars) and is used to weight tariff and freight data (see below). The export data we use, also in millions of U.S. dollars, is the value of exports from the U.S. to Canada. The FDI sales data used comes from the U.S. Bureau of Economic Analysis (2006). We use data on sales by all foreign affiliates by sector as well as country. In particular, we look at U.S. affiliates operating in Canada. Bernard et al. (2005), Feinberg and Keane (2006), and Neary (2009) motivates our interpretation of our FDI sales data. Because a majority of trade, especially between the U.S. and Canada, is classified either as horizontal or within-firm we use our data on FDI sales as a proxy for intra-firm activities. We use total employment by sector as well as non-production workers by sector from the 1997 Economic Census of Manufacturing published by the U.S. Census Bureau (2001). The latter is constructed from the reported data such that: non-production labor equals total employment minus production workers. Of the 20 sectors for which we have data on exports and FDI sales, we only have employment for 15. The excluded sectors are: Instruments and Related Products; Construction, Mining, and Materials Handling Machinery; Other Electrical Equipment, Appliances, and Components; Other Petroleum and Coal Products; and Other Chemicals and Allied Products (see Table 3.1).

5

More details are available upon request.

K.N. Schmeiser, M.F. Ricaurte / International Review of Economics and Finance 22 (2012) 190–200

195

Table 3.1 Industry level data. Industry

SIC (Rev. 1987) Codes

τ

Grain Mill and Bakery Products Other Food and Kindred Products Tobacco Products Textile Products and Apparel Lumber, Wood, Furniture, and Fixtures Paper and Allied Products Chemical Products, nec Soap, Cleaners, and Toilet Goods Other Chemicals and Allied Products

2041+2051+2053 20-(2041+2051+2053+2082) 21 22+23 24+25 26 2819+2869+2879+2899 2841+2842 28-(2813+2819+2821+2822) -(2869+2879+2899+2841) -(2842+2879) 2999 29-2911-2999 301+302+305+306 321+322+323 32-(321+322+323)

1.0638 1.0741 1.1208 1.1420 1.0550 1.0519 1.0676 1.0532 1.0829

Petroleum and Coal Products, nec Other Petroleum and Coal Products Rubber Products Glass Products Other Stone, Clay, and Other Nonmetallic Mineral Products Primary and Fabricated Metals Construction, Mining, and Materials Handling Machinery Household Audio and Video, and Communications Equipment Electronic Components and Accessories Other Electrical Equipment, Appliances, and Components Instruments and Related Products

Exports

FDI sales

Total U.S. employ.

Non-prod. workers

45 4467 24 2881 3252 3054 3874 1240 3514

3106 10,075 698 1414 4600 7560 2497 2438 11,399

357,543 1,109,413 33,594 1,338,136 570,034 574,274 882,645 126,446 nr

128,286 226,363 9153 224,379 93,809 134,178 370,493 48,692 nr

1.0660 1.0552 1.0676 1.0742 1.1280

122 49 2104 1013 744

10 12,844 2306 341 2035

107,625 nr 202,353 128,565 372,906

36,071 nr 41,081 23,514 89,059

33+34 353

1.0455 1.0385

11,337 3046

6815 1283

2,368,857 nr

562,252 nr

365+366

1.0360

3417

1309

294,865

159,552

367 36-363-365-366-367-369

1.0176 1.0448

4523 4881

1530 3921

593,802 nr

161,529 nr

38

1.0351

4973

1410

nr

nr

Exports and FDI sales in millions of U.S. dollars. Employment in thousands. nr = data not reported. Sources = see Section 3.

For our calibration, we use data on freight costs and tariffs to construct a transportation cost satisfying τ > 1. This data is collected by Schott in SIC rev. 1987 format according to year and sector. The tariff rates are import-weighted, implicit averages, calculated for each year as: tariff ¼

duties : customs value

The freight rates are also import-weighted for each year as: freight ¼

cost insurance freightðcif Þ imports : free on board ðfobÞ imports

Thus, the iceberg transportation cost τ is defined by: τ ≡ ðfreightÞð1 þ tariff Þ;

ð3:1Þ

where a higher index implies the good is more costly to trade. We calculate these indices by sectors according to the concordance described in Table 3.1 (through import-weighting). It is important to note there are limitations in the interpretation of the transportation cost calculated. Notice the variation in transport costs is fairly low. This results from our choice of countries (the U.S. and Canada) in order to focus on the importance of other sectoral attributes. This low variation is caused in part from participation in NAFTA which calls for low and fairly unified tariffs. Freight costs, measured similar to Hummels (2001) and Helpman et al. (2004), provide the remaining variation, but again are an imperfect measure of transportation costs. Because this proxy of transportation cost is not ideal, the majority of the literature has turned towards gravity models to further pin down the per-unit cost faced by trading firms. We take a slightly different approach and, focusing on two countries for which transport costs are fairly uniform, we explicitly model the remaining sectoral variations and show that bilateral costs beyond measures of tariff and freight are crucial.

196

K.N. Schmeiser, M.F. Ricaurte / International Review of Economics and Finance 22 (2012) 190–200

4. Numerical experiments In this section, we take our benchmark model, derived above, and examine the explanatory power the model has when productivity and transportation costs drive firm servicing decisions. We then extend the model to allow for a more general concept of tradability – market penetration costs – and show the usefulness of incorporating this into our models. 6 Recall that we identify sectors by their average transportation cost τ as constructed from the data. Moreover, we assume the marginal exporter in each sector has that τ. Alternative approaches to sector classification can be considered however they are outside of the scope of this paper. For expositional simplicity and to focus on sectoral differences, we assume the two countries are symmetric. In the case of autarky, it is trivial to calculate the (endogenously determined) mass M of operating firms. The transportation cost τ is not relevant here, so we can simplify notation. Using the equilibrium quantities that firms produce, we have that for ~ any two firms with productivity φ and φ: xðφÞ ¼ ~Þ xðφ

 1 ~ ρ−1 φ : φ

ð4:1Þ

We can use the above to re-write the total quantity produced domestically and aggregated by the distributor:  ρ  1 ~ Þ 1−ρ pðφ i i ρ ρ Q ¼ ∫φ Mx ðφÞ dφ ⇒M ¼ : P Hence, M is the autarky mass of operating firms in any given country. To calculate the ratio of exports to FDI sales, we construct total sector sales in the model. We fix the probability qτ of being in sector τ as the fraction of that sector's total exports and FDI sales over total sales for the sectors, as found in the data: qτ ¼

fdi salesτ þ exportsτ : ∑s ðfdi saless þ exportss Þ

Additionally, we choose the fixed costs of production for each sector, to match the measure of firms that operate in each sector and mode of servicing. To see this, consider the measure of firms exporting in sector τ:

Mx;τ

   að1−ρÞ 1−G φx;τ ρ M f −a   qτ M⇒ x;τ ¼ d qτ τ ¼ M fx 1−G φd

which relies only on parameters. Then the ratio of aggregate sector exports to FDI sales is:   ~ x;τ M rjx;τ φ exportsτ   x;τ ¼ ¼ fdisalesτ r j Mm;τ ~ φ m;τ m;τ

f x;τ f m;τ

!1−a1−ρ ρ

τ

−a

:

ð4:2Þ

where Mj, t is the mass of firms doing activity j (exports or FDI) in period t. In our benchmark model, the fixed costs are assumed to be county-specific (as opposed to sector-specific). There are two things to note in Table 4.1: the death rate δ and the shape of the Pareto distribution a. Gibson (2006) uses a δ of .05. Eaton, Kortum, and Kramarz (2008) and Gibson (2006) use a Pareto shape a = 1.5 and find it is consistent with French data. While both of these papers use trade models of monopolistic competition, neither allow firms the additional option to participate in FDI. To account for alternative servicing options (and because we don't have appropriate data to measure these for exporting and FDI firms), we decrease δ and increase a. It will be clear that the exact values of these parameters do not drive our results. We use modifications of the parameters shown in Table 4.1 for the reminder of the paper. We perform 4 calibration exercises: 1) sector-specific fixed costs, 2) sector-specific productivity dispersion, 3) sector-specific utility weights, and 4) sector-specific foreign market penetration costs. For each of these exercises, we use the parameter values from Table 4.1 modified so that the variables of interest in each exercise are chosen so that sectoral Export/FDI sale ratios in the model match, if possible, those in the data. Additionally, fixed costs (when relevant) are simultaneously chosen to preserve the relationships in Eq. (2.3). The success of each model is then based on a) its ability to match Export/FDI sales, b) its ability to predict total exports and total FDI sales, and c) its ability to predict labor usage. As in Helpman et al. (2004), we compare non-production workers in the data to fixed FDI costs resulting from the model. Our goal is to assess the power of trade costs in determining whether a sector is export- or FDI-intensive. As one might expect, our baseline model fails to match export/FDI sales. Hence, we begin exploring alternative models. We first consider the ability of 6

Algebraic derivations and numeric results for all calibration exercises are available upon request.

K.N. Schmeiser, M.F. Ricaurte / International Review of Economics and Finance 22 (2012) 190–200

197

Table 4.1 Baseline parameters. Parameter

Value

Description

Target/source

ρ w τ fd fx fm δ^ a b

0.5 1 – 1.3 1.5 2.2 0.03 2 1

Utility function parameter Wages Transportation cost Fixed domestic cost Fixed export cost Fixed FDI cost Death rate of firms Shape parameter Pareto distrib. Lower support Pareto distrib.

(Gibson, 2006) Normalized Table 3.1 Eq. (2.3) Eq. (2.3) Export/FDI sales See text See text See text

All labor units are in thousands. Targets are described in the discussion.

sector-specific fixed costs to explain the variations in the data. From Eqs. (2.3) and (4.2), it is trivial to see that in this variant of the model, calibrating fixed costs does not suffice in matching the data. This is because fm would be so small that no incentive would exist to export, i.e., we would violate φ*x, τ b φ*m, τ. Secondly, we consider the effects of sector-specific productivity dispersion in determining the choice of exports and FDI. In this variant of the model, we test the claim from Helpman et al. (2004) that there is a high cross-sectoral relationship between productivity dispersion and firms' choices of foreign market servicing. Specifically, we allow for a, the shape parameter of the Pareto distribution, to vary among sectors. Eq. (2.5) requires the following restriction on parameters: ρ −aτ b0;∀τ: 1−ρ

ð4:3Þ

Additionally, from (4.2), recall the equation for the export-to-FDI sales ratio is: f x;τ f m;τ

exportsτ ¼ fdi salesτ

!1−a

1−ρ τ ρ

 a 1 τ ; τ

and the fact that f m ≥f x

wi wj

wi τ wj

!

ρ 1−ρ

is needed for consistency with empirical observations of firm productivity cutoffs. For the limit case where the above expression is satisfied with equality, (4.2) reduces to: ρ exportsτ ¼ τρ−1 : fdi salesτ

where we applied symmetry across countries. Because τ > 1, we can see that: ρ

exportsτ ¼ τρ−1 b1; 1. for ρ b 1, the model can only match FDI sales-intensive sectors (i.e., with lower exports than FDI sales): fdi salesτ 2. for the model to match export-intensive sectors (i.e., with higher exports than FDI sales) it would have to be the case that ρ > 1, ρ exportsτ violating the restriction that ρ b 1 and even here, the model would fail. fdi ¼ τρ−1 > 1. salesτ

Hence, it is evident that under the assumptions made, this variant will fail to replicate the intra-sectoral variability in the exports-to-FDI sales ratio as observed in the data. ρ  i 1−ρ i w Now consider the case where f m > f x w τ . Notice that this inequality implies: wj wj exportsτ ¼ fdi salesτ



fx fm

1−a1−ρ ρ

τ

−a

ρ

ρ

> τ ρ−1 :

ð4:4Þ

With ρ b 1 as we assume, τρ−1 b1, such that the ratio of exports to FDI sales could be smaller or larger than 1. Initially, it seems that the story of sector specific productivity dispersion alone can provide a good fit for the data. However, this is

198

K.N. Schmeiser, M.F. Ricaurte / International Review of Economics and Finance 22 (2012) 190–200

where discontinuities appear between empirical and theoretical work. Simulations show the Pareto distribution has a small enough tail that the sectoral variation found in the data can not be replicated in the model. While some alternative productivity distribution, allowing for a larger tail, would allow the model to fit the data, there has been much research (see (Eaton et al., 2008) for an example) that shows that the Pareto distribution provides a good approximation of firm productivity. Hence, while the sectoral productivity dispersion story is empirically promising, it alone is not enough to explain the export/fdi sales in this theoretical framework. For the next model specification, following (Nocke and Yeaple, 2007) we allow for consumers to value products differently through sector-specific weights in the utility function. This allow consumers to perceive differences in goods produced domestically and internationally as well as weight products differently. Hence, these weights allow for consumers to have some measure of home bias and/or allows them to perceive quality differences in goods. In this specification, relative utility weights on products are calibrated to again match the sectoral values of exports/FDI sales. While this model is successful in matching these ratios, it fails to match the absolute sectoral sales or relationship between non-production labor and fixed FDI cost. This is because the fixed cost rules in Eq. (2.3) would need to be violated (and the fixed FDI costs resulting from the calibration are negatively correlated to non-production labor in the data). Finally, we address the issue of the difficulty of foreign market penetration in addition to transportation costs. In the model variations explored above, one restricting factor is that τ cannot be b 1. This is conceptually and empirically obvious as an iceberg transportation cost logically can not be less than one, as a ship would arrive with more goods than it departed with. Additionally, taxes and subsidies, as well as physical transportation costs, do not vary as greatly as would be necessary to match data. If sectoral transport costs varied such that τ ∈ (0, ∞) then we could use any of the above variations to match the data. This would in effect mean that governments greatly subsidize exports (which is not generally the case) or that shipping costs are negative! As this is not a realistic alternative, we propose a measure of the difficulty of foreign market penetration. This embraces all intangible costs associated with doing business (either through exports or FDI) in a foreign country. For our purposes, while this may include country specific features such as non-tariff aspects of trade agreements, language barriers and cultural differences, our focus is on industry-specific features such as advertising barriers, managerial skills, foreign product development, contractual and legal issues, etc. In this sense, the difficulty of foreign market penetration truly is a sector specific barrier that is not restricted to exporters (as are transportation costs and tariffs), but affects both multinationals and exporters (albeit differently). This also allows for taxation exemptions that multinational firms might receive from foreign governments and contractual difficulties that might lead firms to prefer one servicing option over another depending on a product's imitation qualities. We model this sector-specific difficulty of foreign market penetration as the relative cost of an activity through trade or FDI: στ = σm, τ/σx, τ for each sector identified by τ. As before, sector-specific fixed costs are such that φ⁎x, τ b φ⁎m, τ is satisfied. Difficulty of foreign market penetration affects a firm's ability to collect on revenue, so profits are now j

j

j

j j

πk;τ ðφÞ ¼ σ k;τ pk;τ ðφÞyk;τ ðφÞ−w lk;τ ðφÞ: The more difficult it is to provide a good to a foreign market, the more revenue must be paid to trade with exports relative to j

w FDI and the higher a price will be charged pk;τ ¼ ρσ k;τ φ . Higher στ means that it is relatively easier to service through FDI than k;τ

exports, so we expect a lower export/FDI sales ratio. As in Eq. (4.2), we then derive the equilibrium ratio of exports to FDI sales: exportsτ ¼ fdi salesτ

f x;τ f m;τ

!1−a1−ρ  ρ

1

σ ρτ τ

−a

:

ð4:5Þ

From our calibration algorithm, we present the difficulty of foreign market penetration index values in Table 4.2. This model variant predicts total sector exports and FDI sales which are very close to the observed values. The model also closely matches other moments in the data, in particular, total and overhead costs measured as total and non-production employment. Regarding the former, total sector employment arising from the model and observed employment show a positive correlation of 0.845. Additionally, when we follow (Helpman et al., 2004) and compare non-production workers to fixed FDI costs, we find a positive relationship of 0.49. Hence, the model makes an accurate description of sectoral differences in a number of margins. This further suggests that the literature needs to assess the entire scope of the impact of unobservable costs of foreign servicing of goods and services. While empirically, the importance (but not determinants) of these unobserved costs in explaining both trade and FDI flows has been identified, no such thing has been done for the choice of exports and FDI. Explicit modeling of the difficulty of penetration idea embedded in the στ parameter is a simple approach which can theoretically support empirical findings. 5. Conclusions This paper had the objective of explaining the large variations of exports to FDI sales among sectors (also referred to as export- or FDI sales-intensities) observed in the data. While the trade literature has advanced significantly in this direction, we show that sectoral determinants traditionally used fail to account for the observed variations. We do so by building on the (Helpman et al.,

K.N. Schmeiser, M.F. Ricaurte / International Review of Economics and Finance 22 (2012) 190–200

199

Table 4.2 Index of foreign market penetration difficulty. Industry

στ

Grain Mill and Bakery Products Other Food and Kindred Products Tobacco Products Textile Products and Apparel Lumber, Wood, Furniture, and Fixtures Paper and Allied Products Chemical Products, nec Soap, Cleaners, and Toilet Goods Other Chemicals and Allied Products Petroleum and Coal Products, nec Other Petroleum and Coal Products Rubber Products Glass Products Other Stone, Clay, and Other Nonmetallic Mineral Products Primary and Fabricated Metals Construction, Mining, and Materials Handling Machinery Household Audio and Video, and Communications Equipment Electronic Components and Accessories Other Electrical Equipment, Appliances, and Components Instruments and Related Products

0.766 0.736 0.793 0.814 0.957 0.890 1.539 1.368 1.162 15.852 8.046 0.544 0.910 1.041 1.452 0.776 1.735 5.151 1.562 0.866

Note: A larger index means higher difficulty to trade.

(2004) model of monopolistic competition where firms can choose between the foreign market servicing options of exports and FDI sales. To calibrate this model, we aggregate firms into sectors and we show that transportation costs, fixed setup costs, consumer utility weights, and productivity dispersion are not enough to match sectoral data for the U.S. and Canada in 1997. Inspired by the gravity literature employed to explain geographic patterns of trade and FDI flows, we expand the model to account for unobserved foreign market servicing costs, which we call the difficulty of foreign market penetration. We find that the expanded model does a far superior job matching several moments in the data. These market penetration costs are costs that may influence both exporters and multinationals. For example, contractual costs as in Antràs (2003) will be relevant in making the FDI or export choice for some sectors, not for others. These penetration costs allow for inclusion of multinational tax exceptions, government export incentive programs, advertising and product development costs, market knowledge, etc. Thus, we conclude that going beyond traditional transportation costs is needed to understand sectoral differences in foreign market servicing. Based on our findings, we believe that a more general concept that captures intangible sectoral costs has to be incorporated into the profit maximization problem in order to accurately understand choices made by multinational corporations. Our work suggests that future empirical research on the sectoral difficulties of foreign market penetration is needed to formalize its inclusion into individual firm-based models. Acknowledgments We would like to thank the participants of the Trade and Development Workshop at the University of Minnesota and two anonymous referees for their invaluable comments. All errors are ours. References Antràs, P. (2003). Firms, contracts, and trade structure. Quarterly Journal of Economics, 118(4), 1375–1418. Bernard, A. B., Jensen, J. B., & Schott, P. K. (2005). Importers, exporters, and multinationals: A portrait of firms in the U.S. that trade goods. NBER Working Paper, vol. 11404. Blonigen, B. A. (2001). In search of substitution between foreign production and exports. Journal of International Economics, 53(1), 81–104. Buch, C. M., Kleinert, J., Lipponer, A., & Toubal, F. (2005). Determinants and effects of foreign direct investment: Evidence from German firm-level data. Economic Policy, 20(41), 52–110. Eaton, J., Kortum, S., & Kramarz, F. (2008). An anatomy of international trade: Evidence from French firms. NBER Working Paper, vol. 14610. Feenstra, R. C., Romalis, J., & Schott, P. K. (2002). U.S. imports, exports, and tariff data, 1989–2001. NBER Working Papers, vol. 9387. Feinberg, S. E., & Keane, M. P. (2006). Accounting for the growth of MNC-based trade using a structural model of U.S. MNCs. American Economic Review, 96(5), 1515–1558. Gibson, M. J., 2006. Trade liberalization, reallocation, and productivity. Ph.D. thesis, University of Minnesota and Federal Reserve Bank of Minneapolis. Gleason, K. C., Lee, C. I., & Mathur, I. (2002). Dimensions of international expansions by US firms to China: Wealth effects, mode selection, and firm-specific factors. International Review of Economics & Finance, 11(2), 139–154. Head, K., & Ries, J. (2003). Heterogeneity and the FDI versus export decision of Japanese manufacturers. NBER Working Paper, vol. 10052. Helpman, E., Melitz, M. J., & Yeaple, S. R. (2004). Export versus FDI with heterogeneous firms. American Economic Review, 94(1), 300–316. Hummels, D. (2001). Toward a geography of trade costs. Center for Global Trade Analysis, Purdue University GTAP working paper no. 1162. Markusen, J. R. (2004). Multinational firms and the theory of international trade. Vol. 1 of MIT Press Books. The MIT Press. Melitz, M. J. (2003). The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica, 71(6), 1695–1725.

200

K.N. Schmeiser, M.F. Ricaurte / International Review of Economics and Finance 22 (2012) 190–200

Mitze, T., Björn, A., & Gerhard, U. (2010). Trade-FDI linkages in a simultaneous equations system of gravity models for German regional data. Economie Internationale, 2Q, 121–162. Neary, J. P. (Mar 2009). Trade costs and foreign direct investment. International Review of Economics and Finance, 18(2), 207–218. Nocke, V., & Yeaple, S. (2007). Cross-border mergers and acquisitions vs. greenfield foreign direct investment: The role of firm heterogeneity. Journal of International Economics, 72(2), 336–365. Schott, P., 1987. Schotts international economics resource page: Trade data and concordances. Toubal, F., Kleinert, J., & Buch, C. M. (2003). Determinants of German fdi: New evidence from micro-data. Discussion paper series 1: Economic studies 2003, 09. Deutsche Bundesbank, Research Centre. U.S. Bureau of Economic Analysis. (2006). U.S. direct investment abroad, all foreign affiliates, sales by industry of affiliate and country. U.S. Census Bureau. (2001). 1997 economic census manufacturing, general summary. Wagner, J. (2006). International firm activities and innovation: Evidence from knowledge production functions for German firms. Discussion papers on entrepreneurship, growth and public policy. Max Planck Institute of Economics, Group for Entrepreneurship, Growth and Public Policy. Yeaple, S. (2006). Offshoring, foreign direct investment, and the structure of US trade. Journal of the European Economic Association, 4(23), 602–611.