Journal of International Economics 72 (2007) 113 – 127 www.elsevier.com/locate/econbase

Foreign outsourcing, exporting, and FDI: A productivity comparison at the firm level
Eiichi Tomiura ⁎
Department of Economics, Yokohama National University, 79-4 Tokiwa-dai, Hodogaya-ku, Yokohama City, 240-8501, Japan Received 18 August 2004; received in revised form 1 May 2006; accepted 25 November 2006

Abstract This paper documents how productivity varies with globalization modes, based on a firm-level data set covering all manufacturing industries in Japan without any firm-size threshold. Only a small fraction of firms outsource, export, or invest abroad. Foreign outsourcers and exporters tend to be less productive than the firms active in FDI or in multiple globalization modes but more productive than domestic firms. This productivity ordering is robust even when firm size, factor intensity, and/or industry are controlled for. This paper also finds that outsourcers are on average less capital intensive than other globalized firms. © 2007 Elsevier B.V. All rights reserved.
Keywords: Foreign outsourcing; Exporting; FDI; Heterogeneity; Firm-level data; Productivity JEL classification: F12; F23; D20; F14

1. Introduction Cross-border business activities in various forms have been facilitated by trade liberalization and the development of information technology. The globalization of firms is by no means universally observed, however. Even within industrialized countries, the vast majority of firms sell all their output to domestic consumers, have no affiliates overseas, and outsource exclusively to domestic suppliers. This paper uses firm-level data to document the extent to which firms engage in global activities and to evaluate how productivity varies with the choice of globalization modes.
⁎ Tel.: +81 45 339 3563; fax: +81 45 339 3574. E-mail address: tomiura@ynu.ac.jp. 0022-1996/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.jinteco.2006.11.003

114

E. Tomiura / Journal of International Economics 72 (2007) 113–127

Theoretical models have recently incorporated inter-firm heterogeneity into the firm's globalization decision. When firms are heterogeneous, the optimal choice differs across firms. Antràs and Helpman (2004) formalize how a firm sources abroad either through foreign outsourcing (FO) or foreign direct investment (FDI). In their model, the high-productivity firms source overseas by engaging in FDI; the low-productivity firms acquire intermediates only within the home country; and the firms with medium productivity choose FO. Global activities, especially FDI, require larger fixed entry costs, but bring in higher gross profits for productive firms.1 On the other hand, Helpman et al. (2004) analyze the decision to serve foreign markets through exporting or FDI. Their model predicts the following pattern in the access to foreign markets; only the most productive firms find it profitable to produce offshore by FDI; the firms with medium productivity serve foreign markets by exporting; the least productive firms serve only the domestic market. These theories of organizational mode choice have deepened our understanding of trade and FDI from the aggregate sector level to the fundamental firm level. Empirical studies on the globalization decisions of heterogeneous firms remain limited, however, partly because of constraints in the availability of micro data. Though several studies have examined the exporting-FDI choice,2 we are aware of no studies weighing the relative importance between FO and FDI.3 “A firm-level data analysis is needed to answer this question, and no such analysis is available at this point in time” (Antràs and Helpman, 2004, p. 553). This paper compares the firm's productivity across globalization modes by using firm-level data on FDI, export, and foreign outsourcing (explicitly distinguished from domestic outsourcing). The sample covers 118,300 firms without any firm-size threshold, across all manufacturing industries in Japan. This firm-level data set has another advantage, as FO data are consistently derived from the same survey as the export and FDI data. Given that the same firm makes sales and acquisition decisions as a single optimizing unit, it is important to encompass both within an integrated empirical framework. As Grossman et al. (2005) and Yeaple (2003) imply, firms active in a globalization mode are likely to engage in other globalization modes to take advantage of the globalization effects in reducing costs, expanding outputs, and raising the returns from other global activities.4 This paper investigates the complementarity among various globalization modes by exploiting the comprehensive breadth of our data set. The rest of this paper is organized as follows. Section 2 describes the data. Section 3 documents the allocation of firms across eight different globalization modes. Section 4 compares the productivity of the firms. Section 5 closes with final remarks. 2. Description of data 2.1. Data source This paper extracts firm-level data from The Basic Survey of Commercial and Manufacturing Structure and Activity (Sho-Kogyo Jittai Kihon Chosa in Japanese).5 The survey covers 118,300
Grossman and Helpman (2004) predict different ordering in their model with monitoring efforts. See Bernard et al. (2005), Brainard (1997), Girma et al. (2005), Head and Ries (2003), and Helpman et al. (2004), for example. 3 Görg et al. (2004) estimate the relationship between intermediate imports and the productivity of exporters and foreign-owned firms in Ireland. They do not analyze the globalization mode decision or outward FDI, however. 4 Yeaple (2003) examines firms simultaneously engaged in vertical and horizontal FDI. Grossman et al. (2005) examine foreign and domestic outsourcing. Head and Ries (2003) formalize firms exporting to some countries and investing in others. 5 Though the original firm-level data cannot be publicly disclosed, any researcher can gain access to the same data set by obtaining official individual permission from the government in advance.
2 1

E. Tomiura / Journal of International Economics 72 (2007) 113–127

115

firms in all manufacturing industries. No firm-size threshold is imposed, irrespective of the firm's involvement in global activities. The only previous data set to come close to matching this sample size was that by Bernard, Eaton, Jensen, and Kortum (hereinafter BEJK) (2003) on about 200,000 plants derived from U.S. Census of Manufacturers. Thus, this survey is regarded as an accurate overall representation of the whole of manufacturing in Japan. As the survey was conducted only once at 1998, the data set is in a cross-section format. The survey contains various corporate data, such as sales, employment, capital, direct exports, and FDI. As unique data on FO, the survey directly asks sample firms whether they “contract out manufacturing or processing tasks to other firms overseas.” As firms (suppliers, subcontractors) overseas and those within the home country are separated in this questionnaire, foreign outsourcing thus defined is explicitly distinguished from domestic outsourcing.6 2.2. Measures of productivity Before reporting empirical results, this section discusses various measures for productivity. This paper begins with labor productivity, defined by VA (value-added) per-worker, as it certainly is one of the most frequently used measures. The comparison after subtracting costs is essential when outsourcers are involved. If we depend on gross output (sales) per-worker, vertically integrated firms (exporters or investors) appear to be less productive than outsourcers even with identical sales and labor productivity because outsourcers employ fewer workers in-house. Hence, the principal productivity measure in this paper is, in logarithms, h ¼ lnðVA=LÞ ¼ ln½ðSales − CostÞ=LŠ; ð1Þ

where L is the number of regular employees. Cost denotes “the cost of goods sold,” or cost of sales, which is the only cost data available in the survey.7 This paper will also report sales perworker (before subtracting cost) for a robustness check. Second, this paper uses total factor productivity (TFP), an ideal measure for evaluating the contribution of capital. The use of cross-section data, however, makes it practically impossible to estimate the true TFP for each firm. As an alternative, this paper adopts an approximate TFP:8 ATFP ¼ ln Q 1 K − ln : L 3 L ð2Þ

Both value-added and sales are used as Q in calculating ATFP. K denotes tangible fixed assets. This paper uses ATFP only for a robustness check purpose because ATFP is merely a practical approximation of the true TFP and cannot be calculated for the vast number of surveyed firms (21,950 firms, all with capital recorded as zero or unreported). Third, as another robustness check, this paper compares the firm size (L) and the home market share.9 These indices are available for all firms, measured with relatively decent accuracy, and
6 Though it does not cover FO of non-production services, this definition is more appropriate than intermediate import data in excluding intermediates purchased in the marketplace. Tomiura (2005) explains the survey's definition in detail and compares it with other FO measures. 7 This accounting cost concept covers not only outsourcing payment, but also production wages, a component of valueadded rather than economic costs. As globalized firms tend to pay higher wages (e.g., Bernard and Jensen, 1999), however, the productivity premium of globalized firms based on this definition should be a conservative estimate. 8 This productivity proxy adjusts labor productivity by capital intensity, with the importance of capital as weight. See Head and Ries (2003), for example, for ATFP. 9 Home market share is defined as the share of the firm's sales in the domestic market for each two-digit industry.

116 Table 1 Percentage of firms O 1.71 X 4.35

E. Tomiura / Journal of International Economics 72 (2007) 113–127

I 1.08

OX 0.30

XI 1.23

IO 0.25

OXI 0.42

Domestic 90.65

Note: The percentages shown above are of the total number of firms. Categories are mutually exclusive.

informative since they increase monotonically with underlying productivity in standard theoretical models.10 3. Share of firms 3.1. Overall comparisons This section classifies firms by their choice of globalization modes. Table 1 presents the percentage share of each type of firms among the total number surveyed. All firms with cost data are included (98% of surveyed firms). O denotes firms outsourcing to foreign suppliers, and X denotes exporters. FDI firms, I, are identified based on ownership of foreign affiliates, but only if they own shares of at least 20% (to represent FDI in manufacturing affiliates as opposed to FDI in sales branches or portfolio investment). The eight categories in the table are disjoint, i.e., mutually exclusive. IO, for example, represents firms that outsource and invest abroad but never export. We note the following points from this table. First of all, about 90% of the firms are “domestic” (involved in none of these three activities). This level of share is overwhelming, though some of these firms may be linked with global economies through other channels, such as raw material imports or international portfolio investment.11 Moreover, far fewer than 1% of the firms engage in all three globalization modes simultaneously. This finding may suggest that fixed entry costs for globalization are non-negligible. Second, the exporters account for a relatively high share in globalized firms. This may reflect the difference in entry costs across various globalization channels. We must note, however, that if the survey had captured FO of non-production services, the share of FO firms would have far exceeded that of FDI firms. Third, 64% of investors are active in other globalization modes (exporting and/or outsourcing). The most frequent globalization mode among FDI firms is XI rather than I. On the other hand, exporting is the only globalization channel for 69% of exporters, while foreign outsourcing is the only globalization channel for 64% of foreign outsourcers. These figures indicate that firms may find it difficult to own affiliates overseas unless they have accumulated experience in exporting or foreign outsourcing. This paper differs from previous work in identifying outsourcers and in distinguishing overlaps among the three globalization modes. However, the results on exporters and investors are generally consistent with the existing evidence.12 For example, Bernard et al. (2005) report
10 Yeaple (2005) uses home market share and provides a convincing case for it in this context. Head and Ries (2003) also use firm size in the domestic market along with ATFP. 11 This low share of globalized firms is a conservative estimate. Firms with no less than 50 employees are surveyed with certainty, while those with less than 50 (those more likely to be domestic firms) are sampled with a probability of less than one. 12 BEJK (2003) find a higher share of exporters in the U.S. from plant data, yet the exporters are more likely to be multiplant firms than domestic firms. Head and Ries (2003) report that 43% of Japanese firms are XI, but their sample is limited to 1,070 publicly listed firms.

E. Tomiura / Journal of International Economics 72 (2007) 113–127 Table 2 Percentage within each industry Industry 12. Food manufacturing 13. Beverage, tobacco, and feed 14. Textile 15. Apparel and textile products 16. Timber and wooden products 17. Furniture and fixture 18. Paper and pulp products 19. Printing and publishing 20. Chemical products 21. Petroleum and coal products 22. Plastic products 23. Rubber products 24. Leather and fur products 25. Ceramic, stone, and clay 26. Iron and steel 27. Nonferrous metals 28. Metal products 29. General machinery 30. Electric machinery 31. Transport equipment 32. Precision instruments 34. Miscellaneous O 0.39 0.36 1.38 2.37 0.96 1.63 1.78 2.14 0.91 1.12 1.39 2.06 3.48 0.91 1.86 1.43 1.75 2.45 2.31 1.84 2.20 2.94 X 2.04 1.89 4.04 2.82 1.26 1.17 3.21 4.84 10.27 5.34 3.24 3.33 1.74 3.09 5.13 4.86 4.04 7.26 6.36 4.98 7.40 4.42 I 1.05 0.79 0.95 1.35 0.68 0.78 0.76 0.73 2.01 2.25 1.83 0.76 0.62 0.64 1.57 1.59 0.91 0.92 1.23 1.64 0.54 1.16 OX 0.02 0.03 0.05 0.20 0.07 0.04 0.16 0.13 0.39 0.00 0.08 0.17 0.17 0.03 0.25 0.11 0.25 0.76 0.71 0.39 0.95 0.57 XI 0.27 0.27 0.49 0.22 0.05 0.21 0.57 0.23 5.06 3.37 1.57 1.18 0.17 0.70 1.16 2.54 0.83 2.27 2.29 2.58 1.48 1.32 IO 0.07 0.09 0.19 0.67 0.19 0.28 0.16 0.21 0.11 0.00 0.30 0.42 0.17 0.10 0.17 0.21 0.10 0.21 0.39 0.42 0.18 0.59 OXI 0.06 0.03 0.12 0.07 0.02 0.04 0.00 0.13 0.50 0.28 0.38 0.63 0.06 0.15 0.29 0.63 0.19 0.91 1.18 0.96 0.84 0.59 Domestic 96.09 96.53 92.78 92.30 96.77 95.86 93.36 91.59 80.76 87.64 91.20 91.45 93.61 94.38 89.57 88.64 91.93 85.20 85.53 87.19 86.41 88.40

117

No. of firms 12,228 3,288 5,886 8,594 4,276 2,826 3,147 5,330 3,633 356 4,965 2,374 1,783 5,958 2,416 1,892 10,780 11,166 12,185 5,964 3,906 3,872

Note: The percentage of firms within each industry is shown, except in the last column. The ordnance industry (33) is merged into general machinery (29).

that 4.2% of firms surveyed in the U.S. at 2000 were exporters without FDI. This share is in a comparable range with 4.65% (X + OX) in Table 1.13 3.2. Inter-industry comparisons In contrast with the previous table, which combines all firms, Table 2 disaggregates firms by industry.14 Several notable differences in the percentage share of firms among industries emerge from this table. First, the extent of globalization varies considerably across industries. The chemical industry (20) has the lowest percentage of domestic firms, at just over 80%, followed by the four machinery industries (general machinery, electric machinery, precision instruments, and transport equipment). Among these machinery industries, electric machinery and transport equipment industries have the highest share of the most globalized category of firms, OXI. Around 8% of the globalized firms in these two industries are active in all three modes. At the other end of the spectrum, the industries with the five highest percentages of domestic firms (94–97%) are timber and wooden products (16), beverage, tobacco and feed (13), food manufacturing (12), furniture and fixture (17), and ceramic, stone, and clay (25). All of these industries are generally presumed to have high trade costs.
13 14

They define “multinational exporters” as firms with non-zero exports to related parties. Disaggregation on three-digit industries is available upon request.

118 Table 3 Inter-industry correlation O O X I OX XI IO OXI Domestic 1 0.091 −0.254 0.585 −0.185 0.561 0.353 −0.168

E. Tomiura / Journal of International Economics 72 (2007) 113–127

X 1 0.582 0.415 0.772 − 0.103 0.307 − 0.945

I

OX

XI

IO

OXI

Domestic

1 −0.176 0.801 −0.117 0.006 −0.697

1 0.118 0.262 0.449 − 0.420

1 − 0.158 0.104 − 0.011

1 0.197 −0.879

1 − 0.407

1

Note: The correlation is for the share in the number of firms across 75 three-digit industries.

Inter-industry differences are also apparent if we compare different globalization modes. Apparel (15) ranks high in O and IO but low in X and XI, while chemical products (20) rank high in X and XI but low in O and IO. While food manufacturing (12) and beverage, tobacco, and feed (13) are among the least globalized industries, with particularly low percentages of O, about a quarter of globalized firms within these industries choose I. The printing and publishing industry (19), on the other hand, depends mostly on X and engages in very little I. These industry characteristics are generally consistent with our daily observations.15 Finally, there appear to be complementarities between different globalization modes, as indicated by Yeaple (2003) and Grossman et al. (2005). An industry with a high share of exporters, for example, tends to have a high share of investors (e.g., chemical products, electric machinery). Table 3 summarizes the correlation matrix across three-digit industries. The interindustry correlation between I and XI is as high as 0.80.16 The clearly positive, though weaker, correlation is also observed for outsourcing (O and OX, IO). 4. Productivity comparisons 4.1. Overall comparisons Table 4 compares alternative measures of productivity. The average productivity is expressed ¯ ¯ as a percentage logarithm difference from the productivity of domestic firms: 100(θ XI − θ Dom) for ¯ XI firms, for example, where θ indicates the mean productivity. A comparison of various measures suggests the following. First, the gap between global and domestic firms is substantially wider in firm size than in productivity. Globalized firms are larger than domestic firms by at least 70% in O and by more than 300% in XI and OXI.17 Similar regularity is also exhibited in the comparison of home market shares, which are adjusted for size differences directly due to global activities.

15 Antràs and Helpman (2004), for example, contrast Intel, a firm which assembles most of its microchips in-house, with Nike, a firm which outsources most of its manufacturing. Head and Ries (2003) refer to food manufactures as examples of I firms. 16 The correlation rises to 0.897 when calculated between all exporters (X + OX + XI + OXI) and all investors (I + XI + IO + OXI). 17 From a different Japanese firm-level data set, Head and Ries (2003) report that domestic firms are 19–33% smaller and XI firms are 105–332% larger than X firms in size.

E. Tomiura / Journal of International Economics 72 (2007) 113–127 Table 4 Alternative measures of productivity O (A) Q/L (value-added) (B) Q/L (sales) (C) Firm size (L) (D) Home market share (E) TFP (value-added) (F) TFP (sales) (G) K/L 23.27 50.58 70.54 87.03 10.34 28.02 21.00 X 10.04 45.22 145.34 119.70 − 8.20 18.13 41.41 I 39.92 86.69 224.82 318.66 5.83 40.69 86.26 OX 41.94 81.35 173.31 210.89 22.93 50.79 40.95 XI 67.68 104.56 322.27 348.72 27.37 52.39 105.03 IO 34.06 80.96 186.01 283.20 10.70 45.85 53.48

119

OXI 62.70 109.08 341.45 390.41 25.12 59.82 96.48

Note: The percentage logarithm difference from domestic firms is shown. Firms without capital data are excluded from rows E to G.

Second, the productivity gap is considerably larger in sales per-worker but remains generally substantial even after cost is subtracted: the difference in value-added labor productivity from domestic firms is at least 10% for X versus more than 60% for XI and OXI, except when X is measured by value-added ATFP.18 Given that the choice of productivity measure has no significant effect on the productivity ordering of globalization modes, the following sections of this paper concentrate on the value-added labor productivity. The same table also shows the capital–labor ratio since the labor productivity is inevitably affected by K/L. On average, domestic firms are the least capital intensive and investors are the most capital intensive. Outsourcers are more labor intensive than other globalized firms. This finding of labor intensiveness of outsourcers is consistent with the theoretical prediction by Antràs (2003).19 Several findings should be noted on the comparisons across various globalization modes. First, the average productivity of domestic firms is distinctively lower than that of any globalized firms. The magnitude of value-added labor productivity advantage of exporters (from 10% for X to 68% for XI) is also roughly in line with the existing evidence (e.g., BEJK, 2003, find a gap of 33% in the U.S. by combining X together with XI, OX, and OXI). Second, investors are on average more productive than other globalized firms, while exporters appear the least productive among globalized firms. The finding of higher productivity of FDI firms than exporters is in line with previous results (e.g., Girma et al., 2005).20 The finding of higher productivity of FDI firms compared with outsourcers provides rare and direct evidence in support of the theoretical prediction by Antràs and Helpman (2004). Third, the firms active in multiple globalization modes, especially OXI and XI, tend to be particularly productive. The comparison of XI with X or I is consistent with the result by Head and Ries (2003). As far as the author knows, no previous studies have identified OXI firms. 4.2. Productivity comparisons conditional on firm size and factor intensity This section controls for the firm size or capital intensity in comparing productivity, as firm size and capital intensity are both supposed to be critical in determining the globalization of
Note that the surveyed cost, which is subtracted from productivity, includes production wages. Investment cost sharing is relatively easier in physical capital, while hiring/managing workers requires local knowledge/management. Motivated by these observations, Antràs (2003) shows that outsourcing contracts (compared with vertical integration) lose their attractiveness as the capital–labor ratio rises. 20 Head and Ries (2003) find that the size measure strongly supports the theory, while ATFP result in Japan is mixed. They depend, however, on a limited sample of publicly listed firms.
19 18

120 Table 5 Conditional comparisons

E. Tomiura / Journal of International Economics 72 (2007) 113–127

O (A) Productivity gap within size bins (B) Productivity gap within K/L bins (C) Size gap within productivity bins 25.70 16.06 38.52

X 14.51 − 1.80 102.73

I 35.75 13.20 172.78

OX 42.65 30.44 132.54

XI 47.79 35.42 271.03

IO 33.23 18.06 134.09

OXI 34.83 33.89 292.07

Note: The mean of each bin is subtracted. The percentage log deviation from domestic firms is shown.

firms.21 This paper investigates whether the high average productivity we observe in globalized firms stems solely from their capital intensity or large size. This paper sorts all of the firms by firm size and allocates them into approximately 300 bins, as defined by BEJK (2003).22 This paper also controls for K/L by a similar procedure. Table 5 displays the percentage logarithm deviations from domestic firms, after subtracting the average of each bin. See Appendix A for the definitions of bins. As the most notable result from Table 5, this paper finds that the firm's globalization cannot be reliably predicted solely by the size or factor intensity of the firm. The control for the firm size still leaves non-negligible (15–48%) productivity gaps (demonstrated in row A).23 The productivity premium remains considerable (13–35%) for all modes except X,24 even after controlling for the firm's capital–labor ratio (see row B).25 The firm-size gap contracts somewhat but remains very large (39–292%) for all eight globalization channels, even after controlling for the differences in productivity (see row (C)). While Table 5 sums over all firms, more disaggregated results are provided in Table 6. The regularities in this table shed some doubt on the standard simplifying assumption that entry costs for globalization are the same across all firms. The productivity advantage required for many globalization channels tends to be more substantial among smaller sized firms. This result is plausible, as high labor productivity is only one of several attributes of large firms which facilitate globalization. Other attributes, such as strong headquarter functions, rich retained earnings, and established distribution networks and consumer recognition, for example, make it easier for large firms to expand globally. Similarly, though less obviously, the productivity differential appears wider among less capital-intensive firms, and the firm size necessary for globalization with the exception of outsourcing appears larger among firms with higher productivity. 4.3. Intra- and inter-industry productivity comparisons The previous section disaggregated firms by size or capital intensity, but not by industry, an equally important attribute in predicting the firm's globalization choice. Accordingly, this section investigates the productivity premium of globalized firms within and across industries.
This paper does not analyze the relation with skills since the survey contains no data on skills or occupations. BEJK (2003) assign plants into 500 bins, but the size of their bins is approximately the same as ours because their sample size is more than twice as large. 23 The control for the firm size is effective only for I, XI, and OXI. This may be because FDI firms are especially large, as reported in row C of Table 4. 24 BEJK (2003) report the U.S. exporter's premium as 9–20% in labor productivity, controlling for K/L. Bernard and Jensen (1999) report it as 4–18% in TFP. Neither of them distinguishes OX, XI, and OXI firms among exporters, however. 25 The control for K/L significantly decreases the productivity gap. The numerator in our productivity, however, is approximately equal to the economic profit plus capital depreciation, which is inevitably affected by capital intensity.
22 21

E. Tomiura / Journal of International Economics 72 (2007) 113–127 Table 6 Gap within bins O Size bins 1≤L≤5 6 ≤ L ≤ 19 20 ≤ L ≤ 99 100 ≤ L K/L bins K = 0 or no K data K/L ≤ 1 1 ≤ K/L ≤ 2.75 2.75 ≤ K/L ≤ 7.45 7.45 b K/L Productivity bins θ≤0 0 b θ ≤ 1.2040 1.2040 ≤ 1.8052 1.8052 b θ 49.55 32.81 21.94 14.97 X 19.31 24.34 20.67 11.96 I 57.50 72.02 36.48 38.40 OX 68.56 70.36 45.81 35.79 XI 101.41 71.02 65.00 52.21 IO 156.90 64.75 43.00 16.99

121

OXI 74.56 55.65 49.22 39.36

55.97 39.82 14.24 10.93 7.16

15.85 − 9.28 − 10.35 − 4.29 4.71

n. a. 25.97 11.63 6.36 18.54

65.37 53.21 20.80 35.23 25.95

71.03 93.97 35.92 29.38 38.69

112.38 83.79 18.19 12.08 15.55

103.83 99.66 26.63 28.68 37.23

64.83 43.33 42.23 31.56

109.65 86.98 96.01 123.50

137.15 146.03 143.23 217.65

137.27 144.84 102.66 153.72

150.78 209.23 230.12 317.18

158.19 118.06 136.73 144.66

185.07 206.75 249.14 353.41

Note: The mean of each bin is subtracted, as in Table 5. The upper two panels display productivity, while the bottom panel displays firm size. All figures are in % log relative to domestic firms.

Table 7 reports intra-industry variations. Firms are disaggregated into 75 three-digit industries, each of which is further divided into 10 bins according to the firm size or capital intensity.26 The figures in the table are the results after subtracting the average of each industry or of each intra-industry bin. Table 7 demonstrates that the industry characteristics, firm size, and factor intensity are insufficient to predict the globalization choice of each firm. Row A shows that the productivity premium expands rather than contracts in most globalization modes when firms in the same industry are compared. Row B confirms that the firm size differential within each industry helps little in predicting the firm's globalization pattern. As we see in row C, a substantial productivity premium (16–40%) remains even with the tightest controls both for industry and capital intensity, except in the case of mode X. Table 8 compares the intra-industry productivity premium of globalized firms relative to domestic firms across all industries.27 The figures in this table are the industry disaggregation of those in row A of Table 7. Again, globalized firms are more productive than domestic firms in many industries. The average productivity of I and XI firms, for example, is higher than that of domestic firms in every industry.28 Similarly, the productivity premiums of O, IO, and OXI firms are positive in all industries except one.29
26 The three-digit classification is the most detailed level in the survey. The effect of industry control is stronger in BEJK (2003), with 458 four-digit industries. 27 Result with 75 three-digit industries is available upon request. 28 In textile and apparel industries, low-productivity firms are involved in exporting. The small overall productivity premium for X must be strongly influenced by these two industries. Given that they are typical declining industries, many traditional exporters may continue exporting even after they no longer sustain a productivity advantage. 29 The productivity advantage for I and OXI is very high in food and beverage industries, which are characterized by inactive globalization (see Table 2). This finding suggests that industry-specific factors such as trade costs raise entry costs for globalization.

122

E. Tomiura / Journal of International Economics 72 (2007) 113–127

Table 7 Productivity gaps conditional on industries O (A) Within industries (B) Within industries (size bins) (C) Within industries (K/L bins) 26.94 28.48 16.69 X 8.46 9.83 − 3.25 I 41.31 41.36 15.83 OX 46.21 47.73 30.68 XI 65.47 59.26 35.99 IO 45.19 51.78 23.30 OXI 70.85 70.25 40.14

Note: The mean of the respective industry and bin is subtracted. There are 75 three-digit industries and 10 bins within each industry.

These results, controlling for industries, confirm that intra-industry, inter-firm heterogeneity is important in describing a firm's globalization pattern. 4.4. Distribution of productivity Though the previous sections have compared the average productivity, this section explores inter-firm distributions within each globalization mode. Table 9 summarizes the basic statistics for the logarithm labor productivity. Three points are to be noted. First, relatively large cross-section variations compel us to avoid depending exclusively on average values and to investigate the distribution within each globalization category. Second, the difference in standard deviation indicates that most of the XI and OXI firms are all very productive, whereas the firms globalized only through exporting distribute over

Table 8 Productivity gap within each industry Industry 12. Food manufacturing 13. Beverage, tobacco, and feed 14. Textile 15. Apparel and textile products 16. Timber and wooden products 17. Furniture and fixture 18. Paper and pulp products 19. Printing and publishing 20. Chemical products 21. Petroleum and coal products 22. Plastic products 23. Rubber products 24. Leather and fur products 25. Ceramic, stone, and clay 26. Iron and steel 27. Nonferrous metals 28. Metal products 29. General machinery 30. Electric machinery 31. Transport equipment 32. Precision instruments 34. Miscellaneous manufacturing O 31.79 57.41 29.29 80.92 23.02 7.10 36.00 8.01 5.28 18.37 0.39 21.85 59.92 9.24 −13.10 20.24 15.21 3.46 28.60 17.88 36.11 45.56 X 0.39 49.36 − 26.55 − 51.81 − 7.93 − 4.78 22.09 − 6.82 50.28 33.73 − 1.11 22.97 0.20 20.31 − 2.69 15.18 − 4.38 17.60 10.90 − 6.59 36.90 36.60 I 68.95 110.66 24.84 51.96 54.65 17.80 55.69 58.34 63.54 76.92 15.97 43.47 78.90 58.77 17.26 20.13 36.25 10.52 40.93 22.92 23.34 41.72 OX 18.80 −7.16 77.80 93.94 47.70 82.07 47.07 35.20 73.65 n. a. −33.99 47.86 100.47 60.37 −27.61 27.80 25.98 39.31 45.55 34.38 51.86 91.42 XI 122.43 161.58 48.70 86.23 9.94 5.24 58.86 73.62 95.06 86.83 45.61 71.58 77.26 47.75 26.80 46.96 32.83 59.04 82.94 45.04 83.32 60.45 IO 108.43 175.70 46.54 69.66 27.38 1.27 90.89 24.15 100.07 n. a. 48.58 3.04 70.21 70.42 6.42 − 15.87 29.31 7.63 43.28 7.77 59.79 57.13 OXI 136.11 202.66 81.67 95.50 − 34.02 169.58 n. a. 30.78 119.32 89.14 29.18 75.89 84.98 85.90 47.81 45.76 36.35 51.16 84.45 50.32 74.39 109.74

Note: The percentage log difference from domestic firms within each industry is shown.

E. Tomiura / Journal of International Economics 72 (2007) 113–127 Table 9 Summary statistics O Average Median SD 1.45 1.48 0.86 X 1.32 1.45 1.05 I 1.62 1.67 0.93 OX 1.64 1.72 0.86 XI 1.89 1.93 0.75 IO 1.56 1.62 0.96 OXI 1.84 1.87 0.79

123

Domestic 1.22 1.29 0.91

a wide range of productivity. Finally, the median is larger than the mean in all globalization modes, suggesting that firms are distributed with a long tail over the low-productivity range.

Fig. 1. Productivity distribution (OXI vs. Domestic) Note: The intervals of logarithm productivity are on the horizontal axis. The percentage of firms is on the vertical axis. The overall mean is subtracted from each firm's productivity in the top graph, while the three-digit industry mean is subtracted in the bottom graph. The same 18 productivity intervals are applied to all cases.

124

E. Tomiura / Journal of International Economics 72 (2007) 113–127

Fig. 2. Productivity distribution (O, X, I vs. Domestic).

Figs. 1–3 plot the distributions of productivity (the percentage share of firms within each productivity interval), each of which is an empirical counterpart of the probability density function (pdf). See Appendix A for details on the construction of the intervals.30 Two graphs are shown for every case: the upper one of which is for the productivity relative to the overall mean, while the lower one of which normalizes each firm's productivity by the mean productivity in its three-digit industry. Fig. 1 compares domestic firms with the most globalized group of firms (OXI). OXI firms are distributed heavily in relatively high-productivity ranges. Thus, this paper finds again that OXI firms are more productive than domestic firms. The comparison of the two graphs in Fig. 1
The number of intervals chosen is 18, exactly the same as in BEJK (2003), to facilitate comparisons. The same 18 intervals are applied to all cases.
30

E. Tomiura / Journal of International Economics 72 (2007) 113–127

125

Fig. 3. Productivity distribution (OX, XI, IO vs. Domestic).

demonstrates that controlling for industry only negligibly affects the productivity distributions, corroborating the result from BEJK (2003) on exporters. Next, Fig. 2 provides distributional information on firms globalized through only one mode. Domestic firms are shown again to provide a benchmark. I firms are concentrated heavily over high-productivity ranges, whereas X firms are distributed widely over very low-productivity ranges. O firms are located somewhere between. Judging from the bottom graph, the control of industry appears to slightly tighten the distributions. Fig. 3 plots firms active in two globalization modes. The productivity distribution of XI firms shifts considerably to the right of that of other firms. Again, the productivity advantage of globalized firms survives even when looking at disaggregated industries.

126

E. Tomiura / Journal of International Economics 72 (2007) 113–127

These productivity distributions described here confirm our previous findings. Higher productivity of investors compared with exporters and domestic firms is in line with the existing evidence from inter-firm distributions.31 This paper also confirms the robustness of our finding that FDI firms are more productive than FO firms even if the comparison is based on cross-firm distributions. 5. Concluding remarks This paper has demonstrated that FDI firms are distinctively more productive than foreign outsourcers and exporters, which in turn are more productive than domestic firms. This ordering is consistent with theory and has been confirmed to be robust even after controlling for industry, firm size, and factor intensity. Thus, this paper has provided direct evidence in support of the empirical relevance of the heterogeneous firm model in international trade. These findings are informative as a detailed description of firm heterogeneity. The causality between a firm's productivity and its choice of globalization mode, however, will need to be analyzed if similar firm-level data become available in the longitudinal form. Acknowledgment The Ministry of General Affairs of Japan granted me access to the government micro-data files by issuing official approvals. Kei Nara and Mutsuharu Takahashi assisted me in accessing the data. Seiichi Katayama encouraged me to initiate this research. Earlier versions of this paper were presented at the Econometric Society World Congress and international conferences held by Kobe University and Yokohama National University. Helpful comments from participants, especially Craig Parsons and Stephen Yeaple, are acknowledged. The author is deeply grateful to the editor (Jonathan Eaton) and two anonymous referees for their valuable comments. This research was partly supported by the Japan Economic Research Promotion Foundation and Grant-in-Aid for Scientific Research C16530152. Any remaining errors are my own. Appendix A. Construction of bins/intervals [The construction of the bins] All firms with cost data are sorted by L and allocated into approximately 300 size bins. Each bin cannot include exactly the same number of firms, as the firms with exactly the same number of employees are assigned into the same bin. The productivity bins, grouped according to productivity levels, follow the same definition. In sorting by K/L, all the firms lacking data on capital or with capital recorded as zero are grouped into the same bin. The industries with severely limited numbers of firms (Industry Nos. 203, 211, 219, and 231) P  divided à are not into 10 bins. The thresholds for the bins are available upon request. The ¯ value N1X i hi − h ðbðiÞÞ is displayed in the tables for X firms, for example. Firms are indexed ¯ by i. NX is the number of X firms, while b(i) denotes the firm-size bin to which firm i belong. θ is the mean over the firms belonging to this bin b. [The definition of productivity intervals for pdf ] Firms are first sorted by productivity level (after subtracting the mean) within each globalization mode. Next, the domestic firms in the top and bottom 1% in the overall mean case are assigned to the top and bottom intervals, respectively.
Girma et al. (2005) find, in the case of UK, that the productivity distribution of multinational plants stochastically dominates that of exporters, which in turn dominates that of non-exporters.
31

E. Tomiura / Journal of International Economics 72 (2007) 113–127

127

The productivity range between these two thresholds is equally divided into 16 intervals. The same 18 intervals thus constructed are applied to all cases. References
Antràs, P., 2003. Firms, contracts, and trade structure. Quarterly Journal of Economics 118, 1375–1418. Antràs, P., Helpman, E., 2004. Global sourcing. Journal of Political Economy 112, 552–580. Bernard, A., Jensen, B., 1999. Exceptional exporter performance: cause, effect, or both? Journal of International Economics 47, 1–25. Bernard, A., Eaton, J., Jensen, B., Kortum, S., 2003. Plants and productivity in international trade. American Economic Review 93, 1268–1290. Bernard, A., Jensen, B., Schott, P., 2005. Importers, exporters, and multinationals: a portrait of firms in the U.S. that trade goods. NBER Working Paper 11404. Brainard, L., 1997. An empirical assessment of the proximity-concentration trade-off between multinational sales and trade. American Economic Review 87, 520–544. Girma, S., Kneller, R., Pisu, M., 2005. Export versus FDI: an empirical test. Review of World Economics 141, 193–218. Görg, H., Hanley, A., Strobl, E., 2004. Outsourcing, foreign ownership, exporting and productivity: an empirical investigation with plant level data. Research Paper 2004/08. University of Nottingham, Leverhume Center. Grossman, G., Helpman, E., 2004. Managerial incentives and the international organization of production. Journal of International Economics 63, 237–262. Grossman, G., Helpmam, E., Szeidl, A., 2005. Complementarities between outsourcing and foreign sourcing. American Economic Review 95 Papers and Proceedings, pp. 19–24. Head, K., Ries, J., 2003. Heterogeneity and the FDI versus export decision of Japanese manufacturers. Journal of the Japanese and International Economies 17, 448–467. Helpman, E., Melitz, M., Yeaple, S., 2004. Export versus FDI with heterogeneous firms. American Economic Review 94, 300–316. Tomiura, E., 2005. Foreign outsourcing and firm-level characteristics: evidence from Japanese manufacturers. Journal of the Japanese and International Economies 19, 255–271. Yeaple, S., 2003. The complex integration strategies of multinational and cross country dependencies in the structure of foreign direct investment. Journal of International Economics 60, 293–314. Yeaple, S., 2005. Firm heterogeneity and the structure of U.S. multinational enterprise: an empirical analysis. http://www. ssc.upenn.edu/~snyeapl2/papers/composition.pdf.