Globalization and Domestic Operations: Applying the JC/JD Method to Japanese Manufacturing Firms^*

Does the globalization of firms’ activities reduce or enhance domestic employment and operations? In the era of international production networks (Ando and Kimura 2005) or the 2nd unbundling (Baldwin 2011), firms must globalize their activities to maintain or strengthen their international competitiveness. A critical question in academic discourses as well as for policy formulation is whether such moves are beneficial for the home country as a whole. People tend to believe that outward foreign direct investment (FDI) immediately results in a reduction in domestic employment and operations. Such anti-globalization sentiments are sometimes too emotional and possibly mislead the appropriate direction for globalizing firms and policy formulation. Recent empirical literature that uses micro/panel data at the establishment or firm level has mostly claimed that FDI does not necessarily cause job destruction at home and sometimes has positive effects on domestic employment. Nevertheless, these studies do not yet seem to be convincing enough for the general public.

One of the recent strands in the literature focuses on causality from FDI to domestic employment.¹ To rigorously verify causality in econometrics, existing studies often throw away a large number of observations. To isolate the effect of FDI on domestic employment, the investigation tends to focus only on firms with the first FDI in a certain period, sometimes by destination to distinguish vertical FDI from horizontal FDI, and therefore exclude a large number of firms that already have foreign affiliates. Applying a matching technique refines the comparison, but the sample set is further reduced. Although such steps are necessary to detect causality, the overall picture must be sacrificed as firms with the first FDI is just a small subset of the total sample of Japanese manufacturing firms. There are many firms that already have FDI, and such firms will increase, maintain, or decrease the number of foreign affiliates. Many MNEs have foreign affiliates both in East Asia and developed countries, and new establishments of foreign affiliates occur mostly in East Asia. Many firms without foreign affiliates also have various forms of globalizing activities such as exports, imports, and transactions with multinational enterprises (MNEs). These facts suggest that sample sets used by logically rigorous empirical studies may not accurately represent the whole sample.

Another strand in the literature studies MNEs only and estimates a labor demand function to quantify the effects of foreign operations on domestic employment.² This is also a meaningful direction of research to trace changes in the internal structure of MNEs. A comparison with non-MNEs is not explicitly incorporated in this line of empirical studies, however.

To provide useful insights for constructive policy discussion, we believe that a comprehensive study of the complete data set even without rigorous econometric analysis is still necessary. This paper thus does not focus on the causality issue for new MNEs or estimate a labor demand function of a specific group of firms, but instead tries to provide a bird's eye view of the issue of globalizing corporate activities and domestic operations in Japan. We try to keep the samples whole as far as possible based on our database. We present a holistic view of the current status of de-industrialization in the Japanese manufacturing sector, bearing in mind the quality of data. We examine not only domestic employment but also other aspects of domestic corporate activities such as the number of domestic establishments, the number of domestic affiliates, exports, and imports. In addition, we investigate several sample periods, rather than a single specific period, to capture the evolving features of globalizing corporate activities and domestic operations.

This paper applies the job creation (JC)/destruction (JD) method for the Japanese manufacturing firms. The JC/JD method has several advantages for our purposes. First, the method can explicitly take into account the highly heterogeneous characteristics of individual firms and, at the same time, effectively bridge a gap between micro and macro aspects. Based on a series of empirical studies with the U.S. establishment-level data, a seminal work by Davis, Haltiwanger, and Schuh (1996, 17) presents four key facts about JC and JD: magnitude, persistence, concentration, and cyclicality. Magnitude means that gross JC and JD are remarkably large, which are much larger than net changes (net JC/JD) in employment, and present dynamism and firm heterogeneity. In the Japanese manufacturing sector, we find large JC/JD rates, which seem to be much smaller than the case of the United States, and de-industrialization is not relevant except for the period of 1998–2002 though a slight declining trend in manufacturing activities is observed for the recent period. Persistence means that plant-level employment changes are highly persistent and do not easily turn around. Although this is not what our study directly checks, we should not assume simple time sequencing between the expansion of foreign operations and the adjustments in domestic employment and operations because such adjustments take time. Concentration means that large JC/JD concentrate in a subset of plants. This is also what we clearly observe when we draw a density function of JC/JD for subsets of firms, which shows a high peak and narrow tails. Cyclicality means that JD rates exhibit greater cyclical variation than JC rates. This is not clear in our data set, but at least we can see that some manufacturing subsectors present high JD rates in recessions. Bearing in mind these facts, the JC/JD method is appropriate to provide a bird's eye view of the whole manufacturing sector through describing the highly heterogeneous nature of corporate evolution with micro and macro connections.

Second, the JC/JD method is also powerful in comparing different subsets of establishments or firms. The existing literature using the JC/JD method is applied to various subsets of establishments or firms in terms of sectors/subsectors, regions, establishment/firm size, and others. In this context, one of the important findings in the literature is that small firms present more dynamism with larger JC and JD than large firms.³ We will conduct a comparative study in the following three dimensions: manufacturing subsectors, SMEs versus large firms, and expanding MNEs (increasing the number of foreign affiliates) versus non-expanding MNEs versus local firms (without foreign affiliates).⁴ To investigate differences in the distribution of firms, we also apply the Kolmogorov-Smirnov (K-S) test besides the JC/JD method. We find that multinational SMEs expanding foreign operations tend to enlarge domestic employment, domestic operations, and trade, compared with other types of SMEs. Moreover, we find that expanding MNEs intensifies headquarter activities.

The paper is organized as follows: The next section introduces our data set for the Japanese manufacturing firms, and basic statistics is presented in Section 3. Section 4 applies the JC/JD method to investigate gross and net changes in domestic operations and trade. Empirical observations based on the full decomposition with entry and exit of firms is first presented, and then a detailed analysis based on the panel decomposition without entry and exit is conducted for different subsets of firms in terms of the firm size and the status of holding foreign affiliates. Section 5 conducts the K-S test for changes in domestic operations and trade by different subsets of firms. The conclusion is presented in Section 6.

Our empirical analysis is based on firm-level statistics, which are compiled by the Ministry of Economy, Trade, and Industry, Government of Japan (the former name was the Ministry of International Trade and Industry): The Basic Survey of Business Structure and Activity (Kikatsu hereafter). This database provides detailed information on (parent) firms located in Japan as well as the number, industry, and regional location of their foreign affiliates with no less than 20 percent Japanese ownership. Note that the location of foreign affiliates is not identified on a country basis; instead, the questionnaires have East Asia, North America, and Europe as regional categories.⁵ Moreover, although the information on trade is available, the destination/origin of exports/imports is not identified on a country basis; only trade data for some major regions are available besides trade as a whole.

The sample in the survey covers firms with more than 50 workers, capital of more than 30 million yen, and having establishments in mining, manufacturing, wholesale/retail trade, and restaurants. Our study uses this survey data from 1997 to 2011, which is the latest available year for us.⁶

Because the Kikatsu incorporates firm-level statistics, rather than plant-level, some useful information on the internal structure of a firm for our study is available. For instance, it provides information on the allocation of workers in headquarters (HQ) services and manufacturing activities, the number of domestic/foreign establishments, and the number of domestic/foreign affiliates. By making use of the strength of the Kikatsu data, we investigate not only domestic employment but also employment engaged in HQ services, employment involved in manufacturing activities, and other aspects of domestic corporative operations such as the number of domestic establishments, the number of domestic affiliates, exports, and imports.

Another advantage of the Kitatsu data is that the coverage of the manufacturing sector is claimed to be at the “census” level (see manufacturing subsectors in the next section). On the other hand, the coverage of the services sector is incomplete, though it has been expanded over time. Therefore, this study concentrates on manufacturing firms in investigating globalizing corporate activities and domestic operations by Japanese firms. The overall trend of the coverage for Japanese manufacturing firms by Kikatsu data is presented in Table A.1 and it is briefly discussed in Appendix  A.

While the Kitatsu data have several advantages, they also have limitations related to the “census” coverage, which is particularly serious for our study. One issue is the size truncation. As mentioned earlier, the survey covers firms with 50 or more workers. Thus, firms with fewer than 50 workers are not included in the survey. If a firm has workers close to 50 and lowers employment below 50, it is dropped from the survey even if it continues to exist. If this firm increases workers and this exceeds 50, it may re-appear in the survey. Moreover, although the percentage of collecting an effective questionnaire is relatively high, some firms that continue to exist may not return the questionnaires in some years. Although the establishment year of a firm is available, a specialized survey for exit does not exist. Furthermore, information on mergers and acquisitions is not available.

Our empirical results based on the Kitatsu data must thus be carefully interpreted, considering these limitations. Our analysis, however, tries to provide valuable information, which has not yet been sufficiently studied, by utilizing the strengths of the database as mentioned previously.

This section presents the basis statistics of Japanese manufacturing firms in our database. We first discuss subsectoral features of Japanese manufacturing firms, based on Tables 1 and 2. Table 1 shows subsectoral shares of domestic operations and trade in 2010, and Table 2 presents shares of SMEs in each subsector for each variable.⁷ In terms of the number of firms, domestic employment, domestic establishments, and domestic affiliates, major subsectors are food processing (sector 1), chemicals (sector 9), and machineries (sectors 18–21)—particularly general machinery, electric machinery, and transport equipment (18–20). Major subsectors in terms of exports are general machinery, electric machinery, and transport equipment, and major subsectors in terms of imports are petroleum and coal products (sector 10) and iron and steel (sector 15) in addition to machineries (sectors 18–21, mainly 18–20).

The majority of manufacturing firms in Japan are SMEs in terms of the number—close to 80 percent are SMEs (Table 2). SMEs’ portion becomes lower from the perspective of domestic operations; shares of SMEs are around a quarter for domestic employment (not only domestic employment in total but also employment engaged in HQ services and manufacturing activities), close to a half for domestic establishments, and a quarter for domestic affiliates. On the other hand, trade is dominated by large firms—the portion of SMEs is only five percent for exports and 10 percent for imports.

To capture the overall pattern of Japanese manufacturing MNEs, let us look at the composition of Japanese manufacturing MNEs in 2010 (Table 3). Around 90 percent of Japanese manufacturing MNEs go at least to East Asia, regardless of whether they are SMEs or large firms, indicating Japanese active investment in East Asia. Although some MNEs go to North America and/or Europe in addition to East Asia, SMEs are relatively active in East Asia (51 percent of MNEs with affiliates in East Asia) and large firms are relatively active in North America and Europe (70 percent of MNEs with affiliates in North America and 81 percent of MNEs with affiliates in Europe). In terms of subsectoral composition, around half of manufacturing MNEs are machineries, respectively. These subsectors are one of the major subsectors of Japanese manufacturing firms in general, but subsectoral shares among manufacturing MNEs are larger, compared with those in Table 1. It suggests that these subsectors are more than proportionally active abroad.

In our sample based on the panel data set for each period, the number of manufacturing MNEs slightly increases from 2,621 in 1998–2002, 2,863 in 2002–06, 2,999 in 2006–08, to 3,185 in 2008–10 in the sample, which are the sum of the number for MNE1 and MNE2 (Table 4).^8,9 MNE1 and MNE2 refer to manufacturing MNEs that increase the number of foreign affiliates in each period (expanding MNEs) and manufacturing MNEs that do not increase the number of foreign affiliates (non-expanding MNEs), respectively, and manufacturing firms other than MNEs in our data set are regarded as local firms (“Local” hereafter). As one can see in Table 4, not only large firms but SMEs are also aggressively expanding their operations abroad until the first half of the 2000s, and they are still active even in the latter half of the 2000s with the 2008–09 global financial crisis (GFC) in shorter periods (two-year-span). The shares of MNE1 in total MNEs are 45 percent in 1998–2002, 50 percent in 2002–06, 28 percent in 2006–08, and 27 percent in 2008–10. In particular, the electric machinery and transport equipment sectors are vigorous in expanding operations abroad; subsectoral shares for MNE1 in Table 4 tend to be larger than those for all-sizes of manufacturing firms in Table 3. Moreover, most of the expanding MNEs are expanding their operations at least in East Asia, except for the period 2006–08 (Table 5). In the period of 2002–06, 95 percent of expanding MNEs increase in the number of affiliates in East Asia, suggesting active expansion of operations in East Asia. On the other hand, during the same period, corresponding shares for North America and Europe in Table 5 are smaller than those in Table 3. This indicates that the share of expanding MNEs is relatively small for these regions, unlike the case of East Asia.

This section applies the JC/JD method to Japanese manufacturing firms and investigates gross and net changes in their domestic operations and trade. In particular, we shed light on changes in domestic operations and trade by MNE1, comparing with those by MNE2 and Local. In addition to gross and net changes in domestic employment, the paper also analyzes gross and net changes in employment engaged in HQ services, employment engaged in manufacturing activities, domestic establishments, domestic affiliates, exports, and imports. Furthermore, to capture the evolving features of globalizing corporate activities and domestic operations, we examine these for several sample periods, rather than focusing on only one period.

This JC/JD method is usually for the analysis of employment, but we apply it to other variables representing domestic operations and trade, in addition to domestic employment. Note that we apply the JC/JD method to the firm-level data, not the establishment/plant-level data, and thus we do not capture the JC/JD within a firm.

This subsection analyzes general trends of gross and net changes in domestic employment.¹¹ Our empirical approach is basically the “panel” decomposition, using a panel database for each period. To grasp the aggregate picture of domestic employment, however, let us first show the results of the “full” decomposition, taking entry/exit of firms into consideration. Figure 1a shows the decomposition of net change rates of domestic employment in 22 manufacturing subsectors as well as the whole manufacturing sector (shown as “total”) into four categories of gross changes, together with net changes for each subsector. Four categories for the “full” decomposition are gross job creation (C) by firms that exist at the beginning and the end of each period, gross job creation (C) by entry firms that do not appear at the beginning but exist at the end, gross job destruction (D) by existing firms, and gross job destruction (D) by exiting firms that exist at the beginning but do not appear at the end.¹² On the other hand, Figure 1b presents the contribution of each subsector to the net change rates of the whole manufacturing sector, with a distinction of four categories.

The results of “full” decomposition for domestic employment provide several interesting insights. First, domestic employment is dynamic, and the heterogeneity across firms in the adjustment of domestic employment is huge (Figure 1a). Both gross changes (C and D) are large (much larger than the net changes) not only in the whole manufacturing sector but also at subsectoral levels. For instance, C/D (−) for the whole sector are 12 percent/−25 percent in four years (1998–2002), 19 percent/−15 percent in four years (2002–06), and around ±10 percent in two years (2006–08; 2008–10).¹³ Although we need to consider possible over-counting of entry and exit, large gross changes suggest the dynamism of domestic employment. Note that whereas a number of subsectors present active gross changes, aggregate changes (at the whole manufacturing sector level) are dominated by large subsectors, namely, food processing, chemical, and machineries (general machinery, electric machinery, transport equipment, and precision machinery) (Figure 1b).

Second, the Net G extensively changes over time, reflecting changes in internal and external economic conditions. The net changes for the whole manufacturing sector in the “full” decomposition are net job destruction (−12.6 percent) for the period 1998–2002 (after the 1997–98 Asian financial crisis), net job creation (4.8 percent) for 2002–06, net job creation (2.3 percent) for 2006–08 (almost before the GFC), and even slightly net job creation (0.8 percent) for 2008–10.¹⁴ This suggests that although de-industrialization advanced in 1998–2002, it is not relevant after 2002. It also confirms that, as discussed in Section 1, it is important to investigate not only a specific sample period but also several sample periods.

Third, as Davis, Haltiwanger, and Schuh (1996) claim in the word “cyclicality,” business cycles or boom and bust in the economy affect D more than C. This is particularly noticeable at the subsectoral level; we observe extremely high D in some subsectors in downturn periods such as 1998–2002 and 2008–10. Rising industries are built up step by step whereas the scrapping of declining industries is abrupt in a recession.¹⁵

Since gross changes induced by entry/exit of firms seem to be too big in Figure 1, we focus only on the results of “panel” decomposition in the rest of the paper.¹⁶ Figure 2 shows the results of “panel” decomposition for domestic employment by distinguishing (1) SMEs from (2) large firms.

Regardless of whether we look at SMEs or large firms, net changes vary across subsectors, and gross job creation and destruction (C, D) are much larger than net changes not only at the whole sector level but also at subsectoral level, which confirms huge heterogeneity across firms again. Nevertheless, the general trend seems to differ to some extent between SMEs and large firms. In the period of 1998–2002, net changes are net job destruction for both SMEs and large firms, but net job destruction is much larger for large firms than SMEs not only for the whole manufacturing sector (C, D (−), and Net G are 5.8 percent, –13.3 percent, –7.6 percent for SMEs; and 4.7 percent, –15.6 percent, and –10.9 percent for large firms) but also for most subsectors. In the period of 2002–06, net changes are net job creation for SMEs and large firms, but both gross and net job creation are larger for large firms than SMEs (C, D (−), and Net G for the whole sector are 9.4 percent, –7.1 percent, and 2.2 percent for SMEs; and 11.1 percent, –7.4 percent, and 3.7 percent for larger firms).¹⁷ Even in the latter two periods, net changes are greater for large firms than SMEs. Although large firms have large net job creation in 2006–08 (2.2 percent for the whole manufacturing sector) and slight net job creation (0.8 percent) in 2008–10, SMEs have no net change in 2006–08 (almost zero percent) and slight net job destruction (−1.0 percent) in 2008–10.

Considering such differences in trends between SMEs and large firms, the following analysis focusing on the differences among the firm type, that is, MNE1, MNE2, and Local, is conducted separately for SMEs and large firms.

This subsection analyzes gross and net changes in domestic employment, other domestic operations, and trade by three types of firms (i.e., MNE1, MNE2, and Local) to capture distinct features of MNE1. Table 6 summarizes gross and net changes in domestic employment by the type of firm and the size of firm. Table 6 also presents those changes in employment engaged in HQ services and manufacturing activities. Figures 3a to 6a show subsectoral gross and net changes in domestic employment by the type of firm and the size of firm, and Figure 3b to 6b show the contribution of each subsector to the net change rates of the whole manufacturing sector, with a distinction of two categories, C and D.

The most interesting insight for MNE1/SMEs is that net changes for the whole manufacturing sector are larger than MNE2/SMEs or Local/SMEs in all periods, though net changes are net job destruction in 1998–2002 (−5.4 percent) and slight net job destruction in 2008–10 (−0.7 percent), and large net job creation in 2002–06 (7.7 percent) and in 2006–08 (2.6 percent). C is larger and D is smaller for MNE1/SMEs than MNE2/SMEs or Local/SMEs in all periods as well. It suggests that compared with MNE2/SMEs or Local/SMEs, MNE1/SMEs tend to expand domestic employment. In addition, net change rates for employment engaged in HQ services are larger in all periods except the first period (1998–2002) than those for domestic employment. It indicates that MNE1/SMEs intensify HQ services more than proportionally. Furthermore, although net changes (Net G) for employment engaged in manufacturing activities are net job destruction in all periods, Net G for MNE1/SMEs are larger than or almost at the same level as those for MNE2/SMEs or Local/SMEs except the last period; Net G in 2002–06 and 2006–08 are almost zero and larger for MNE1/SMEs, and Net G in period 1998–2002 (−6.9 percent) is almost at the same level of Local/SMEs. Furthermore, gross job creation (C) for employment engaged in manufacturing activities is the largest for MNE1/SMEs among three types of firms in all periods. Therefore, the size of manufacturing activities tends to slightly shrink only in the recent period. Overall, MNE1/SMEs tend to increase domestic employment in total, compared with MNE2/SMEs or Local/SMEs, and intensify HQ services, while they do not significantly decrease manufacturing employment except the period after the GFC.

On the other hand, large firms present a different picture. Whereas net job destruction in 1998–2002 for MNE1/large firms is huge (−11.9 percent), they present net job creation in the other periods, 2002–06 (4.9 percent), 2006–08 (2.7 percent), and 2008–10 (1.6 percent) even after the GFC. Net G for the whole manufacturing sector for MNE1/large firms are smaller than Local/large firms but larger than MNE2/large firms in all periods. However, except the first period (1998–2002), gross job destruction (D) for MNE1/large firms is the smallest among three types of firms not only for domestic employment but also for employment engaged in HQ services and employment involved in manufacturing activities. These suggest that the period 1998–2002 seems to be a restructuring period for MNE1/large firms, but after that, the smaller gross job destruction (D) contributes to net job creation. Similar to the case of MNE1/SMEs, HQ services by MNE1/large firms seem to be strengthened in both absolute and relative terms, particularly recently. Manufacturing activities in terms of employment significantly shrink in 1998–2002 (huge net job destruction or –17.5 percent), but they tend to expand in the absolute term in 2006–08 and 2008–10 even after the GFC, though they are likely to shrink relatively.

Before moving to the results of other domestic operations and trade, let us discuss some subsectoral features of changes in domestic employment. In 1998–2002, net change for the whole sector is net job destruction for SMEs (Figure 2) and for MNE1/SMEs (Figure 3a–i). However, 9 out of 22 subsectors have net job creation for MNE1/SMEs, which is completely different from MNE2/SMEs or Local/SMEs with net job destruction in all subsectors. In 2002–06, MNE1/SMEs have net job creation not only in the whole sector but also in most subsectors (17 out of 22 subsectors). Moreover, although net job creation for the whole sector is smaller for MNE1/large firms (4.9 percent) than Local/large firms (6.0 percent), MNE1/large firms have many subsectors with net job creation (14 out of 22 subsectors), which is slightly more than 13 subsectors for Local/large firms. Subsectoral variation seems to be quite large in both periods, 2006–08 and 2008–10, except the case of MNE1/large firms in 2008–10, which may be partly influenced by the smaller number of firms due to a shorter period.

If we look at subsectoral contribution of MNE1 (Figures 3b to 6b), the general machinery and electric machinery sectors significantly induce net job destruction and transport equipment partially compensate for it for SMEs in 1998–2002. Electric machinery and transport equipment significantly induce huge net job destruction for large firms. In 2002–06, machineries, particularly general machinery, electric machinery, and transport equipment, contribute to large net job creation by SMEs, while transport equipment significantly contributes to huge net job creation by large firms. The electric machinery contributes significantly and positively to the net change in the whole sector for SMEs in 2006–08 and 2008–10, while electric machinery and transport equipment contributes to net job creation positively and negatively, respectively, in 2006–08 and vice versa in 2008–10 for large firms.

Table 7 summarizes gross and net changes in other domestic operations and trade for the whole manufacturing sector by the type and size of firms. The major findings are as follows: First, the heterogeneity across firms is huge in terms of domestic establishments, domestic affiliates, exports, and imports, and these domestic operations and trade are also dynamic, similar to domestic employment. Both gross changes (C and D) are much larger than net changes for MNE1/SMEs. In particular, large gross creation (C) contributes to net creation for all of other domestic operations and trade in all periods, unlike MNE2/SMEs and Local/SMEs.

Second, net changes in the number of domestic establishments and domestic affiliates are net creation for MNE1/SMEs, which are greater for MNE2/SMEs or Local/SMEs, in all periods. It suggests that MNE1/SMEs tend to increase in the number of domestic establishments and domestic affiliates more significantly than other SMEs. There is a difference between domestic establishments and domestic affiliates, however; gross destruction (D) for domestic affiliates is the smallest for MNE1/SMEs among the three types of firms, whereas D for domestic establishment is not the smallest for MNE1/SMEs though they are close to the smallest types of firms.

Third, net changes for domestic establishments are net destruction in all periods for both MNE1/large firms and MNE2/large firms, although they are net creation in all periods for Local/large firms. It suggests that multinational large firms are likely to restructure (shrink) domestic operations in terms of domestic establishments while conducting activities abroad. In contrast, Local/large firms tend to increase the number of domestic establishments.

Fourth, net changes for domestic affiliates are net creation in all periods except 1998–2002 for MNE1/large firms, whereas they are net destruction in all periods for MNE2/large firms and Local/large firms. It indicates that MNE1/large firms tend to be active in expanding domestic operations in terms of domestic affiliates while expanding activities abroad, though other large firms tend to decrease the number of domestic affiliates. This might be because of the necessity of complimentary operations abroad in the case of MNE1/large firms.

Fifth, exports and imports, particularly in 1998–2002 and 2002–06, expand not only for MNEs but also local firms. Although trade expansion slows down in 2006–08 and 2008–10, both exports and imports by MNE1/SMEs still grow unlike other firms. In the case of MNE1/SMEs, gross creation (C) per se is large and close to that for Local/SMEs with the largest C. Moreover, gross destruction (D) is the smallest for both exports and imports in most of the cases. This suggests expanding export and import activities or back-and-forth transactions within the production networks by MNE1/SMEs. Note that in the case of large firms, changes in imports in some sectors, particularly the petroleum and coal products, significantly influence net change rate at the aggregate level. Also, note that data for trade in 2006–08 may already partially reflect the negative impacts of the GFC since the fiscal year for around a half of the firms is from April to March. That is, the data in 2008 for these firms is based on activities from April 2008 to March 2009. Indeed, trade changed rapidly just after the GFC, unlike other variables such as other domestic operations, though the recovery of trade was also rapid particularly in machinery sectors.¹⁸ Thus, trade data in 2006–08 may already partially reflect the negative impacts.

Figure 7 shows kernel density estimates of net change rates of domestic employment by the type of firms, separately for SMEs and large firms, in all periods. MNE1/SMEs’ density is lower at the peak around a zero change rate and seems to be biased toward the right compared with densities of other SMEs. On the other hand, MNE1/large firms’ density in 2002–06 in particular seems to be biased toward the right compared with densities of other large firms, although such a bias is not clear in other periods.

To compare net changes in domestic operations and trade for MNE1 with those for MNE2 and Local, this section applies the K-S test to their net change rates. Let F and f denote the cumulative distribution functions (CDFs) of net change rates (g) that correspond to two groups of firms to be compared. Stochastic dominance can be tested by evaluating two null hypotheses as follows:¹⁹

Tables 8 and 9 present the results of the K-S test. Similar to the analysis in the previous section based on the JC/JD method, MNE1/SMEs tend to have higher growth rates of domestic employment, compared with other SMEs, except the last period. In addition, net change rates of employment engaged in HQ services and manufacturing activities are larger for MNE1/SMEs than other SMEs until the mid 2000s. In the latter 2000s, however, it no longer holds.

As for large firms, MNE1/large firms tend to expand HQ services, compared with other large firms, except the first period. Moreover, though Local/large firms have the largest net change rates at the aggregate level in the JC/JD analysis in all periods, growth rates of MNE1/large firms tend to be higher than those of other large firms during the mid 2000s. Regarding manufacturing activities, MNE1/large firms have large change rates than other large firms only in the first period. Since the mid 2000s, we could not identify any differences in CDF between MNE1/large firms and other large firms.

As for other domestic operations and trade, net change rates are higher for MNE1/SMEs than MNE2/SMEs or Local/SMEs for domestic affiliates and exports in all periods except the last period and for imports in 1998–2002 and 2002–06. On the other hand, in the case of large firms, net changes are in most cases higher for MNE1/large firms than MNE2/large firms, but we could not find CDF for MNE1/large firms on the right side of CDF for Local/large firms.

This paper applies the JC/JD method and the K-S test to the micro data of Japanese manufacturing firms and provides a bird's eye view of the dynamism of domestic employment and domestic operations with globalizing corporate activities. Firms are classified into subsets in three dimensions: subsectors, small or large size, and multinationals expanding foreign operations/multinationals not expanding/local firms. Major findings are the following: (1) gross changes in domestic employment and domestic operations are much larger than net changes, showing restructuring dynamism and firm heterogeneity; (2) de-industrialization or the shrinkage of manufacturing sector is not relevant except for the period 1998–2002 though a slight declining trend in manufacturing activities is recently observed; (3) multinational SMEs expanding foreign operations tend to enlarge domestic employment and domestic operations, compared with other types of SMEs; (4) multinationals expanding foreign operations are likely to intensify headquarters activities within production networks; and (5) multinational SMEs expanding foreign operations tend to expand exports and imports more than other types of SMEs, suggesting that they extend active operations in international production networks, particularly in East Asia.

These results carry profound policy implications. First, in principle, de-industrialization can be stopped or at least delayed if firms are in a favorable environment for effectively utilizing the mechanics of production networks. In Japan, there exists strong public support for the globalization of corporate activities. Indeed, outward FDI, particularly for extending production networks in East Asia, is pursued not only by large firms but also by SMEs. Both central and local governments aggressively promote such FDI, and even labor unions do not oppose it. This is because people intuitively know that globalizing firms have actually generated domestic employment and operations. This paper confirms the intuition that globalizing corporate activities are not necessarily destroying jobs but can instead create domestic employment and operations.

In this regard, the improvement of location advantages is important. In an international production network, a firm allocates production processes and tasks to multiple production blocks located at home and abroad. To retain some of the domestic economic activities, the home country must be a favorable place for them; otherwise, all economic activities may move out of the country. It is the responsibility of central and local governments to investigate what sort of economic activities would be appropriate to be kept at home and enhance location advantages for them. Economic activities suitable for locating in developed countries are listed as follows: (1) headquarters functions, (2) research and development activities and pilot/mother plants, (3) highly capital-intensive activities such as system LSI (large scale integration) manufacturing plants, (4) activities utilizing agglomeration effects such as automobiles, and (5) activities that use a number of patented and black-boxed technologies such as original equipment manufacturing production of laser printers. These are just general notions, however, and we have to examine the combination of firm-specific assets of leading firms and the niches of location advantages as firms and location advantages are highly heterogeneous.

Second, the expansion of headquarters function and a sign of relative shrinkage of manufacturing activities indicate a gradual shift in the nature of domestic activities. Whether such a skill shift can be efficiently adjusted within a firm or a firm tends to replace labor in the labor market is one of the issues that we must investigate. For this, we need a specialized study because our JC/JD approach at the firm level does not directly observe the movements of individuals within a firm, either across different tasks or across establishments located in different places, or possible firing and new recruitment of labor. The accommodation of skill shift may become an important policy issue.

Another concern is the implication of the shrinkage of manufacturing activities. Can manufacturing firms survive without manufacturing activities at home? According to the theory of production fragmentation, fragmented production blocks cannot be purely capital-intensive or purely labor-intensive; to make fragmentation efficient, each production block must carry the combination of various inputs. In this regard, retained production blocks in Japan may need some manufacturing activities with factory workers. This can be a policy concern because whether a firm can keep some manufacturing activities depends on location advantages at home that include the supply of factory workers. The recent debate on possible introduction of unskilled labor from abroad may be interpreted in this context, too.

Third, our study based on the Kikatsu data for more than a decade provides some optimism over the possible de-industrialization of the Japanese economy. Nevertheless, recent policy debates after the GFC, the Great East Japan Earthquake, and formidable yen appreciation raise great concern over the poor performance of small domestic firms, particularly located in rural areas. The Kikatsu data covering firms with 50 or more workers do not show any shrinkage of workers (and regular workers as well) in the manufacturing sector as shown in Table A.1, even in 2011 and 2012. Another data source, the Economic Census, that also covers small manufacturing firms, presents quite different figures (Figure A.1), however; employment peaked in 2007, and a drastic decrease is observed up to 2011. We cannot tell what happens, but one possibility is the poor performance of small manufacturing firms after the GFC. Although this is beyond the scope of our study, more investigation is needed beyond the Kikatsu data for an assessment of recent economic performance of manufacturing firms in Japan.

Table A.1 presents the trend of corporate structure of Japanese manufacturing firms, based on the aggregated data of the Kikatsu. According to the Kikatsu data, the number of manufacturing firms gradually decreased in the latter half of the 1990s and reached a trough around 2003. After that, however, the number of firms slightly increased in the latter half of the 2000s before the GFC and slightly decreased in 2009 and 2010, and then in 2011 and 2012 the number seems to return to the level before the GFC. The number of establishments also shows a similar trend. Employment also bottomed out around 2002–03, but what is interesting is that employment tends to increase after that (even after the GFC), though it dropped in 2009. Regarding affiliates, the number of domestic affiliates seems to have a decreasing trend, whereas the number of foreign affiliates apparently tends to increase; the increase seems to have accelerated after the GFC. All of these facts suggest that, at least based on the Kikatsu data, the manufacturing sector experienced a restructuring period after the 1997–98 Asian financial crisis until 2002–03, but after that, the shrinking of the sector, including employment, does not seem to continue.

As mentioned in Section 2, however, the Kitatsu has a size cutoff; the survey covers firms with more than 50 workers. Firms with fewer than 50 workers are not included in the survey. Figure A.1 shows a rapid declining trend of hiring by manufacturing establishments with no fewer than four employees after the GFC, based on the Economic Census (note that this figure is based on the industrial classification of an establishment, not on the industrial classification of a firm). Such a large difference in trend of employment in Table A.1 and Figure A.1 may be partly explained by the decline in employment by firms with fewer than 50 workers. Because our analysis and discussion are based on the Kikatsu data, the possible trend by these small firms cannot be fully captured.

Ratios of non-regular workers corresponding to our data set are presented in Table B.1. Non-regular workers consist of part-time workers (Type 1), Hiyatoi or day workers (Type 2), and Haken or temporary agency workers (Type 3). Part-time workers are counted with regular workers in our data set whereas the latter two are not. Although part-time workers occupy 7–10 percent of total workers, they are highly concentrated in food processing. The number of Hiyatoi is limited. The significance of Haken goes up with the deregulation for the manufacturing activities in 2004 and gradually comes down after the economic downturn due to the GFC. Overall, at least in manufacturing firms covered by the Kikatsu data, non-regular workers have not been very significant over our sample period, not replacing a large portion of regular workers, except in the food processing industry.