January 2005

Exports, FDI, and Productivity of Firm: Cause and Effect*

Fukunari Kimura†

Faculty of Economics, Keio University

Kozo Kiyota‡ International Graduate School of Social Sciences, Yokohama National University

Abstract

This paper examines the relationship among exports, FDI, and productivity of firm. Using longitudinal panel data of Japanese firms, we found that the most productive firms engaged in both exports and FDI, relatively productive firms engaged in either exports or FDI, and the least productive firms stayed in the domestic market. Good firms will engage in exports and/or FDI. Furthermore, both exports and FDI improve firm’s productivity once we control for the productivity convergence effect. Firms that continue to stay in foreign markets via exports or FDI show the highest productivity growth, which also contributed to the aggregated productivity growth. (100 words) JEL Classification Code: F10 (International Trade, General), F20 (International Factor Movements and International Business, General), D21 (Firm Behavior) Keywords: Multinational Enterprises, Panel Data, Firm Heterogeneity, Total Factor Productivity, Firm Survival

* The METI database was prepared and analyzed in cooperation with the Japan Center for Economic Research and the Research and Statistics Department, the Ministry of Economy, Trade, and Industry (METI), the Government of Japan. We wish to thank Kyoji Fukao, Keiko Ito, Masahito Kobayashi, Yoshiaki Omori, Shujiro Urata, Ryuhei Wakasugi, and seminar and conference participants at METI, Research Institute of Economy, Trade, and Industry, the 2nd Yokohama Conference, the 63rd annual meeting of the Japan Society of International Economics, and Yokohama National University. Kozo Kiyota gratefully acknowledges the financial support of Grants-in-aid (B-2: 14330012) by the Ministry of Education, Culture, Sports, Science and Technology. The views expressed herein are those of the authors. † Corresponding author: Tel: +81-3-3453-4511 ext. 23215; Fax: +81-3-5427-1578; E-mail: fkimura@econ.keio.ac.jp; Address: Faculty of Economics, Keio University, 2-15-45, Mita, Minato-ku, Tokyo 108-8345, Japan. ‡ Address: International Graduate School of Social Sciences, Yokohama National University, 79-4, Tokiwadai, Hodogaya-ku, Yokohama 240-8501, Japan.

1

1. Introduction With the growing importance of firm heterogeneity, which is often represented as productivity difference among firms, recent theoretical and empirical studies in international trade began to incorporate firm heterogeneity into the model. Bernard, Eaton, Jensen, and Kortum (2003) and Melitz (2003) are the pioneering studies that theoretically clarify the relationship between exports and firm heterogeneity, represented as productivity, in the general equilibrium framework. Although these studies based on different models with each other, both explain the fact that higher the productivity of the firm, the more it is likely to be exporter.

This prediction is confirmed in many empirical studies that utilize firm/plant level longitudinal panel data.1 For instance, Bernard and Jensen examined the benefits of exports at the plant level in the United States (Bernard and Jensen, 1999, 2001, and 2004). They found that exporters presented 4-18 percent higher productivity than non-exporters and these differences were confirmed before firms exported although the impact of exports on productivity was not clear.

Similar findings are observed in many other countries: Clerides, Lach, and Tybout (1998) for Colombia, Mexico, and Morocco; Bernard and Wagner (2001) for Germany; Hallward-Driemeier, Iarossi, and Sokoloff (2002) for Indonesia, Korea, Malaysia, Philippines, and Thailand; Baldwin and Gu (2003) for Canada. Their results are summarized in Table 1. On average, the previous studies confirm that exporters show by 10-15 percent higher productivity than non-exporters. These differences are observed before firms became exporters: good firms become exporters. However, it is not always the case that a firm improves its performance after entering the export market.

=== Table 1 ===

In addition to exporting activities, foreign direct investment (FDI) recently became another important issue of discussion on globalization.2 Helpman, Melitz, and Yeaple (2004) extend Melitz’s study to incorporate FDI into trade and firm heterogeneity model. Representing firm heterogeneity by productivity difference, their model predicts that the least productive firms serve only in domestic market. Relatively more productive firms exports, and the most productive firms engage in FDI. The empirical analysis based on European firm data in 1996 partly confirms the validity of theoretical prediction in that multinational enterprises (MNEs) are substantially more

1 The pioneering micro study on exports and productivity was Aw and Hwang (1995). They used the cross-sectional firm-level data of the Taiwanese electronics industry in 1986 and found significant differences in productivity between exporters and non-exporters. 2 Lewis and Richardson (2001) for an extensive survey of the related literature as well as an heuristic explanation of the supporting argument for the benevolent effects of various channels of global commitment by corporate firms, including FDI.

2

productive than non-MNE exporters. Similarly, productivity difference between MNEs and non-MNEs, or between

foreign-owned firms and domestically-owned firms are extensively investigated in other countries: Globerman, Ries, and Vertinsky (1994) for Canada; Doms and Jensen (1998) for the United States; Grima, Thompson, and Wright (2002) for the United Kingdom; Hallward-Driemeier, Iarossi, and Sokoloff (2002) for Indonesia, Korea, Malaysia, the Philippines, and Thailand; Kimura and Kiyota (2005) for Japan. All of these studies except Globerman, Ries, and Vertinsky (1994) confirmed that foreign-owned firms presented higher productivity than domestic firms in the static sense.

The scope of most analyses is, however, limited to the static aspect, and the dynamic aspect of FDI is not fully explored yet. Do good firms engage in FDI? Does FDI improve firm’s performance? Or Both? These questions have not been tackled yet. Furthermore, the relationship between exports and FDI at the firm level is still ambiguous. As Helpman, Melitz, and Yeaple (2004) assumed, the sunk cost of starting global commitment would be different between exports and FDI, which is not explicated examined in the previous empirical studies though. More analysis is needed to clarify the relationship among trade, FDI and productivity.

This paper has three contributions. First, we investigate the dynamic aspects of FDI as well as exports. Questions addressed in this paper are whether or not good firms conduct exports/FDI, and whether or not exports/FDI improves the firm’s productivity. Second, this paper clarifies differences between exports and FDI in the link with corporate characteristics and performances. Third, we also examine the link among exports, FDI and aggregated productivity. Data used in our analysis are firm-level longitudinal panel data of Japanese firms from 1994 to 2000, which covers not only manufacturing firms but also non-manufacturing firms for approximately 22,000 observations each year.

Our major findings are summarized as follows: the majority of firms engage in neither exports nor FDI. Most of firms that engage in FDI are exporters, but exporters do not always engage in FDI. Good firms engage in FDI as well as exports. The most productive firms engage in both exports and FDI. The relatively productive firms conduct either exports or FDI while the least productive firms stay in the domestic markets. These findings support the proposition addressed by Helpman, Melitz, and Yeaple (2004). Productivity is an important factor in explaining the decision to invest abroad as well as exports. Moreover, we confirm clear evidence that exports and FDI do improve firms’ productivity once we control for a productivity convergence effect.

We further investigate gains from exports and FDI, especially focusing on the aggregated level productivity. Both exports and FDI contribute to the aggregated TFP growth through two channels. One is the own-productivity effect, and the other is the entry/exit effect. To continue exports or FDI is an important source of the own-productivity effect. In addition, exports contribute to firm survival in domestic market although FDI does not.

3

The paper is organized as follows: next section discusses the characteristics of our sample set consisting of approximately 22,000 firms, and presents static differences between exporters and non-exporters, and between firms that engage in FDI and that do not. In section 3, we examine the relationship among exports, FDI and productivity, from the dynamic aspects. Section 4 extends the analysis to the link between exports, FDI and aggregated level productivity. Section 5 concludes with implications of the results for policy and future research. 2. Exports, FDI and Productivity of Firm: Static Aspects 2.1. Data We use the micro database of Kigyou Katsudou Kihon Chousa Houkokusho (The Results of the Basic Survey of Japanese Business Structure and Activities) by the Research and Statistics Department, the Ministry of Economy, Trade and Industry (METI). This survey was first conducted in the 1991 F/Y, then in the 1994 F/Y and annually afterwards. The main purpose of the survey is to statistically capture the overall picture of Japanese corporate firms in light of their activity diversification, globalization, and strategies on R&D and information technology.

One of the strength of this survey is that it is based on firms rather than establishments. Since manufacturing census does not cover non-manufacturing establishments such as headquarter or sales blanches, the aggregation of manufacturing plants does not always depict “firm” unless it is a single-plant firm. In that sense, our data capture the activity of firms more accurately than manufacturing plant census.

Another strength is its coverage of samples and the reliability of figures. The survey is made on all firms with more than 50 employees and with more than capital of 30 million yen, covering both manufacturing and non-manufacturing firms.3 The limitation of this survey is the lack of information on some financial information.

From this survey, we develop a longitudinal panel data set for years from 1994 to 2000, based on each firm’s permanent number. We drop the firms from our sample set for which the data of age (questionnaire-level year minus establishment year), total wages, tangible assets, value-added (sales minus purchases) or the number of regular workers (including temporary workers) are not positive and the firms that re-enter the market due to incomplete replies. The number of samples ends up with over 22,000 each year. 2.2. Stylized Facts Before turning into dynamic aspects, we provide an overview of the basic statistics. Figure 1 reports the number of exporters and firms that engage in FDI in 1994 and 2000.

3 Some of non-manufacturing industries such as finance, insurance and software services are not included.

4

Exporters are in the minority. In 2000, of the roughly 22,000 firms in the database, only 20 percent (4,382 firms) reports exporting. This portion is almost the same as that of the United States (Bernard, Eaton, Jensen, and Kortum, 2003). Firms that engage in FDI are much minor compared with exporters. Only 13 percent (2,765 firms) report as engaging in FDI. Of Japanese firms, 76 percent do neither exports nor FDI. Similar results are confirmed in 1994.

=== Figure 1 ===

FDI has close relationship with exports. In 2000, out of the 2,765 firms that engage in FDI, roughly two-thirds (1,988 firms) export. On the other hand, exporters do not always engage in FDI. Out of the 4,382 exporters, only 45 percent (1,638 firms) engage in FDI. Firms that engage in FDI but not export are in the minority, accounting for 4 percent (777 firms) in our database. 2.3. Do Firms That Engage in Exports and FDI Perform Better? To make comparison across firms and time-series, we employ deterministic (non-stochastic) method in computing total factor productivity (TFP) developed by Caves, Christensen, and Diewert (1982) and extended by Good, Nadiri, Roller, and Sickles (1983). This index uses a separate hypothetical firm as a reference point for each cross-section of observations and chain-links the reference points together over time in the same way as the conventional Theil=Törnqvist index of productivity growth. Detailed of the methodology and data are summarized in the appendix.

The advantage of the multilateral index is that we do not assume any specific production function. This is because the multilateral index only requires information on output quantity as opposed to the parametric method that must specify production function. This is particularly useful when we cannot observe the production function of each firm.4

Table 2 presents export and FDI premia.5 The premium means the productivity

4 The multilateral index method is also employed by Aw, Chung, and Roberts (2000, 2003). Olley and Pakes (1996) discuss alternative productivity measure that is consistent with a dynamic and stochastic model of industry development. Levinsohn and Petrin (2003) extend Olley and Pakes (1996) such that intermediate input demand functions can play the same role as investment. Due to the data availability (our data do not cover intermediate inputs except “purchase”) though, this study employs deterministic method only. 5 Export/FDI premia are measured as the coefficients for export and FDI dummies in a regression of the form:

itititit Y εγβαθ +++= Char.sln , where θ is TFP, Y is export/FDI status that take value one if the firm is an exporter/FDI, Char.s are the vectors of firm characteristics and ε is an error term. Firm characteristics include capital-labor ratio, firm age, the number of workers, R&D expenditure-sales ratio, TFP, foreign-owned dummy as well as year and industry dummies. Foreign-owned dummy takes value one if foreign ownership is larger than 10 percent and zero

5

difference between exporters/firms that engage in FDI and non-exporter/ firms that do not, in the same industry and the same year with same characteristics. There are two major findings in this table. First, firms that engage in exports and FDI present good performance. Columns (1) and (2) present differences between exporters and non-exporters while columns (3) and (4) show differences between firms that engage in FDI and firms that do not. The differences are about 3.5-5.0 percent with statistical significance.

=== Table 2 ===

Second, there is a systematic relationship among exports, FDI, and productivity. The most productive firms engage in both exports and FDI, the second highest productive firms engage in FDI only while the third groups engage in exports only. Columns (5) and (6) present difference among firms that export only, firms that engage in FDI only and firms that engage in both exports and FDI. The largest differences are observed for firms that engage in both exports and FDI, presenting 6.4-7.9 percent higher productivity than firms do not exports and FDI. The next largest coefficients are observed for firms that engage in FDI only, indicating 4.8-6.0 percent. Firms that export but not engage in FDI present relatively low productivity, 3.9-4.2 percent. The lowest productive firms engage in neither exports nor FDI, and stay in the domestic market. 3. Exports, FDI and Productivity of Firm: Dynamic Aspects 3.1. Do Good Firms Become Exporters/Engage in FDI in the Future? We now examine the pre- and post-entry performance of Japanese firms coming into foreign markets. We focus on two channels for firms to enter the foreign market: exports and FDI. This section investigates possible determinants with which firms decide to export and/or invest abroad, based on Roberts and Tybout (1997).

Suppose that firm i exports if current and expected revenues by doing so are

greater than the costs. Denote current revenue as itR . Costs are defined as sunk cost,

F , which the firm must pay if it did not export last period, plus variable cost at period

t , itc . Assume that variable itY takes value one if firm i exports at period t and

zero for otherwise.6 That is,

otherwise. The coefficient β indicates the gap (%) between exporters and non-exporters (between firms that engage in FDI and firms that do not. Fixed-effects model is employed for the estimation (based on the results of the Hausman specification test). 6 In our dataset, exports include the sales by establishments abroad but exclude those by affiliates abroad.

6

−+>

= −

otherwise, 0);1( if 1 1

*ititit

itYFcR

Y (1)

where *itR is firm i ’s current and expected revenues if the firm exports at period t .

( )]0|)([]1|)([ 11* =•−=•+= ++ itittitittitit YVEYVERR δ , (2)

where δ means discount rate and )]([ 1 •+itVE is the expected values of maximized

pay-off conditioned by itY . The estimation method is in a binary choice of the form

>+−−+

= −= −∑otherwise, 0

;0)1( if 1 11k 10 ititK

iktkit

YFZY

µββ (3)

where 1−iktZ indicates firm-specific variables that might affect the probability of

exporting at period t . itµ represents the error term with logit distribution.

There are several estimation strategies for this dynamic binary-choice model with unobserved heterogeneity. For instance, Roberts and Tybout (1997) and Bernard and Wagner (2001) employ a probit model with random effects while Bernard and Jensen (1999) and Bernard and Wagner (2001) use a linear probability model with fixed effects. However, a linear probability model requires instruments such as two-period lags of the levels of right-hand-side variables (Bernard and Wagner, 2001). Since our sample period is not long enough to use such instruments, we employ the probit model with random effects of the form:

.11 10 ititK

k iktkit FYZY µββ +++= −= −∑ (4)

We introduce two-digit industry dummies for some of the regressions to control for industry-wise characteristics such as comparative advantages and market

conditions.7 Additional firm characteristics 1−itZ include capital-labor ratio, firm age,

the number of workers, R&D expenditure-sales ratio, TFP, and foreign-owned dummy as well as year and industry dummies.8 In order to avoid possible simultaneity problems, we lag all firm characteristics and other exogenous variables by one year.9

7 Foreign market conditions could also be important factors to affect on the decision to export and/or conduct FDI. We however do not introduce any variable for them except industry dummies due to the difficulty in obtaining detailed relevant data. 8 These firm characteristics are common to those used in estimating export/FDI premia in section 2.3. 9 For more detail, see Bernad and Jensen (1999, p.12 and footnote 19).

7

Summary statistics and correlation matrix of variables are summarized in Appendix Table 1.

We also apply this analytical framework for the decision to engage in FDI. As

for the FDI analysis, variable itY takes value one if a firm engages in FDI and zero

otherwise. In this paper, the firm that engages in FDI is defined as the firm holding one or more affiliates abroad.

Table 3 presents the regression results of equation (4) with random effects probit estimation. Columns (1) through (3) present the results for exports while columns (4) through (6) indicate the results for FDI. For exports, three main features stand out. First, good firms become exporters. The coefficients of TFP are significantly positive in all equations. This result implies that the higher the productivity the higher the probability of the firms to enter the export market. Second, sunk cost, reflected on the coefficients for lagged export dummies, also seems to be an important factor when a firm decides whether or not to enter the export market. This finding is consistent with the ones that Roberts and Tybout (1997) and Bernard and Jensen (1999) conclude. Third, potential exporters are research intensive and old. Foreign ownership is another significant factor for the decision to export.

=== Table 3 ===

Similar results are observed for the decision to FDI. Good firms will engage in FDI. Sunk costs are required to engage in FDI. Potential candidates are research intensive and old. The coefficients for the number of workers and capital-labor ratio are significantly positive in all regressions. These features are common in both the decision to exports and FDI. The difference between exports and FDI is confirmed only in the ownership structure. Domestically-owned firms are more likely to invest abroad. The sunk cost for FDI, reflected on the coefficients for lagged FDI dummies, does not seem to be very much different from the coefficients for exports. However, once separating sample firms by employment size, we find that the sunk cost for FDI seems slightly higher than that for exports for small-scale firms (less than 500 workers), which is presented in Appendix Table 2. This finding partly supports the assumption employed by the model in Helpman, Melitz, and Yeaple (2004). 3.2. Do Exports and FDI Contribute to the Productivity Growth? Next, we examine the reverse causal arrow: whether or not exports and FDI contribute to productivity growth. To test the effects of exports and FDI on productivity growth,

we run a simple regression of changes in TFP, itθ , on initial export/FDI status, itY , and

other firm characteristics, itChar.s , following the method of Bernard and Jensen (1999),

8

as

.Char.s lnln%

11

1

ititit

ititit

Y εγβαθθθ

+++=−=∆

−−

− (5)

The coefficient, β , presents the gaps in the annual average growth rate of TFP

performance between exporters/firms that invest abroad and other firms. If exports/FDI improve productivity more rapidly than non-exporters/firms that invest domestic market

only, β will be positive. Other firm characteristics for the initial year are the same as

those used in section 3.1. Table 4 presents the regression results of equation (5) with the fixed-effects

model.10 Coefficients indicate the gap of annual average TFP growth rate between exporters/FDI firms and others. Columns (1) through (3) indicate the results without controlling for initial TFP level while columns (4) through (6) are results with controlling for initial TFP level. We control for initial productivity level since higher the productivity, the lower the subsequent growth if the productivity convergence effect across firms over time works.11

=== Table 4 ===

Results of columns (1) through (3) do not strongly support the hypothesis that

exports and/or FDI improve productivity. However, once we control for the initial TFP level, a totally different story emerges. From columns (4) through (6), we can clearly confirm the positive impacts of exports and FDI on productivity growth. Exporters indicate 2.4 percent higher growth than non-exporters. Firms that engage in FDI indicate 1.8 percent higher growth than that do not. Firms that engage in both exports and FDI indicate the highest growth.

These results imply that, without controlling for initial TFP level, the regression might generate omitted variable bias such that previous studies could not fully identify

10 We choose fixed-effects model based on the results of Hausman specification test. 11 The productivity convergence model assumes that productivity in firm i evolves according to:

itititiit D εθλλγθ lnlnlnln 1 +++= − , where iγ is the asymptotic rate of productivity growth

of firm i , itD represents the catch-up variable, and itε stands for firm specific productivity

shock. The catch-up variable, itD , is defined as a function of the productivity in firm i , 1−itθ ,

relative to that in the most productive firm, *1−itθ : )ˆln()/ln(ln 1

*11 −−− −=−= ititititD θθθ , where

*111 /ˆ−−− = ititit θθθ . This formation leads to the productivity convergence equation:

ititiitit µθβαθθ ++=− −− 11ˆlnˆlnˆln with negative β . For the analysis of TFP convergence, see

Bernard and Jones (1996), Wolff (1991), and Pascual and Westermann (2002).

9

the gains from trade. Once we control for the initial TFP level, or productivity convergence effect, we can confirm that both exports and FDI do improve firm’s productivity. 4. Alternative Aspect of Gains from Exports and FDI 4.1. Aggregated Productivity: Is Reallocation Effect Important?

Recent studies also confirm that one of the gains from export activities is found in the impacts on aggregated productivity. Bernard and Jensen (2001) confirmed that the reallocation effect (i.e., the effect generated by shifting shares from low-productivity firms to high-productivity firms) was quite large in the United States such that, in the aggregate, 42 percent of overall TFP growth is from it. Moreover, over 87 percent of overall TFP growth results from the expansion of continuing exporters.

To examine the importance of reallocation effect, we decompose the annual aggregate TFP growth into within-firm, or own-productivity, and between-firm, or reallocation, effects. Define aggregate TFP as

∑≡Θi ititt v ,lnln θ (6)

where itv stands for the value added share of firm i at year t .

( ) ( ),2lnln

lnln2

lnlnlnln

11

11

111

∑∑

∑∑−

+

+−

+

≈

−=Θ−Θ

++

++

+++

i itititit

i itititit

i ititi itittt

vvvv

vv

θθθθ

θθ (7)

where the first term of the right-hand-side variables indicates the own-productivity effect while the second term is reallocation effect. The own-productivity effect is positive if the value added weighted within firm productivity growth is positive. The positive reallocation effect results from an increasing share of value added at firms with higher than average productivity.

Equation (7), however, is not applicable to unbalanced panel data containing firms entry/exit. Griliches and Regev (1995) proposed more general decomposition formula as follows:

( ) ( )

( ) ( ),2lnln

lnln2

2lnln

lnln2

lnln

11

11

11

11

1

∑∑ ∈ ++

∈ ++

++

++

+

−

+

+−

+

+

−

++−

+≈Θ−Θ

Ci itititit

Ci itititit

Xt

Et

Xt

EtX

tEt

Et

Xt

tt

vvvv

vvvv

θθθθ

θθθθ

(8)

where superscripts E and X stand for entry and exit firms, respectively. The set of all firms that remain in operation in both years is denoted as C . The first term of the right-hand-side variables indicates the TFP differences between entry and exit firms, and it is positive if the productivity of entry firms is higher than that of exit firms. The

10

second term indicates reallocation effects for entry/exits firms. The third term represents the contribution of surviving firms while the fourth terms indicate reallocation effects among surviving firms. To distinguish the entry/exit effects with surviving firms’ effects, we call the summation of first and the second terms as entry/exit effects, the third term as own effect and the forth term as reallocation effect.12

One of the advantages of this decomposition is that we can group firms into categories in another dimension (Bernard and Jensen, 2001). We further decompose equation (8) into four groups based on firms’ export/FDI status in two consecutive years (start, both, stop and neither) as

( ) ( )

( ) ( ),2lnln

lnln2

2lnln

lnln2

lnln

11

11

11

11

1

∑ ∑∑ ∑ ∈ ++

∈ ++

++

++

+

−

++−

++

−

++−

+≈Θ−Θ

j Cij

itj

it

jit

jit

j Cij

itj

it

jit

jit

Xt

Et

Xt

EtX

tEt

Et

Xt

tt

vvvv

vvvv

θθθθ

θθθθ

(9)

where }Neither Stop, Both, Start,{∈j represents the status of firm i between years t

and 1+t . Table 5 presents the decomposition exercise results. Contrary to the finding of

Bernard and Jensen (2001), our results do not support the importance of reallocation. Overall TFP grew at average annual rate of 5.7 percent from 1994 to 2000. While the dominant source of aggregate TFP growth was the own-productivity effect, accounting for 69 percent of the total, changes results from entry/exit were important in overall growth. Of aggregate TFP growth, 42 percent came from the entry/exit effect. On the other hand, the reallocation effect was relatively small and shows negative contribution, accounting for –10 percent of the total. These estimates suggest that own-productivity growth and entry/exit rather than reallocation are important sources of aggregate productivity growth in Japan.13

=== Table 5 ===

We also confirm positive contributions of exports and FDI in the own-productivity effect. The decomposition results indicate that 48 percent of aggregate TFP growth result from the growth of continuing exporters and 46 percent of aggregate TFP growth come from the growth of continuing firms that invest abroad. Surviving in foreign markets seems to be an important source for the aggregated productivity growth. 4.2. Is “Survival” in Foreign Markets Important? 12 Various permutation of this decomposition are found in the literature. For more detail, see Aw, Chen, and Roberts (2003). 13 Aw, Chen, and Roberts (2003) also find that within-firm productivity growth and entry/exit play large roles in productivity growth in Taiwan while reallocation across plants plays only a minor role.

11

To statistically test the importance of surviving in foreign markets, we run the similar regression to equation (5), including the following three dummy variables:

,Char.sStopBothStart lnln%

1321

1

ititititit

ititit

εγβββαθθθ

+++++=−=∆

−

− (10)

where )1(*)0( if 1Start 1 === − ititit YY

)1(*)1( if 1Both 1 === − ititit YY

)0(*)1( if 1Stop 1 === − ititit YY

Other firm characteristics variables are the same as those used in section 3.2, with controlling for the productivity convergence effect. If continue to export and FDI, or to “survive” in foreign market is important for the productivity growth of the firm, the

coefficients 2β should be positive and larger than 1β and 3β .

Table 6 presents the regression results. All coefficients of “Start,” “Both,” and “Stop” indicate the positive and statistically significant signs for both exports and FDI. However, the coefficients of “Both” are larger than those of “Start,” and the coefficients of “Stop” are the smallest. These results imply that firms that continue to export and/or conduct FDI present the highest growth while firms that just enter foreign market via exports and/or conduct FDI show relatively modest growth. The firms that exit from foreign markets indicate relatively low TFP growth though their growth is still higher than the firms that stay in domestic market. To continue to export and FDI, or to “survive” in foreign market, is particularly important for the productivity growth of a firm, which ends up with a major source of aggregated productivity growth.

=== Table 6 === 4.3. Do Exports and FDI Contribute to Survive in the Domestic Market?

In section 4.2, Table 5 confirms that entry/exit is another important source of aggregate productivity growth in Japan. One may ask whether or not exports and FDI contribute to the firm survival. To answer this question, we estimate Cox’s proportional hazard model of the form

),exp(0 itititit ZYhh γβ += (11)

where ith is firm i ’s hazard rate and ith0 is a baseline hazard rate.14 Hazard rate is

14 The reason why the equation does not have intercept is related to the fact that the baseline hazard cannot be explicitly specified (and therefore not estimated). Since

12

defined as “the rate at which spells are completed after duration t , given that they last

at least until t (Greene, 2003).” itY represents the export/FDI status in year t , and

itZ is a vector of corporate characteristics in year t and other variables used in

Section 3.1. The advantage of this analysis is that the analysis requires no prior assumption

on the distribution of exit timing. Also, the hazard rate directly captures the probability that a firm will exit in the next short interval of time given that it survives until time t . Estimated coefficients have the interpretation of the ratio of the hazards for one-unit change in the corresponding covariate (vector). For instance, suppose that we focus on

the FDI status and we have 09.0=β . This means that firms that engage in FDI face a

hazard 10 percent greater than firms that do not since 10.1)09.0exp( ≈ . On the other

hand, if we have 11.0−=β , firms that engage in FDI face a hazard 10 percent lower

than firms that do not ( 9.0)11.0exp( ≈− ). If exports/FDI have positive contribution to

the firm survival, the coefficient β must be a significantly negative coefficient.

Table 7 presents the regression results of equation (11). Exports have positive impacts on firm survival. The hazard rates of exports are 0.82-0.93, meaning that exporters face the hazard rate by 7-18 percent lower than non-exporters.

=== Table 7 ===

On the other hand, FDI has negative impacts on firm survival. The hazard rates

of firms that engage in FDI present 1.09-1.23, meaning that FDI firms face hazard rate by 9-23 percent greater than non-FDI firms. FDI might impose financial or managerial burden on some firms, particularly on small- and medium-sized enterprises, so that the probability of exiting would be higher.15 5. Concluding remarks

)exp()}exp({)exp( 00 ititititititit ZYhZYhh γβαγβα +=++= , it simply changes the baseline

hazard from ith0 to )}exp({ 0 αith , both of which we do not define. Hence, the any value works for α . 15 Kimura and Fujii (2003) found that small firms could benefit from exports but conducting FDI aggravated rather than assisted performance.

13

In this paper, we examine the relationship between exports, FDI and productivity using the firm level longitudinal panel data in Japan between 1994 and 2000. Our major findings are summarized as follows: firms that engage in exports and FDI are in the minority. Most of firms that engage in FDI are exporters while exporters do not always engage in FDI. Good firms engage in FDI as well as exports. The most productive firms engage in FDI while the relatively productive firms export. The least productive firms neither export nor invest abroad, and stay in the domestic markets. Productivity is an important factor in explaining the decision to FDI as well as export. Moreover, both exports and FDI do improve firm productivity once we control for the productivity convergence effect.

Gains from exports and FDI are confirmed at the aggregated level, too. Export and FDI contribute to the aggregated TFP growth through two channels. The one is through own-productivity effect, and the other is through the entry/exit effect. The own- productivity growth is found in firms that continue to export and/or invest abroad. Exports have positive impact on firm survival while FDI does not.

The importance of reallocation effect between high-productive and low-productive firms is not confirmed in our study. One possible explanation is the existence of the high adjustment costs of factor reallocation in Japan. For instance, the unobservable barriers to labor movement across firms might lower reallocation effect. Japanese structural reform is still under way. To get full gains from exports and FDI, we might need to promote reallocation through various policy channels.

Positive impacts of exports and FDI on the productivity growth have another policy implication. To “survive” in foreign market are likely to result in higher productivity growth for both each firm and aggregated level. Hence, to facilitate exports or to enhance global commitment would have some validity though we need more discussion on its implementation and sequencing. We should also keep it in mind that FDI possibly has negative impacts on firm survival.

Further research agenda must include more thorough investigation on the mechanics of interaction between global commitment and productivity. One possible clue may be found in examining differential effects of various types of global commitment. We know that various types of international trade exist; inter-industry trade, vertical and horizontal intra-industry trade, arm’s-length/intra-firm trade, and so on. There also exist different types of FDI; horizontal FDI, vertical FDI, manufacturing/non-manufacturing FDI, and so on. Although serious data limitation exists in this field, more penetrating fact-finding would provide a useful feedback for theoretical thought. References Aw, B.-Y. and A.R. Hwang. 1995. “Productivity and the Export Market: A Firm Level

Analysis,” Journal of Development Economics, 47(2): 313-332.

14

Aw, B.-Y., X. Chen, and M.J. Roberts. 2003. “Firm-level Evidence on Productivity Differentials and Turnover in Taiwanese Manufacturing,” Journal of Development Economics, 66(1): 51-86.

Aw, B.-Y., S. Chung, and M.J. Roberts. 2000. “Productivity and Turnover in the Export Market: Micro Evidence from Taiwan and South Korea,” World Bank Economic Review, 14(1): 65-90.

Aw, B.-Y., S. Chung and M.J. Roberts. 2003. “Productivity, Output, and Failure: A Comparison of Taiwanese and Korean Manufactures,” Economic Journal, 113: F485-510.

Baldwin, J.R. and W. Gu. 2003. “Export-market Participation and Productivity Performance in Canadian Manufacturing,” Canadian Journal of Economics, 36(3): 634-657.

Bernard, A.B., J. Eaton, J.B. Jensen, and S. Kortum. 2003. “Plants and Productivity in International Trade,” American Economic Review, 93(4): 1268-1290.

Bernard, A.B. and J.B. Jensen. 1999. “Exceptional Exporter Performance: Cause, Effect, or Both?” Journal of International Economics, 47(1): 1-26.

Bernard, A.B. and J.B. Jensen. 2001. “Exporting and Productivity: The Importance of Reallocation,” U.S. Census Bureau, Center for Economic Studies, Working Paper 01-02.

Bernard, A.B. and J.B. Jensen. 2004. “Why Some Firms Export?” Review of Economics and Statistics, 86(2): 561-569.

Bernard, A.B. and C.I. Jones. 1996. “Comparing Apples to Oranges: Productivity Convergence and Measurement Across Industries and Countries,” American Economic Review, 86(5): 1216-1239.

Bernard, A.B. and J. Wagner. 2001. “Export Entry and Exit by German Firms,” Weltwirtschaftliches Archiv (Review of World Economics), 137(1): 105-123.

Caves, D.W., L.R. Christensen, and W.E. Diewert. 1982. “Output, Input, and Productivity Using Superlative Index Numbers,” Economic Journal, 92(365): 73-86.

Clerides, S.K., S. Lach, and J.R. Tybout. 1998. “Is Learning-by- Exporting Important? Micro-Dynamic Evidence from Colombia, Mexico and Morocco,” Quarterly Journal of Economics, 113(3): 903-948.

Doms, M. and J.B. Jensen. 1998. “Comparing Wages, Skills, and Productivity between Domestically and Foreign-Owned Manufacturing Establishments in the United States,” in R.E. Baldwin, R.E. Lipsey, and J.D. Richardson. (eds.), Geography and Ownership as Bases for Economic Accounting, Chicago, IL: University of Chicago Press.

Globerman, Steven, J. Ries and I. Vertinsky. 1994. “The Economic Performance of Foreign Affiliates in Canada,” Canadian Journal of Economics, 27(1): 143-156.

Good, D.H., M.I. Nadiri, L.-H. Roller, and R.C. Sickles. 1983. “Efficiency and Productivity Growth Comparisons of European and U.S. Air Carriers: A First

15

Look at the Data,” Journal of Productivity Analysis, 4(1-2): 115-125. Greene, W.H. 2003. Econometric Analysis, 5th edition, Upper Saddle River, NJ: Prentice

Hall. Grima, S., S. Thompson, and P.W. Wright. 2002. “Why Are Productivity and Wages

Higher in Foreign Firms,” The Economic and Social Review, 33(1): 93-100. Griliches, Z. and H. Regev. 1995. “Firm Productivity in Israeli Industry, 1979-1988,”

Journal of Econometrics, 65(1): 175-203. Hallward-Driemeier, M., G. Iarossi, and K.L. Sokoloff. 2002. “Exports and

Manufacturing Productivity in East Asia: A Comparative Analysis with Firm-level Data,” NBER Working Paper, No. 8894.

Helpman, E., M.J. Melitz and S.R. Yeaple. 2004. “Export versus FDI with Heterogeneous Firms,” American Economic Review, 94(1): 300-316.

Kimura, F. and T. Fujii. 2003. “Globalizing Activities and Rate of Survival: Panel Data Analysis on Japanese Firms,” Journal of the Japanese and International Economies, 17(4): 538-560.

Kimura, F. and K. Kiyota. 2005. “Foreign-owned versus Domestically-owned Firms: Economic Performance in Japan,” forthcoming in Review of Development Economics.

Levinsohn, J. and A. Petrin. 2003. “Estimating Production Functions Using Inputs to Control for Unobservables,” Review of Economic Studies, 70(2): 317-341.

Lewis, III, H. and J.D. Richardson. 2001. Why Global Commitment Really Matters! Washington, DC: Institute for International Economics.

Melitz, M.J. 2003. “The Impact of Trade on Aggregate Industry Productivity and Intra-Industry Reallocations,” Econometrica, 71(6): 1695-1725.

Ministry of Economy, Trade and Industry (METI) (Research and Statistics Department) (various years) Kigyou Katsudou Kihon Chousa Houkokusho (the Results of the Basic Survey of Japanese Business Structure and Activities), Tokyo: Shadanhoujin Tsuusan Toukei Kyoukai.

Nishimura, K.G., T. Nakajima, and K. Kiyota. 2005. “Does the Natural Selection Mechanism Still Work in Severe Recessions? - Examination of the Japanese Economy in the 1990s,” forthcoming in Journal of Economic Behavior and Organization.

Olley, G. S. and A. Pakes. 1996. “The Dynamics of Productivity in the Telecommunications Equipment Industry,” Econometrica, 64(6): 1263-1297.

Pascual, A.G. and F. Westermann. 2002. “Productivity Convergence in European Manufacturing,” Review of International Economics, 10(2): 313-323.

Roberts, M.J. and J.R. Tybout. 1997. “The Decision to Export in Colombia: An Empirical Model of Entry with Sunk Costs,” American Economic Review, 87(4): 545-564.

Wolff, E.N. 1991. “Capital Formation and Productivity Convergence Over the Long Term,” American Economic Review, 81(3): 565-579.

16

Appendix: Construction of Multilateral TFP Index Methodology

We use the multilateral TFP index developed by Caves, Christensen, and Diewert (1982) and extended by Good, Nadiri, Roller, and Sickles (1983). The multilateral index relies on a single reference point, which is constructed as a hypothetical firm that has the arithmetic mean values of log output, log input, and input cost shares over firms in Japan in each year. Each firm’s logarithmic output and inputs are measured relative to this reference point each year, and then the reference points are chain-linked over time. Suppose that the TFP of the hypothetical firm is equal to one. The TFP index for firm i in year t is defined as

( ) ( )( )( ) ( )( )∑ ∑∑

∑

= = −−=

= −

−+−−+−

−+−≈

t J

j jjJ

j titjtijt

ttitit

XXssXXss

QQQQ

2 1 111

2 1

lnln21lnln

21

lnlnlnlnln

τ ττττ

τ ττθ

where itQln , itXln , and ijts are the log output, log input of factor j , and the cost

share of factor j for firm i , respectively. tQln , tXln , and jts are the values of

the hypothetical reference firm in year t and are equal to the arithmetic means of corresponding variables over all firms in the year. The first term on the right hand side variable is the deviation of the firm’s output from the output of the reference point in the industry in year t , and the second term is the cumulative change in the output reference point between year t and the initial year, 1=t . The last two terms perform the same

operation for each factor input j , and are weighted by the average of the cost shares

for firm i and the reference point in year t . The index measures TFP of each firm in each year relative to that of the hypothetical firm in the initial year. Data

Outputs are defined as value-added while inputs are capital and labor. The advantage of the use of value added is its aggregation property. It is difficult to aggregate firm level productivity to the whole economy level if we use gross output; value added can be directly comparable across industries. As for other additional data and their manipulation, we adopt the methodology described in the Appendix in Nishimura, Nakajima, and Kiyota (2005).

Figure 1. Exports and FDI of Japanese Firms, 1994 and 2000

1994 (21,232 firms) 2000 (21,661 firms)

Source: METI database.

FDI only(777)3.6%

Exports andFDI (1,988)

9.2%

Exports only(2,394)11.1%

Domesticfirms

(16,502)76.2%

FDI only(664)3.1%

Exports andFDI (1,638)

7.7%

Exports only(2,485)11.7%

Domesticfirms

(16,445)77.5%

Table 1. Export, FDI and Productivity of Firm: Summary Results of Previous Studies

Exports and productivityAuthors Roberts and Tybout (1997) Clerides, Lach, and Tybout

(1998)Bernard and Jensen (1999) Aw, Chung, and Roberts (2001) Bernard and Jensen (2001)

Country Colombia Colombia, Mexico, and Morocco United States Korea and Taiwan United States

Unit Plants Firms and plants Plants Firms and plants PlantsPeriod 1981-1989 1981-1991 (Colombia); 1986-

1990 (Mexico); 1984-1991(Morocco)

1984-1992 1981-1991 (Taiwan); 1983-1993(Korea)

1983-1992

Industry Manufacturing Manufacturing (Table 7) Manufacturing Manufacturing ManufacturingNot examined Yes 3.5-18.1% (Table 1) 3.9-31.1% (Table 2) 8-9% (p.9)

With control? Not examined Yes Yes No NoYes Yes Yes Yes for Taiwan YesNot examined No No No for both countries No

Authors Bernard and Wagner (2001) Baldwin and Gu (2003) Bernard, Eaton, Jensen, andKortum (2003)

Hallward-Driemeier, Iarossi, andSokoloff (2002)

Bernard and Jensen (2004)

Country Germany Canada United States Indonesia, Korea, Malaysia,Philippines, and Thailand

United States

Unit Plants Plants Plants Plants PlantsPeriod 1978-1992 1974-1996 1992 1996-1998 1984-1992Industry Manufacturing Manufacturing Manufacturing Manufacturing Manufacturing

Not examined 4.0-15.0% (p.641) 15-33% (Table 2) 1) Yes 11-12% (Table 3) 1)With control? Not examined Yes Yes Yes Yes

Yes 1) Yes Not examined Not examined Yes (but not robust)Not examined Yes Not examined Not examined Not examined

FDI and productivityAuthors Globerman, Ries, and Vertinsky

(1994)Doms and Jensen (1998) Grima, Thompson, and Wright

(2002)Hallward-Driemeier, Iarossi, andSokoloff (2002)

Kimura and Kiyota (2004)

Country Canada United States United Kingdom Indonesia, Korea, Malaysia,Philippines, and Thailand

Japan

Unit Plants Plants Firm Plants FirmsPeriod 1986 1987 1989-1994 1996-1998 1994-1997Industry Manufacturing Manufacturing Manufacturing Manufacturing All industry 2)

Not significant difference (Table2) 1)

2.3-2.4% (Table 7.4) 14% (Table 2) Yes 67.2-101.8% (Table 3)

With control? Yes Yes Yes Yes NoNotes:

2) Except some non-manufacturing industries such as agriculture, finance and insurance, software.1) Labor productivity

Do exporters perform better?

Do good firms (plants) become exporters?Do exports improve productivity?

Do exporters perform better?

Do foreign-owned firms perform better thandomestic firms?

Do good firms (plants) become exporters?Do exports improve productivity?

Table 2. Export and FDI Premia Table 3. Decision to Exports and FDI, 1994-2000

Dependent variable: TFP level All firms(1) (2) (3) (4) (5) (6) Dependent variable: Export dummy (t) FDI dummy (t)

Export dummy 3.95 3.54 (1) (2) (3) (4) (5) (6)[7.05] [6.37] Export dummy 3.10*** 3.01*** 2.91*** 0.62***

FDI dummy 5.04 3.79 [216.98] [204.35] [192.44] [34.40][7.79] [5.89] FDI dummy 0.59*** 3.05*** 3.00*** 2.82***

FDI & Export dummy 8.00 6.43 [28.77] [176.54] [171.73] [155.89][9.38] [7.57] TFP 0.18*** 0.16*** 0.13*** 0.17*** 0.14*** 0.10***

FDI only dummy 5.96 4.82 [14.64] [12.52] [10.43] [12.82] [10.16] [6.60][6.81] [5.54] Number of workers 0.09*** 0.13*** 0.06*** 0.22*** 0.26*** 0.24***

Export only dummy 4.16 3.86 [12.37] [17.69] [7.67] [28.42] [31.45] [28.95][6.83] [6.37] Capital-labor ratio (millions of yen, 1994 prices) 0.02*** 0.02*** 0.01 0.07*** 0.07*** 0.07***

Year dummy No Yes No Yes No Yes [3.95] [3.30] [1.05] [10.02] [9.47] [9.07]Industry dummy No Yes No Yes No Yes R&D expenditure-sales ratio (%) 0.05*** 0.04*** 0.04*** 0.03*** 0.02*** 0.01**Firm characteristics Yes Yes Yes Yes Yes Yes [17.76] [12.06] [10.98] [11.90] [7.36] [2.47]N 153,147 153,147 153,147 153,147 153,147 153,147 Age 0.11*** 0.09*** 0.07*** 0.12*** 0.10*** 0.05***R2 0.040 0.050 0.040 0.050 0.040 0.050 [8.28] [6.61] [5.20] [7.37] [5.88] [3.05]Notes: 1) Fixed-effect model is used for estimation. Foreign owned dummy 0.27*** 0.24*** 0.27*** -0.08 -0.09* -0.21***

2) All differences are significant at 1% level and figures in brackets indicate t-statistics. [6.28] [5.46] [6.28] [1.53] [1.88] [4.43]Constant -2.79*** -3.19*** -2.76*** -3.75*** -4.12*** -3.89***

[51.88] [40.34] [34.14] [58.30] [46.37] [43.86]Source: METI Database Year dummy No Yes Yes No Yes Yes

Industry dummy No Yes Yes No Yes YesN 121,825 121,825 121,825 121,825 121,825 121,825AIC 0.323 0.312 0.305 0.244 0.240 0.230Log-Likelihood -19682.0 -18971.9 -18554.2 -14878.1 -14571.6 -13994.0Notes: 1) Random-effect probit model is used for estimation.

2) ***, **, * indicate level of significance at 1%, 5% and 10% and figures in brackets indicate t-statistics.

Source: METI Database

3) Estimated coefficients indicate the gap (%) of TFP level between exporters/firmsengages in FDI and other firms

3) All independent variables are at period t-1. We take natural log for TFP, number of workers, capital-laborratio and age.

Table 4. Effects of Exports and FDI on TFP Growth Table 5. Decomposition of Aggregated TFP Growth by Export/FDI Status

Dependent variable: annual average TFP growth by Export status by FDI statusAll firms Export/FDI status Growth rates % of total growth Growth rates % of total growth

Overall 0.0573 100.0% 0.0573 100.0%(1) (2) (3) (4) (5) (6) Entry/exit 0.0238 41.5% 0.0238 41.5%

Export dummy -0.20 2.41*** Own-Productivity effect 0.0393 68.5% 0.0393 68.5%[0.24] [3.86] Stop (1, 0) 0.0010 1.8% 0.0011 2.0%

FDI dummy -0.69 1.83** Both (1, 1) 0.0277 48.2% 0.0264 46.1%[0.74] [2.57] Start (0, 1) 0.0009 1.6% -0.0004 -0.8%

FDI & Export dummy -0.31 3.91*** Neither (0, 0) 0.0097 16.8% 0.0121 21.2%[0.25] [4.14] Reallocation effect -0.0057 -10.0% -0.0057 -10.0%

FDI only dummy -2.19* 1.84* Stop (1, 0) 0.0009 1.5% 0.0001 0.2%[1.74] [1.91] Both (1, 1) -0.0048 -8.3% -0.0079 -13.8%

Export only dummy -0.76 2.39*** Start (0, 1) -0.0004 -0.7% 0.0000 -0.1%[0.86] [3.52] Neither (0, 0) -0.0014 -2.5% 0.0021 3.6%

Year dummy Yes Yes Yes Yes Yes Yes Source: METI database.Industry dummy Yes Yes Yes Yes Yes YesFirm characteristics Yes Yes Yes Yes Yes YesInitial TFP level No No No Yes Yes YesN 121,825 121,825 121,825 121,825 121,825 121,825R2 0.02 0.02 0.02 0.43 0.43 0.43Notes: 1) Fixed-effect model is used for estimation.

Source: METI Database

2) ***, **, * indicate level of significance at 1%, 5% and 10% and figures inbrackets indicate t-statistics.3) Estimated coefficients indicate the gaps of growth rate betweenexporters/firms engage in FDI and other firms.

Without control of"convergence" effect

With control of "convergence"effect

Table 6. Effects of Exports and FDI on TFP Growth Table 7. Firm Survival

Dependent variable: annual average TFP growth (1) (2) (3) (4) (5)Export dummy -0.20*** -0.07* -0.12***

(1) (2) [5.41] [1.78] [2.94]Exports, Start 2.19*** FDI dummy 0.09* 0.16*** 0.21***

[2.67] [1.85] [3.18] [4.00]Exports, Both 3.89*** TFP -0.20*** -0.19*** -0.22*** -0.20*** -0.19***

[4.74] [10.04] [9.07] [11.07] [9.62] [9.27]Exports, Stop 2.49*** Number of workers -0.64*** -0.65*** -0.66*** -0.66*** -0.66***

[2.85] [35.85] [35.45] [36.25] [35.84] [35.63]FDI, Start 2.41** Capital-labor ratio (millions of yen, 1994 prices) -0.13*** -0.09*** -0.14*** -0.10*** -0.09***

[2.44] [14.19] [9.73] [14.52] [10.00] [9.95]FDI, Both 4.36*** R&D expenditure-sales ratio (%) -0.01 0.00 -0.03** 0.00 0.00

[4.58] [1.19] [0.53] [2.49] [0.00] [0.36]FDI, Stop 0.43 Age -0.75*** -0.72*** -0.76*** -0.73*** -0.73***

[0.39] [38.20] [29.60] [38.35] [29.66] [29.62]Year dummy Yes Yes Foreign owned dummy 0.56*** 0.48*** 0.50*** 0.46*** 0.50***Industry dummy Yes Yes [5.80] [4.96] [5.19] [4.76] [5.09]Firm characteristics Yes Yes Year dummy No Yes No Yes YesInitial TFP level Yes Yes Industry dummy No Yes No Yes YesN 121,825 121,825 Hazard rateR2 0.131 0.130 Exports 0.82 0.93 0.89For notes and sources, see Table 4. FDI 1.09 1.17 1.23

N 118,415 118,415 118,415 118,415 118,415AIC 0.907 0.871 0.871 0.907 0.871Log-Likelihood -53690.1 -51514.2 -51506.5 -53703.7 -51510.9Notes: 1) Cox's proportional hazard model is used for estimation.

2) ***, **, * indicate level of significance at 1%, 5% and 10% and figures in brackets indicate t-statistics.

Source: METI Database

Appendix Table 1. Summary Statistics and Correlation MatrixSummary statistics Correlation matrix

VariableN Mean S.D. FDID EXPD FDIEXP

DFDIonlyD

EXPonlyD

L KL RDS Age FOD TFP

FDI dummy 153,147 0.12 0.32 1.000Export dummy 153,147 0.20 0.40 0.470 1.000FDI & Export dummy 153,147 0.09 0.28 0.829 0.608 1.000FDI only dummy 153,147 0.03 0.18 0.511 -0.095 -0.058 1.000Export only dummy 153,147 0.12 0.32 -0.134 0.722 -0.111 -0.068 1.000L 153,147 5.17 0.99 0.367 0.231 0.328 0.151 0.003 1.000Capital-labor ratio (millions of yen, 1994 prices) (KL) 153,147 1.65 1.31 0.140 0.103 0.117 0.070 0.027 0.101 1.000R&D expenditure-sales ratio (%) (RDS) 153,147 0.51 1.81 0.206 0.268 0.229 0.017 0.136 0.198 0.087 1.000Age 153,147 3.47 0.57 0.153 0.137 0.145 0.051 0.046 0.129 0.300 0.056 1.000Foreign owned dummy (FOD) 153,147 0.02 0.15 0.103 0.164 0.104 0.023 0.114 0.159 0.030 0.123 -0.030 1.000Total factor productivity (TFP, natural log) 153,147 -0.01 0.62 0.148 0.193 0.144 0.043 0.116 0.065 -0.118 0.096 -0.014 0.139 1.000Variable N Mean S.D. FDID EXPD TFP L KL RDS AGE FODFDI dummy (t-1) 131,486 0.12 0.32 1.000Export dummy (t-1) 131,486 0.20 0.40 0.467 1.000TFP (t-1) 131,486 -0.02 0.60 0.148 0.191 1.000L (t-1) 131,486 5.17 0.98 0.370 0.236 0.066 1.000Capital-labor ratio (millions of yen, 1994 prices) (KL) (t-1) 131,486 1.65 1.28 0.141 0.103 -0.096 0.108 1.000R&D expenditure-sales ratio (%) (RDS) (t-1) 131,486 0.57 1.86 0.220 0.287 0.105 0.207 0.094 1.000Age (t-1) 131,486 3.47 0.56 0.151 0.135 -0.010 0.129 0.292 0.062 1.000Foreign owned dummy (FOD) (t-1) 131,486 0.02 0.14 0.093 0.162 0.133 0.152 0.030 0.132 -0.030 1.000Note: For the definition of variables, see main text.Source: METI database.

Appendix Table 2. Decision to Exports and FDI, by the Employment Scale

less than 100 workers 100-500 workers 500-1000 workers greater than 1000 workers

Dependent variable:Exportdummy (t)

FDIdummy (t)

Exportdummy (t)

FDIdummy (t)

Exportdummy (t)

FDIdummy (t)

Exportdummy (t)

FDIdummy (t)

(a1) (a2) (a3) (a4) (a5) (a6) (a7) (a8)Export dummy 3.07*** 2.97*** 3.04*** 3.04***

[109.12] [150.50] [60.39] [53.65]FDI dummy 3.12*** 3.00*** 2.86*** 2.91***

[77.49] [126.69] [61.67] [55.52]TFP 0.18*** 0.15*** 0.16*** 0.13*** 0.14*** 0.13*** 0.06 0.16***

[7.50] [4.87] [9.07] [6.89] [3.30] [3.28] [1.32] [3.54]Number of workers 0.08 0.31*** 0.14*** 0.27*** 0.09 0.22*** 0.00 0.26***

[1.41] [4.11] [6.95] [11.91] [0.96] [2.62] [0.12] [6.39]Capital-labor ratio (millions of yen, 1994 prices) 0.01 0.05*** 0.03*** 0.07*** 0.05* 0.12*** 0.05* 0.15***

[1.15] [3.41] [3.01] [6.36] [1.83] [4.82] [1.82] [5.17]R&D expenditure-sales ratio (%) 0.03*** 0.03*** 0.04*** 0.02*** 0.06*** 0.04*** 0.06*** 0.03***

[3.88] [3.15] [9.20] [4.84] [4.63] [3.35] [5.06] [2.99]Age 0.08*** 0.04 0.09*** 0.08*** 0.19*** 0.25*** 0.09 0.21***

[3.26] [1.11] [4.77] [3.69] [3.64] [4.83] [1.59] [3.76]Foreign owned dummy 0.41*** -0.22 0.30*** 0.04 0.16 -0.22* 0.14 -0.14

[3.95] [1.31] [4.68] [0.51] [1.15] [1.66] [1.53] [1.49]Constant -2.96*** -4.03*** -3.15*** -4.15*** -3.35*** -4.21*** -2.23*** -4.53***

[9.76] [10.52] [22.70] [25.71] [5.29] [7.19] [6.51] [12.49]Year dummy Yes Yes Yes Yes Yes Yes Yes YesIndustry dummy Yes Yes Yes Yes Yes Yes Yes YesN 39,837 39,837 64,397 64,397 9,675 9,675 7,916 7,916AIC 0.268 0.153 0.330 0.252 0.343 0.402 0.349 0.378Log-Likelihood -5299.8 -3022.2 -10601.8 -8086.9 -1623.1 -1908.4 -1348.6 -1462.5Note: See Table 3.Source: METI Database