Internal capital market and the growth and survival of Japanese plants in the United States

J. Japanese Int. Economies 19 (2005) 366–385 www.elsevier.com/locate/jjie

Internal capital market and the growth and survival of Japanese plants in the United States Tatsuo Ushijima Aoyama Gakuin University, Graduate School of International Management, Shibuya 4-4-25, Shibuya-ku, Tokyo 150-8366, Japan Received 27 May 2003; revised 19 October 2003 Available online 28 February 2004

Ushijima, Tatsuo—Internal capital market and the growth and survival of Japanese plants in the United States This article examines the growth and survival of foreign plants based on a cross-section of Japanese plants in the United States. Results show that plant turnovers are significantly related to plant age and size in a manner highly consistent with the Bayesian learning model of firm dynamics (e.g. Jovanovic [Econometrica 50 (1982) 649]). They also indicate that financial conditions prevailing at the Japanese parent affect plant dynamics importantly, suggesting the working of internal capital markets across the Pacific. Results, however, are ambiguous with regard to the efficiency of internal capital allocation. J. Japanese Int. Economies 19 (3) (2005) 366–385. Aoyama Gakuin University, Graduate School of International Management, Shibuya 4-4-25, Shibuya-ku, Tokyo 150-8366, Japan.  2004 Elsevier Inc. All rights reserved. JEL classification: F23; G31; L11 Keywords: Multinational corporation; Firm turnover; Foreign direct investment; Internal capital market

1. Introduction It is often argued that Japanese firms are extremely slow in discarding failing operations. The management buzz word “sentaku to shuchu” meaning selection and concentration, E-mail address: [email protected]. 0889-1583/$ – see front matter  2004 Elsevier Inc. All rights reserved. doi:10.1016/j.jjie.2003.12.004

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

367

which swept Japan’s management community in the late 1990s in the wave of corporate restructuring, is highly suggestive of this problem. It implies that an abundance of nonperforming operations, which should have been discarded earlier, is a factor contributing to the lackluster performance of many Japanese companies. However, whether slow selection is a manifestation of this alleged managerial deficiency or an outcome of rational economic processes has not been studied very carefully. Firms can be slow in selecting out operations for economically legitimate as well as illegitimate reasons. For instance, knowing an operation’s long-term viability with a high degree of confidence may take time if considerable uncertainty exists with regard to external and internal conditions impacting its performance. Whether in fact Japanese firms are exceedingly inert in reaching a divestment decision, or, more generally, sub-optimally commit themselves to underperforming businesses, is an important yet undervisited question. This study contributes to filling this gap by studying the growth and exit behavior of Japanese firms in the United States. Though the problem of excessive commitment might be particularly ubiquitous in the management of Japanese firms, the growing literature on the internal capital market suggests that it is not entirely endemic to them. It suggests that the problem commonly arises when “corporate socialism” (Scharfstein and Stein, 2000; Stein, 2001) is distorting a firm’s investment and divestment decisions. Firms are said to be socialistic when they are relocating capital internally to finance the growth and survival of failing operations at the cost of more successful ones. Studies, such as Lamont (1997), Lang et al. (1996), Shin and Stulz (1998), Rajan et al. (2000), show that US firms actively cross-subsidize operations in a seemingly inefficient way. Following these studies, this article investigates the growth and survival of US plants in relation to the financial condition prevailing at the Japanese parent to gain insight into the working of internal capital markets across the Pacific. An important caveat is that mere dependence of plant dynamics on parental financial conditions, though suggestive of subsidization, does not clearly point to the inefficiency implied by socialism. In fact, the internal capital market is efficiency increasing if it is used for “winner picking,” where capital is relocated for financing operations that have a solid growth prospect but will be underfunded if operated as a stand-alone unit (Williamson, 1975; Stein, 1997). Claiming inefficiency therefore demands to show otherwise: i.e. it is used more for supporting “losers” rather than prospective “winners.” An empirical challenge in this regard is to identify winning and losing plants, since current accounting practices do not require firms to disclose the performance of individual foreign operations. For this reason, in performing the task, I draw insights from models of efficient firm growth and exit pioneered by Jovanovic (1982). At the heart of these models is the assumption that firm success depends on idiosyncratic efficiency, which is ex ante uncertain. Firms grow and exit as they learn about the efficiency through post-entry operations. A key prediction from the models, which is strongly supported by empirical studies such as Evans (1987a, 1987b), Hall (1987), Dunne et al. (1989), is that the turnover of small firms (plants) is more turbulent than that of large firms (plants): i.e. small firms have a higher exit rate, but, conditional on survival, grow faster. This is because small firms are highly heterogeneous in terms of efficiency. They include both future winners (efficient firms), which grow fast, and losers (inefficient firms) destined to fail, while large firms mostly consist of proven winners. In a later section, I will show that these patterns also characterize the turnover of Japanese plants operating in the United States.

368

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

Also contributing to the greater turbulence of small firms is the fact that they typically face tight financial constraints, which make them vulnerable to even small shocks (Carpenter and Petersen, 2002; Cabral and Mata, 2002). When a business is a subpart of a larger organization, however, as in the case of foreign plants, the internal capital market potentially mitigates the constraint and moderates its dynamics, but differently depending on the efficiency of internal fund use. Specifically, efficient use of funds (winner picking) dictates that the growth and survival of young small plants are most heavily financed. Inefficiency (socialism), on the other hand, implies that plants having failed to grow large despite their long-time existence receive relatively greater financing because it aims to safeguard known losers from the imminent risk of failure. My empirical results reveal that internal capital markets importantly drive plant dynamics in that parental financial conditions significantly influence plant growth and survival. Specifically, ceteris paribus, plants owned by a firm whose financial condition is healthier and/or improving grow faster and are less likely to exit. This finding explains a seemingly puzzling observation that Japanese divestment in the United States increased in the late 1990s in the face of booming US economy. In that period, the financial health of Japanese firms deteriorated due to the prolonged stagnation of the domestic economy. It consumed internal funds that firms would have mobilized to support ailing US operations, leading to the wave of divestment. On the other hand, results provide no systematic evidence that firms discriminate across plants in providing supports based on future growth prospects, either in the sense of socialism or winner picking. In this sense, the internal capital markets of Japanese firms may be best characterized as egalitarian. The organization of this article is as follows. The next section reviews recent theoretical and empirical research on firm dynamics and internal capital markets to draw insights for empirical analysis. Section 3 introduces data and econometric methodology. Section 4 presents estimation results. The final section concludes.

2. Background 2.1. Growth and survival of stand-alone firms Firm growth and survival crucially depend on expectation. Economically rational firms would grow only if they anticipate that doing so creates greater profits than staying at the current size and exit only if the anticipated value of staying in business is less than the termination value. They grow (decline) as they adjust size to the anticipated optimal scale and exit when it is zero. Since the optimal firm size generally increases with a firm’s productive efficiency, firm growth and survival are importantly governed by the evolution of the firm’s expectation on its efficiency. When there is no uncertainty in this regard, firm growth and survival would be entirely determined by external business conditions such as product demand growth. More generally, however, empirical resolution of internal uncertainty importantly affects the dynamics. In an influential article, Jovanovic (1982) presented a model in which firms do not know their efficiency with full certainty. They, as a Bayesian learner, enter a market based on a prior belief on the efficiency and continuously update it through performance feedback

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

369

from post-entry operations. Firms who received favorable signals indicative of superior efficiency grow large and those who received unfavorable signals decline or cease to operate (exit). As they receive more feedback and the precision of expectation increases, firm dynamics become increasingly stabilized. This model, along with others developed by Lippman and Rumelt (1982), Cabral (1995), Hopenhayn (1992), and Pakes and Ericson (1998), effectively highlights the role of internal uncertainty as a driver of firm growth and exit. A key insight from these models is that the turnover of small firms is more volatile than that of large firms: i.e. small firms have a high rate of failure yet, conditional on survival, grow faster than large firms. This is because the distribution of productive efficiency is highly heterogeneous in small firms. Without financial frictions, which are empirically important as we will see shortly, small firms stay small either because they are not sure enough to upscale production or the past experience indicates that they are not efficient. The first group of firms includes future winners (efficient firms) who would eventually grow large, contributing to the faster growth, while the latter group is responsible for the greater incidence of exit. Evidence strongly supports the insight. Studies such as Evans (1987a, 1987b) and Hall (1987) show that, in a cross-section of US firms, conditional on age, firm growth and exit rates and the variance of growth rates decrease with firm size, and, conditional on size, they decrease with age.1 These tendencies are evident also in the turnover of a firm’s operating units, such as manufacturing plants, rather than the firm as a whole. Dunne et al. (1988, 1989) provide evidence from US establishment data. A recent study by Blonigen and Tomlin (2001), which closely parallels the present study, supplies evidence on the growth of US plants operated by Japanese firms. The recent literature also suggests that the greater turbulence of small firms is also attributable to financial constraints. Following Fazzari et al. (1988), a number of studies have shown that the availability of internally generated funds impacts real business activities (see Hubbard, 1998 for a review). Carpenter and Petersen (2002) point out that the limited access to external financing is particularly binding to the growth of small firms. Their evidence indicates that small firm growth (in total assets) is almost entirely financed by internal cash flows. Cabral and Mata (2002) revealed that financial constraints play a key role in the evolution of firm size distribution, which critically hinges on small firm growth and survival. From a theoretical standpoint, Cooley and Quadrini (2001) show that introducing financial frictions into the learning model of firm dynamics (e.g. Jovanovic, 1982) makes it more consistent with empirical regularities reported in the literature. 2.2. The role of internal capital markets When a business is financially constrained, it may fail to grow large or even survive however efficient it is due to the limited access to external financing. If it is a stand-alone firm, the story ends here. However, when it operates as a subpart of a larger organization, such as US plants of Japanese firms, the story has one more twist, since it potentially has 1 Sutton (1997) and Caves (1998) provide a comprehensive overview of empirical studies on firm turnover.

370

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

access to funds raised in other parts of the firm, if the headquarter supervising the internal fund use is willing to relocate capital across operations. Such relocation, known as the internal capital market, can be efficient when a firm’s operating units have varied growth prospects and ability to finance growth. Stein (1997) coined the efficient use of internal capital markets as “winner picking.” It involves transferring capital to promising operations, which would be underfunded if financed solely by their own funds. More generally, the efficiency of overall fund use is maximized when capital is allocated such that the marginal value of investment is equalized across operations. The economic value of internal capital markets was recognized also by earlier authors, including Williamson (1975), and management consultants such as the Boston Consulting Group. The recent discovery of the diversification discount, however, casts serious doubt on the efficiency of internal capital markets. Studies such as Lang and Stulz (1994) and Berger and Ofek (1995) report that diversified US firms trade at a substantial discount relative to specialized firms.2 These studies evoked a widely held suspicion that ill-disciplined managers of diversified firms allocate capital inefficiently across divisions.3 At the opposite end of winner picking on the efficiency spectrum is “corporate socialism” (Scharfstein and Stein, 2000; Stein, 2001). It arises when managers allocate funds excessively to unpromising operations at the cost of promising ones. Such relocation might be used to merely sustain failing operations, which efficiency considerations suggest should be discarded. Where socialism prevails, the wedge in the marginal value of investment among operations increases because of the internal capital market, causing a decline (discount) in firm value. Evidence indicates that firms actively use internal capital markets. For instance, Lamont (1997) shows that a non-oil division’s investment in petroleum companies is positively affected by oil prices, even that of divisions who use oil as an input. Lang et al. (1996) show that the firm’s overall leverage significantly reduces a division’s investment and growth, whether it is the core (largest) or non-core division. Shin and Stulz’s (1998) results indicate that a division’s investment is affected by cash flow in other divisions as well as its own. Furthermore, divisional investment patterns revealed by these and other studies, including Rajan et al. (2000), suggest that the internal fund use of diversified US firms is, on average, inefficient.4 2.3. Implications The internal capital market literature has important yet mixed implications for the behavior and performance of Japanese firms. On the one hand, it suggests that they may be 2 But see Campa and Kedia (2002) and Villalonga (2000) who suggest that the discount is an artifact rather than real. 3 Note that, as Rajan et al. (2000) pointed out, the agency problem at the top management supervising overall fund use does not necessarily lead to the allocative inefficiency. This is because executives interested in creating a large empire would reach their goal most effectively by allocating capital toward operations with good growth prospects, though the firm, as a whole, would still overinvest. For this reason, the theoretical literature focuses on the self-interested behavior of lower rank managers who supervise individual operations. See, for instance, Scharfstein and Stein (2000) and Rajan et al. (2000). Boot (1992) provides an agency model of underdivestment. 4 See, however, Whited (2001) and Maksimovic and Phillips (2002) for a criticism and results suggesting otherwise.

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

371

inert in discarding operations because their capital allocation is such that unprofitable operations are protected by financial supports from elsewhere. Since such supports are made only at the cost of more successful operations, the overall firm performance would decline. On the other hand, the winner picking strategy suggests that, even if supports are present, they are efficient if they go to prospective winners, improving firm performance, at least in the long run. An important question, therefore, is who, if any, gets funded. The case of US operations provides an interesting study ground for this issue. As Urata (1998) notes, the average profitability of Japanese firms’ US operations has been disappointedly low. Nonetheless, Japanese firms had been seemingly very reluctant to discard them until the late 1990s, which marked a significant increase in divestment activities. The recent increase in US divestment is somewhat puzzling because the late 1990s witnessed an improvement in these operations’ profitability thanks to the strong US economy.5 A possible explanation for this “puzzle” is that a number of operations that cannot stand on their own had been sustained by financial supports from the parent. The prolonged stagnation of Japan’s economy, however, eroded away the parent’s ability to continue supports, resulting in the wave of divestment. Assuming that the above scenario is correct, which itself is a matter of empirical question, in a cross-section of operations, the winner picking and socialism scenarios point to the same type of operations as least affected by the parent’s capacity to provide supports, but different types of operations as most affected. Specifically, least affected are those who are solid enough to sustain themselves. If one follows the learning model of firm dynamics, they are mostly likely plants having grown large because of superior efficiency. Most affected are plants whose viability is yet to be seen but can potentially grow profitably if funded adequately (winner picking scenario), or those whose inability to sustain themselves is already known to the parent with a high degree of certainty (socialism scenario). The former is most likely young small plants, while the latter is most likely old small plants having failed to grow large despite their long-time existence. Therefore, one would expect that the dependency of plant dynamics on parental financial conditions are differently moderated by plant size and age depending on the efficiency of internal capital markets. The growth and survival of foreign operations have attracted increasing attention in recent years. Blonigen and Tomlin’s (2001) study of the growth of Japanese US plants demonstrates that grow rates decline with plant size and age, consistent with the learning model. Feinberg and Phillips’ (2002) study of the growth of US firms’ foreign affiliates reports a negative effect of affiliate size and positive effect of affiliate efficiency. Their results also indicate that the internal capital market impacts the relative growth of affiliates belonging to the same parent. Mata and Portugal (2000) studied the survival of foreign affiliates and showed that mortality rate declines with affiliate age. To the best of my knowledge, however, no studies have examined the growth and survival of foreign operations simultaneously, even though they may be best viewed as integral parts of the same economic 5 According to MITI’s Basic Survey of Overseas Business Activities, the profitability of Japanese US operations

measured by the ratio of after-tax profits to sales was averaged at −0.47% in 1990–1994 and 0.93% in 1995– 1999. Nonetheless, according to Toyo Keizai statistics, the divestment of US operations more than doubled in the latter period. Of course, the parallel increase in divestment and profitability is partially attributable to the fact that the exit of low performers lifts the mean profitability of remaining ones.

372

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

process as implied by the learning model. In what follows, I take on this task based on a cross-sectional sample of US plants, paying particular attention to their dependence on parental financial conditions.

3. Data and methodology 3.1. Sample The sample consists of US manufacturing plants that were active in 1994 and owned by Japanese manufacturers. Affiliates not engaged in manufacturing activities, such as distribution and R&D subsidiaries and holding companies, were excluded to ensure the homogeneity of functions sample affiliates mainly perform. According to Toyo Keizai’s Kaigai Shinshutu Kigyo Soran (Japanese Overseas Investment), there were 767 such plants in 1994. Since the analysis below requires the financial information on the side of parents, plants owned by private firms were excluded from the sample. After further eliminating plants owned by firms who ceased to exist as an independently listed firm due to failure or M&A over the next five years and plants whose detailed profile is not available in the Toyo Keizai directory, I ended up with 510 plants owned by 380 parents.6 These plants constitute the sample for this study. For each plant in the sample, the following information was collected: industry (defined at the SIC 2-digit level), employment, entry mode (greenfield vs. acquisition), equity share held by the parent, and age (years after foundation) in 1994.7 Employment is the plant size variable in this study, and plant growth is measured by the growth in employment. Using employment in measuring size and growth is consistent with past studies of firm and plant dynamics, including Blonigen and Tomlin (2001) who examined US plant growth in the late 1980s. The growth and survival of sample plants over the next five years were tracked based on the Toyo Keizai directory. In 1999, 407 plants were still active, but 103 plants had ceased to exist by that year.8 For surviving plants, employment in 1999 was collected to calculate a five year growth in size (gr) defined as the difference in the log of employment, i.e. gr = ln(size1999 ) − ln(size1994 )

(1)

where size represents the number of employees. Table 1 summarizes the size and age distributions of sample plants. The mean age is nine years. Blonigen and Tomlin (2001) report that the mean logged age in their 1987 sample is 1.04. In the present sample, the corresponding figure is 2.04, indicating the overall aging of Japanese US plants in the 1990s. Analysis of variance (ANOVA) reveals slight 6 Some joint ventures have multiple Japanese parents. For these plants, I treat the firm owning the largest equity stake as the parent. 7 In the case of acquired plants, age was defined as years after acquisition. The Toyo Keizai directory does not contain information on the founding year of the acquired plants. 8 Sometimes, a plant ceased to exist merely because of its changing the name, but still actively operating in the same location. I did not treat such a plant as failed-one.

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

373

Table 1 Age and size distribution of sample plants (n = 510) Mean Age Size (employment) raw size relative to US establishment2 relative to US affiliate3 relative to Japanese parent4

Industry difference1

Quartile 25%

Median

75%

9

6

7

10

1.56*

507 8.55 0.75 0.11

58 0.87 0.08 0.01

154 2.40 0.22 0.04

400 6.59 0.56 0.12

1.00 1.14 1.06 0.50

1 F -statistics for the inter-industry difference in mean. 2 Ratio of plant size to the average size of US establishments in the same SIC 2-digit industry taken from 1992

Census of Manufacturers, US Census Bureau. 3 Ratio of plant size to the average size of US foreign affiliates in the same 2-digit industry taken from Foreign Direct Investment in the United States (revised 1994 estimates), US Department of Commerce. 4 Ratio of plant size relative to the Japanese parent. * Significant at the 10% level.

inter-industry difference in the age distribution of sample plants. For instance, plants in the transportation equipment industry (SIC37) are significantly younger, by about three years on average, than plants in other industries. The average employment is 507, which is substantially larger than the 180 reported in Blonigen and Tomlin (2001). The plant size distribution is highly skewed toward the right, with the median being less than a third of the mean. Table 1 also reports distributions of plant size relative to the average size of US establishments and foreign owned US affiliates operating in the same 2-digit industry and the size of Japanese parents. On average, Japanese plants in the United States are eight times as large as the average US establishment size in the same industry. However, when compared to the mean size of foreign-owned US affiliates, they are 25% smaller on average.9 Meanwhile, the average size of US plants relative to their Japanese parent is about 10%. The skewness of size distribution is evident also with these relative measures, with the mean exceeding the median by a large margin. The mean growth of surviving plants (gr) is 0.239. Table 2 takes a first-cut look at the effect of size and age on plant growth and survival. It divides plants into four categories according to a two-way classification scheme based on plant age and size. For instance, plants are classified as new if their age is less than or equal to the median age of sample plants. Small plants are those whose size relative to the parent is no greater than the sample median. Reported in each cell are within average growth and survival rate. The table clearly demonstrates that plant growth declines with plant age and size, with new small plants achieving the highest growth. It also shows that survival rates increase with plant size and, 9 Note that, since firm and plant size distributions are usually very skewed (Sutton, 1997), even if Japanese

plants follow the same size distribution of foreign-owned US plants, when compared to the mean, most Japanese plants appear smaller.

374

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

Table 2 Growth and survival by plant size and age New age
median

Old age
median

Small size
median

n = 157 mean gr = 0.46 survival rate = 0.73

n = 96 mean gr = 0.23 survival rate = 0.72

Large size > median

n = 123 mean gr = 0.16 survival rate = 0.85

n = 134 mean gr = 0.11 survival rate = 0.90

Note. Plant size is measured relative to the size of parent. Mean gr is based only on surviving plants. Survival rate is the ratio of plants who were still active in 1999.

albeit weakly, with age. These patterns conform to the empirical regularities reported by past research on firm and plant dynamics. 3.2. Econometric specification and variables To study the dynamics of sample plants in more depth, I estimate a system of plant growth and survival equations simultaneously based on the maximum likelihood (ML) method. ML joint estimation is used because, since the growth rate is observable only for surviving plants, analyzing it based only on a sample of surviving plants may introduce the sample selection bias (Heckman, 1979). Mansfield (1962) was the first to point out that estimating a growth model without attention to the selection (survivorship) bias runs the risk of spuriously detecting a negative effect of size (age) if low growth small (young) firms disappear with a higher probability than large (old) firms. Though later studies controlling for the bias found that the effect of omitting non-survivors is rather small, I estimate the following models jointly to minimize the potential hazard of spurious correlation. The models are specified as follows. Growth model:
ηk · SICk + β1 ln(age) + β2 ln(rsize) + β3 ln(exp) + β4 FC gr = k

+ β5 FC + δ · controls + u;

(2)

Survival model: Prob(srv = 1) = Φ(z + e > 0),
πk · SICk + γ1 ln(age) + γ2 ln(age)2 + γ3 ln(rsize) + γ4 ln(exp) z=

(3)

k

+ γ5 FC + γ6 FC + ω · controls, where srv is a dummy variable taking one if a plant survives through 1999 and zero otherwise; gr is observable only if srv = 1. The growth model is a linear regression with a normally distributed error term u, while the survival model is a conventional probit model with an error term e (Φ denotes the standard normal density function). The selection bias arises via the correlation in error terms. I denote the correlation as ρ.

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

375

The variable age measures plant age in 1994 and enters the models as a logarithm. Note that the squared log age appears in the survival model. Pakes and Ericson (1998) suggest that age–survival relationship can be non-monotonic in the initial stage of learning if a certain minimal time is required to process information received after entry in arriving at an exit decision. The term captures such non-linearity. The variable rsize measures plant size in 1994 relative to the parent’s employment in the same year. Plant size was rescaled by parent size to control for industry-specific factors affecting the optimal plant scale, such as economies of scale, and firm-specific factors, such as the parent’s endowment of valuable resources supporting foreign operations (Caves, 1971; Buckley and Casson, 1976). Results reported below are qualitatively invariant when raw plant size or other relative size measures in Table 2 are used instead. The variable exp represents the parent’s experience in the United States before the focal plant was established. Such experience, if any, may substitute for a plant’s own learning and stabilize its growth and survival. It is defined as years elapsed since the foundation of the parent’s first US subsidiary to the establishment of the focal plant. FC is a variable representing the parent’s financial condition in 1994 and FC is its subsequent change. {SICk } is a set of industry dummy variables defined at the SIC 2-digit level. They are included to control for industry-specific factors, such as demand growth and the intensity of competition, impacting plant growth and survival. The controls is a vector of plant and parent-level control variables. More detailed descriptions of FC and control variables follow. 3.2.1. Parental financial conditions (FC) When a plant is financially constrained, parents may provide financial supports in an effort to promote its growth and survival. Such supports can take many forms including transfer pricing, lower and/or delayed profit remittance, and the provision of internal debts. Regardless of the form they take, a parent’s ability to support plants is constrained by its own financial conditions, in particular, the availability of easy-to-mobilize funds whose usage is relatively unfixed. To gauge the existence and magnitude of such financial resources, I rely on the following balance sheet variables and use them interchangeably as an empirical representation of FC: Working capital (wk): current assets minus current liabilities divided by assets. Current ratio (cr): ratio of current assets to current liabilities. Quick ratio (qk): ratio of quick assets to current liabilities. Leverage (lev): short- and long-term liabilities relative to assets. Current and quick ratios are a standard measure of liquidity in the accounting literature, measuring the size of easy-to-liquidate assets relative to liabilities that must be paid off very soon. Quick ratio, also known as the acid test ratio, is more stringent in the measurement of liquidity than current ratio in that its numerator excludes relatively fixed assets such as inventories. Working capital is akin to the current ratio in the definition of liquidity. I standardize it by assets to adjust for firm size differential. Leverage measures the extent of debt dependence in corporate financing. Since, ceteris paribus, greater leverage increases a firm’s bankruptcy risk, more leveraged firms are expected to be less flexible in fund use. These variables were constructed from Japan Development Bank’s financial database.

376

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

In analyzing the effect of above financial variables on plant growth and survival, I consider the effects of their initial level in 1994 and subsequent change simultaneously. The change (FC) is measured as the difference between 1994 and 1998 (i.e. FC = FC1998 − FC1994 ), where FC is one of the four financial measures. The expected sign of FC and FC is positive for liquidity variables and negative for leverage, if plants are financially constrained and parents provide supports to mitigate the constraint. Otherwise, they should not affect plant growth and survival significantly. 3.2.2. Control variables (controls) Two-parent level variables are included to control the parent’s capacity to compete successfully in foreign markets. The first is R&D expenditure divided by sales (rd). The established literature on multinational firms, such as Caves (1971) and Buckley and Casson (1976), posits that a precondition for successful foreign market entry is the ownership of proprietary assets, such as superior technological capability and know how, which can be put to use in multiple locations at low cost. Consistently, empirical research on Japanese FDI, such as Kimura (1989), Kogut and Chang (1996), and Berry and Sakakibara (2002), shows that R&D investment significantly increases firm-level FDI. The second is the number of US plants owned by the parent (plants). The fact that a firm operates multiple plants in a foreign country is indicative of its superior capacity to compete in the market. In addition to the parent-level controls, two plant-level variables are included: aqs is a dummy variable that takes one if a plant was established via acquiring a local firm and zero for greenfield plants; owns is the share of equity held by the parent, measuring the extent of control the parent can exercise over a plant’s operations and decisions. Yamawaki (1997) and Mata and Portugal (2000) report that acquired and joint venture affiliates have a greater propensity to exit than fully owned affiliates established via greenfield FDI. In terms of growth, however, Blonigen and Tomlin (2001) revealed that acquired and joint venture plants do not differ much in their behavior from wholly owned greenfield plants. Ideally, in addition to the above variables, one would like to have a variable measuring plant performance, such as profit rate and Q ratio, in the growth and survival regressions. It is natural to expect that high-performing plants grow faster and thrive longer than low-performing ones for at least two reasons. First, high-performers are mostly likely economically viable plants, which can operate efficiently in a large scale. Second, they tend to face milder financial constraints because of their internal cash flows and ability to raise external funds in the local capital market.10 The second reason is noteworthy since it implies that omitting a plant performance variable runs the risk of obtaining a biased estimate for the importance of parental financial supports.11 10 According to MITI’s Basic Survey of Overseas Business Activities, in 1995, US manufacturing affiliates, on average, spent 595 million yen for capital investments, of which 400 million yen was financed by affiliate’s own funds. The borrowing from the parent, on the other hand, accounted for less than 10% of the expenditure. These figures should not be seen as undermining the importance of parental financial supports, since parents can manipulate affiliate cash flows through various channels of intra-firm transactions. 11 The bias’s direction is hard to predict since it depends on the efficiency of internal fund use, i.e. who benefits most from the financial support from the parent.

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

377

Unfortunately, however, as I noted in introduction, plant performance data is not publicly available. This is a major drawback to studying the dynamics of foreign operations. Nonetheless, I stress that the present research design of combining the models of learning dynamics and capital market imperfections mitigates the problem, at least partially. This is because, if plant learning is efficient in the sense of the Bayesian learning model, plant age and size reflect the history of performance a plant has achieved from its inception. Since plant performance is expected to be more and less persistent, a plant’s current size and age would contain the information on its present performance as well. Including plant age and size in regressions therefore reduces the omitted variable bias. Evidence indicates that the profitability of foreign affiliates is positively associated with their size and age. For instance, Lupo et al. (1978) report that the profitability of foreign plants operated by US multinationals increases with plant age and size. In a rare study using MITI’s proprietary microeconomic data, Sakakibara and Yamawaki (2000) show that the profitability of Japanese foreign affiliates increases with affiliate size. These studies suggest that plant size and age serve as a reasonable, if not perfect, surrogate for plant performance. Table 3 presents summary statistics for the dependent and independent variables. The vast majority (79%) of sample plants are majority-owned (63% are wholly owned by a single parent).12 The dominant mode of plant establishment is greenfield investment, with acquired plants accounting for less than a quarter of the sample. Parents typically operate only one or two plants in the United States, though the distribution of plants is highly skewed and a few firms own as many as nine plants. The ratio of plants owned by multi-

Table 3 Summary statistics of independent and dependent variables n

Mean

Standard deviation

510 407 510 510 510 510 510 510 510 510 510 510 510 510 510 510 510

0.80 0.24 2.04 −3.35 1.63 0.04 2.86 0.24 0.84 0.16 1.76 1.33 0.59 −0.02 −0.17 −0.17 −0.02

0.40 0.67 0.57 1.61 1.29 0.03 2.23 0.43 0.24 0.16 1.07 0.94 0.17 0.11 0.68 0.60 0.08

Quartile 25%

srv gr ln(age) ln(age) ln(exp) rd plants aqs owns wk cr qk lev wk cr qk lev

1 −0.02 1.79 −4.42 0 0.02 1 0 0.64 0.07 1.17 0.79 0.48 −0.07 −0.29 −0.26 −0.06

Median 1 0.18 1.95 −3.24 1.79 0.04 2 0 1 0.15 1.47 1.10 0.60 −0.02 −0.07 −0.08 −0.02

75% 1 0.47 2.30 −2.14 2.83 0.06 4 0 1 0.27 1.98 1.48 0.71 0.02 0.08 0.05 0.01

12 Removing minority joint ventures from the sample does not affect the estimation results reported below.

378

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

plant firms is about 65%. The table also shows that parents typically experienced a decrease in liquidity after 1994.

4. Estimation results 4.1. Main results Table 4 shows ML estimation results of the plant growth and survival models. Since I have four variables representing FC, results are reported for each variable used in turn. The estimated ρ is positive in all four specifications (i.e. the selection bias is positive, consistent with Mansfield, 1962), but hardly significant. The effect of plant size is significantly negative for growth and positive for survival, indicating that larger plants are less prone to exit but grow slower. The effect of age on growth is negative, though not as significant as the size effect. Meanwhile, the effect of age on plant survival is significantly non-linear: the survival rate initially declines for about five years but increases with plant aging after that period. The observed pattern is consistent with Pakes and Ericson (1998) who suggest that firms may require a certain minimum time to process information received after entry before reaching an exit decision. The literature has shown that the variance of growth rate, as well as its mean, decreases with firm size (Sutton, 1997), probably reflecting greater uncertainty confronting small firms. To see if such a pattern exists also in the growth of sample plants, the logged squared residual from the growth model was regressed on the log of rsize. When residuals from Specification (1) is used, the slope coefficient is estimated at −0.205 (p = 0.011). Thus, the negative dependence of growth variability on size is present also in this sample.13 Overall, size- and age-dependency patterns in the turnover of US plants are highly consistent with the learning model, implying that empirical resolution of uncertainty with regard to the optimal scale of production importantly drives the dynamics. The effect of parent’s previous experience in the United States (exp) is negative for growth and positive for survival, but is limited in magnitude and not significantly different from zero. The results weakly suggest that prior experiences held by the parent substitute for a plant’s own learning, but to a very limited extent. Concerning the control variables, none are significant in the growth regression. In the survival models, however, the effect of aqs is negative and marginally significant and that of owns is significantly positive. Thus, as revealed by past studies, acquired plants are more prone to exit than greenfield plants, and joint ventures, especially minority owned ones, have a greater exit rate than wholly owned plants. Turning to FC variables, one will find that the effects of working capital (wk) and its change (wk) on plant survival and growth are significantly positive. They indicate that plants belonging to a parent owning greater liquidity at the beginning and that having experienced a subsequent increase in liquidity grow faster and are less likely to disappear. 13 Results based on other specifications are virtually identical. I also tried other variables, including plant age, as regressions, but none of them were significantly related to the variance.

Table 4 Estimation results of the plant growth and survival models Specification FC measure Depend. variable ln(rsize)

ln(age)2

−0.116*** (−4.282) −0.107* (−1.719)

0.210*** (4.088) −1.417*** (−2.563) 0.429*** (2.958) 0.027 (0.395)

−0.035 (−1.167) Parental financial condition FC 0.574** 1.307*** (2.353) (2.588) 1.230** FC 0.617* (1.920) (1.972) Control variables rd 0.601 0.477 (0.463) (0.154) plants −0.008 0.021 (−0.487) (0.624) aqs −0.029 −0.280 (−0.334) (−1.619) owns −0.115 0.669** (−0.776) (2.254) constant −0.146 1.393** (−0.525) (2.117) P 0.172 (0.659) Log likelihood −600.3 Model χ 2 80.6*** n 510 ln(exp)

(2) Current ratio (cr) gr srv −0.122*** (−4.448) −0.106* (−1.716) −0.031 (−1.043) 0.130*** (3.016) 0.091 (1.440)

0.199*** (3.939) −1.419*** (−2.569) 0.429*** (2.961) 0.031 (0.455) 0.174 (1.636) 0.169 (1.078)

0.582 1.318 (0.451) (0.428) −0.006 0.015 (−0.322) (0.446) −0.034 −0.295* (−0.393) (−1.722) −0.117 0.649** (−0.783) (2.205) −0.289 1.325** (−0.999) (1.970) 0.143 (0.495) −601.1 84.81*** 510

−0.127*** (−4.635) −0.110* (−1.769) −0.032 (−1.105) 0.158*** (3.276) 0.110 (1.551)

0.196*** (3.905) −1.440*** (−2.604) 0.433*** (2.988) 0.031 (0.445) 0.155 (1.305) 0.156 (0.884)

0.601 1.686 (0.467) (0.553) −0.005 0.012 (−0.275) (0.359) −0.032 −0.291* (−0.365) (−1.703) −0.117 0.639** (−0.781) (2.175) −0.284 1.439** (−0.995) (2.167) 0.136 (0.462) −600.8 86.62*** 510

(4) Leverage (lev) gr

srv

−0.124*** (−4.630) −0.106* (−1.710) −0.036 (−1.205)

0.202*** (3.961) −1.417** (−2.555) 0.429*** (2.948) 0.010 (0.140)

−0.410* (−1.922) 0.069 (0.141)

−0.923* (−1.927) −2.753*** (−2.913)

0.718 0.770 (0.553) (0.248) −0.008 0.022 (−0.428) (0.626) −0.033 −0.271 (−0.384) (−1.556) −0.141 0.673** (−0.947) (2.250) 0.187 2.053*** (0.680) (2.938) 0.160 (0.661) −599.4 75.98*** 510

379

Note. All specifications include industry dummies. In parentheses are t -values. * Significant at the 10% level. ** Idem., 5%. *** Idem., 1%.

(3) Quick ratio (qk) gr srv

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

ln(age)

(1) Working capital (wk) gr srv

380

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

When current (cr) and quick (qk) ratios are used instead, results are weaker, but they tell the same story. The result based on leverage is also consistent. Plants belonging to more leveraged parents and those having increased reliance on debts grow slower and are less likely to survive (however, the effect of leverage change on growth is not significantly different from zero). In sum, these results indicate that, other things being equal, plants owned by firms endowed with rich internal funds grow faster and are more safeguarded from the risk of immediate failure than those held by firms facing a tight financial constraint. They suggest that Japanese plants in the United States are financially constrained non-trivially and their dynamics are conditioned by the availability of internal funds parents can put to use for their support. They also imply that the surge of US divestment in the late 1990s was induced, at least in part, by the deteriorating financial health on the side of parents due to the prolonged stagnation of Japan’s domestic economy. 4.2. Who gets funded? Given that internal capital markets significantly impact the fate of US plants, the next question is who benefits most from financial supports from the parent. In addressing this question, I employ a methodology similar to Shin and Stulz’s (1998), which uses industrylevel Tobin’s Q as a proxy for growth opportunities confronting a division in a diversified firm. To capture growth opportunities of individual plants, however, I rely on the two-way plant classification scheme introduced earlier rather than Q. Specifically, I create four plant category dummies and make them interact with FC and FC. For instance, newsmall takes one for plants whose age and size (relative to the parent) are no greater than the sample median, and zero otherwise. Similarly, oldsmall, newlarge, and oldlarge are dummy variables for old-small, new-large, and old-large plants, respectively (see Table 2 for the exact definition of each category). Admittedly, a more direct test of efficiency would be to use a plant performance variable instead of the above categorical variables. Given the lack of performance data, however, the present specification is a reasonable alternative to address the problem. Since the last subsection’s analysis revealed that the turnover of Japanese US plants is consistent with the learning model of firm dynamics à la Jovanovic (1982), categorizing plants based on size and age is expected to sort out plants according to their profitability, if not perfectly. As discussed in Section 2, the internal capital market has varied implications on plant dynamics depending on whether it is efficient or not. To recapitulate, efficiency in the sense of winner picking implies that supports target prospective winners who may be unable to realize their growth potential or even survive unless supported. They are most likely new small plants whose survival rate is relatively low but grow very fast if they survive. Therefore, one would expect, in the presence of a winner picking strategy, the growth and survival of new small plants to be most sensitive to the parent’s financial conditions because they are the ones who are most heavily supported. On the other hand, inefficiency in the sense of socialism implies that support goes to losing plants having failed to grow

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

381

Table 5 ML estimation results of the plant growth and survival models including interaction terms Specification FC measure FC ∗ newsmall FC ∗ oldsmall FC ∗ newlarge FC ∗ oldlarge FC ∗ newsmall FC ∗ oldsmall FC ∗ newlarge FC ∗ oldlarge Wald FC Wald FC FC ∗ newsmall FC ∗ oldsmall FC ∗ newlarge FC ∗ oldlarge FC ∗ newsmall FC ∗ oldsmall FC ∗ newlarge FC ∗ oldlarge Wald FC Wald FC

(1) wk

(2) cr

(3) qk

(4) lev

1.335*** (3.338) 0.464 (1.060) 0.385 (1.024) 0.393 (1.207) 0.380 (0.568) 1.816* (1.929) 0.675 (1.201) 0.663 (1.242) 5.42 [0.14] 1.68 [0.64]

Growth model 0.178*** 0.257*** (2.603) (3.060) 0.141** 0.180** (2.218) (2.346) 0.102 0.124 (1.596) (1.563) 0.058 0.065 (1.080) (1.060) 0.021 0.137 (0.178) (0.974) 0.380* 0.355 (1.805) (1.538) 0.115 0.097 (0.938) (0.603) 0.022 0.033 (0.226) (0.310) 2.80 [0.42] 4.35 [0.23] 2.91 [0.41] 1.65 [0.65]

−0.445* (−1.673) −0.723*** (−2.586) −0.408* (−1.763) −0.389 (−1.630) 1.846 (1.626) −0.006 (−0.005) −1.101 (−1.247) 0.330 (0.443) 3.20 [0.36] 4.36 [0.11]

1.379* (1.703) 0.903 (0.912) 0.649 (0.812) 1.483 (1.636) −0.674 (−0.527) 2.618 (1.221) 1.364 (0.966) 2.048** (2.199) 0.95 [0.81] 3.49 [0.32]

Survival model 0.243* 0.219 (1.821) (1.334) 0.225 0.226* (1.704) (1.397) 0.352** 0.469** (2.118) (2.223) 0.543** 0.430** (2.495) (2.429) −0.555** −0.772** (−2.023) (−2.197) 0.767** 0.873** (2.012) (2.139) 0.673** 0.973** (2.198) (2.263) 0.717** 0.879** (2.423) (2.380) 1.54 [0.67] 2.69 [0.44] 15.7 [0.00] 16.0 [0.00]

−0.893 (−1.627) −1.483*** (−2.569) −0.584 (−1.042) −0.944 (−1.632) −2.146 (−1.039) −1.592 (−0.858) −4.280* (−1.813) −2.817* (−1.762) 3.22 [0.36] 0.85 [0.84]

Note. In parentheses are t -values. Wald FC and Wald FC report the Wald test statistics of parameter homogeneity (χ 2 (df = 3)) for interaction terms involving FC and FC, respectively. In brackets are p-values. * Significant at the 10% level. ** Idem., 5%. *** Idem., 1%.

382

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

large, making the growth and survival of old small plants most dependent on the parent’s financial conditions.14 Table 5 presents estimation results. I report only the coefficients of interaction terms and relevant test statistics since other parts of the results do not differ much from the baseline results in Table 4. Results are mixed and defy easy interpretation. For instance, in the specifications using liquidity variables, the growth of new small plants is most sensitive to the initial financial condition, suggestive of efficiency, though a Wald test does not reject the null of parameter homogeneity among four categories. The growth of old small plants, however, is most sensitive to changes in liquidity, implying socialism. When leverage is used in lieu, the growth of old small plants is most sensitive to the initial condition. The results for survival are too mixed to draw any meaningful interpretation. The null of homogeneity is rejected for the effect of liquidity change on survival in Specifications (2) and (3), but this is mainly due to the significantly negative effect for new small plants, which does not make any economic sense. Overall, perhaps, the only conclusion one can draw from Table 5 is that Japanese firms, as a whole, are not very systematic in providing supports to US operations. Neither socialism nor winner picking provides a good description of the nature of their internal fund use as far as the US operation is concerned. In this sense, egalitarian might be the best term to describe the nature of internal fund allocation in Japanese firms.

5. Concluding discussion This article highlights two important facts in the dynamics of foreign operations based on a sample of Japanese manufacturing plants in the United States. First, the growth and survival of foreign plants are closely related to plant size and age in a manner consistent with the learning model of firm dynamics. This implies that foreign market entry is an uncertain conduct whose success is hard to predict ex ante. As in the case of indigenous firms (plants), only efficient plants would thrive in the long run. However, knowing the level of efficiency unambiguously is harder for foreign plants, which operate in a distant unfamiliar environment. As such, the importance of experiential learning may be even greater in foreign operations than in domestic operations, as suggested by Blonigen and Tomlin (2001). Second, the growth and survival of foreign plants are significantly influenced by the financial condition prevailing at the parent. This suggests that they are financially constrained non-trivially, but able to seek supports from parents to mitigate the constraint through the internal capital market. The flip side is, when a plant is heavily dependent on parental supports, deterioration in the parent’s financial health threatens its very existence. This explains why US divestment by Japanese firms increased dramatically in the late 1990s despite the booming US economy. This finding parallels Klein et al. (2002) that US investment by Japanese firms allied with an ailing bank significantly decreased in the 14 Viewing old small plants as losing might be a little problematic, since they still grow faster than large plants

conditional on survival (see Table 2). However, their survival rate is low despite their long-time existence, implying that, on average, their economic viability is rather low.

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

383

early 1990s. They both suggest that the availability of internal and external funds exerts a large influence on the firm’s real foreign activities. Concerning the efficiency of the internal capital market, however, results of this study are indecisive. If anything, they seem to suggest that, when providing supports, parents do not discriminate across plants based on the future growth prospect very systematically. Neither winner picking nor socialism provides a good description of the internal capital market in Japanese firms as far as the US operation is concerned. This result resembles Shin and Stulz’s (1998) in that divisional investment in diversified US firms is not responsive to the growth prospects of individual divisions as reflected in industry-level Q. Based on this finding, they contend that the internal capital market of US firms is inefficient, since efficiency considerations demand firms to discriminate such that divisions with a good growth prospect receive disproportionately large funding. In this sense, though not too socialistic, Japanese firms might be also deemed inefficient in selecting out foreign operations. A quick caveat is in order, however. Whited (2001) and Maksimovic and Phillips (2002) suggest that industry-level Q is a poor proxy for a division’s growth opportunities. They show that, when a more adequate measure is used (Maksimovic and Phillips, 2002) or the measurement error is accounted for (Whited, 2001), divisional investment patterns of US firms appear efficient. A similar criticism may apply to the present study. Even though plant age and size are important determinants of the growth and survival chance plants can, on average, expect, they are admittedly crude proxies for plant-specific factors affecting future growth and survival. More research is necessary before reaching a definitive conclusion. This and other limitations notwithstanding, this study visited an important yet understudied question and points to interesting extensions future research might take. For instance, though this study looked at only supports going from the parent to foreign operations, the reverse flow of financial resources is of course feasible. In fact, recent anecdotal evidence on Japanese firms suggests that firms enjoying superior performances in large foreign markets, such as the United Sates, enjoy greater strategic flexibility in the domestic market. Studying the mutual dependence of domestic and foreign operations would increase our understanding of the behavior of multinational corporations and their impact on market competition. Furthermore, one can increase the scope of this study to encompass operations in multiple national markets. In a study based on a large sample of US foreign affiliates, Feinberg and Phillips (2002) suggest that there is a tradeoff in the growth of affiliates owned by the same firm yet located in different places because of the resource constraint the parent and affiliates collectively face. The international operation of Japanese firms grew substantially in the last two decades. It provides an interesting alternative avenue for studying this intriguing issue. Acknowledgments I am grateful to Marvin Lieberman, Olav Sorenson, Mariko Sakakibara, Takeo Hoshi (editor), and an anonymous referee for their helpful comments. Remaining errors are mine. Financial supports from the doctoral program and Center for International Business Education and Research at Anderson Graduate School of Management at UCLA are highly appreciated.

384

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

References Berger, P.G., Ofek, E., 1995. Diversification’s effect on firm value. J. Finan. Econ. 37, 39–65. Berry, H., Sakakibara, M., 2002. Resource accumulation and overseas expansion by Japanese multinationals. Working paper. UCLA. Blonigen, B.A., Tomlin, K., 2001. Size and growth of Japanese plants in the United States. Int. J. Ind. Organ. 19, 931–952. Boot, A.W., 1992. Why hang on to losers? Divestitures and takeovers. J. Finance 47, 1401–1423. Buckley, P.J., Casson, M., 1976. The Future of the Multinational Enterprises. Macmillan, London. Cabral, L., 1995. Sunk costs, firm size, and firm growth. J. Ind. Econ. 43, 161–172. Cabral, L., Mata, J., 2002. On the evolution of the firm size distribution: facts and theory. CEPR discussion paper. Campa, J.M., Kedia, S., 2002. Explaining the diversification discount. J. Finance 57, 1731–1762. Carpenter, R.E., Petersen, B.C., 2002. Is the growth of small firms constrained by internal finance? Rev. Econ. Statist. 84, 298–309. Caves, R., 1971. International corporations: The industrial economics of foreign investment. Economica 38, 1–27. Caves, R., 1998. Industrial organization and new findings on the turnover and mobility of firms. J. Econ. Lit. 36, 1947–1982. Cooley, T.F., Quadrini, V., 2001. Financial market and firm dynamics. Amer. Econ. Rev. 91, 1286–1310. Dunne, T., Roberts, M.J., Samuelson, L., 1988. Patterns of firm entry and exit in US manufacturing industries. RAND J. Econ. 19, 495–515. Dunne, T., Roberts, M.J., Samuelson, L., 1989. The growth and failure of US manufacturing plants. Quart. J. Econ. 104, 671–698. Evans, D., 1987a. Tests of alternative theories of firm growth. J. Polit. Economy 95, 657–674. Evans, D., 1987b. The relationship between firm growth, size, and age: estimates for 100 manufacturing industries. J. Ind. Econ. 35, 567–581. Fazzari, S.M., Hubbard, R.G., Petersen, B.C., 1988. Financing constraints and corporate investment. Brookings Pap. Econ. Act. 1, 141–195. Feinberg, S., Phillips, G., 2002. Firm-specific resources, financial market development and the growth of US multinationals. NBER working paper 9252. Hall, B., 1987. The relationship between firm size and firm growth in the manufacturing sector. J. Ind. Econ. 35, 115–140. Heckman, J.J., 1979. Sample selection bias as a specification error. Econometrica 47, 153–161. Hopenhayn, H., 1992. Entry, exit, and firm dynamics in long run equilibrium. Econometrica 60, 1127–1150. Hubbard, R.G., 1998. Capital market imperfections and investment. J. Econ. Lit. 36, 193–225. Jovanovic, B., 1982. Selection and evolution of industry. Econometrica 50, 649–670. Kimura, Y., 1989. Firm-specific strategic advantages and foreign direct investment behavior of firms: the case of Japanese semiconductor firms. J. Int. Bus. Stud. 20, 296–314. Klein, M.W., Peek, J., Rosengren, E.S., 2002. Troubled banks, impaired foreign direct investment: The role of relative access to credit. Amer. Econ. Rev. 92, 664–682. Kogut, B., Chang, S.J., 1996. Platform investments and volatile exchange rates: Direct investment in the US by Japanese electronic companies. Rev. Econ. Statist. 78, 221–231. Lamont, O., 1997. Cash flow and investment: evidence from internal capital markets. J. Finance 52, 83–110. Lang, L., Stulz, R., 1994. Tobin’s q, corporate diversification, and firm performance. J. Polit. Econ. 102, 1248– 1280. Lang, L., Ofek, E., Stulz, R., 1996. Leverage, investment, and firm growth. J. Finan. Econ. 40, 3–30. Lippman, S.A., Rumelt, R.P., 1982. Uncertain imitability: An analysis of interfirm differences in efficiency under competition. Bell J. Econ. 13, 418–438. Lupo, L.A., Gilbert, A., Liliestedt, M., 1978. The relationship between age and rate of return of foreign manufacturing affiliates of US manufacturing parent companies. Surv. Curr. Bus. 58 (8), 60–66. Maksimovic, V., Phillips, G., 2002. Do conglomerate firms allocate resources inefficiently across industries? Theory and evidence. J. Finance 57, 721–767. Mansfield, E., 1962. Entry, Gibrat’s law, innovation, and the growth of firms. Amer. Econ. Rev. 52, 1023–1051. Mata, J., Portugal, P., 2000. Closure and divestiture by foreign entrants: the impact of entry and post-entry strategies. Strategic Manage. J. 21, 549–562.

T. Ushijima / J. Japanese Int. Economies 19 (2005) 366–385

385

Pakes, A., Ericson, R., 1998. Empirical implications of alternative models of firm dynamics. J. Econ. Theory 79, 1–45. Rajan, R., Servaes, H., Zingales, L., 2000. The cost of diversity: The diversification discount and inefficient investment. J. Finance 55, 35–80. Sakakibara, M., Yamawaki, H., 2000. What determines the profitability of foreign direct investment? A subsidiary-level analysis of Japanese multinationals. In: Academy of Management Proceedings 2000, vols. 1– 6. Scharfstein, D.S., Stein, J.C., 2000. The dark side of internal capital markets: Divisional rent-seeking and inefficient investment. J. Finance 55, 2537–2564. Shin, H.H., Stulz, R.M., 1998. Are internal capital markets efficient? Quart. J. Econ. 113, 531–552. Stein, J.C., 1997. Internal capital markets and the competition for corporate resources. J. Finance 52, 111–134. Stein, J.C., 2001. Agency, information and corporate investment. NBER working paper 8342. Sutton, J., 1997. Gibrat’s legacy. J. Econ. Lit. 35, 40–59. Urata, S., 1998. Explaining the poor performance of Japanese direct investment in the United States. Japan World Economy 10, 49–62. Villalonga, B., 2000. Does diversification cause the “diversification” discount? Working paper. UCLA. Whited, T.M., 2001. Is it inefficient investment that causes the diversification discount? J. Finance 56, 1667–1691. Williamson, O.E., 1975. Markets and Hierarchies. Free Press, New York. Yamawaki, H., 1997. Exit of Japanese multinationals in US and European manufacturing industries. In: Buckley, P.J., Mucchielli, J.L. (Eds.), Multinational Firms and International Relocation. Edward Elgar, Brookfield, VT, pp. 220–237.