Diversification patterns and performance of large established Japanese firms

Tatsuo Ushijima* Aoyama Gakuin University Graduate School of International Management Shibuya 4-4-25, Shibuya-ku Tokyo 150-8366 JAPAN Yoshitaka Fukui Aoyama Gakuin University Graduate School of International Management Shibuya 4-4-25, Shibuya-ku Tokyo 150-8366 JAPAN * Corresponding author Tel: +81-3-3409-8544; Fax: +81-3-3409-4167 E-mail: ushijima@gsim.aoyama.ac.jp

This version: December 11, 2004

Acknowledgement: We would like to thank seminar participants at the University of Tokyo and the 22nd Nikkei conference on firm behavior for their helpful comments. Remaining errors are ours. Financial supports from the Graduate School of International Management at Aoyama Gakuin University are greatly appreciated.

ABSTRACT This article examines the industry diversification of the largest Japanese manufacturers in 1973-98. Results show that 118 sample firms steadily increased diversification, a trend continued from earlier periods. Nevertheless, the relatedness of their constituent businesses gauged based on the Input-Output table remained high and stable throughout the study period. Econometric analysis reveals that firms pursuing the “constrained diversification” exploiting inter-business links centered on the core industry segment tend to achieve a higher profitability than firms engaged in the “linked diversification” exploiting links distant from the core. JEL classification: L23; L25; L29 Keywords: Diversification; Industry relatedness; Coherence; Japanese firm

1. Introduction In the last decade, corporate diversification across industries has attracted a great deal of attention by economists and management scholars. Research on the diversification discount, such as Berger and Ofek (1995) and Lang and Stulz (1994), shows that diversified U.S. firms trade at a substantial discount relative to specialized firms. Active debate is still

going on as to whether diversification itself is responsible for the discount (Martin and Sayrak, 2003). Recent studies, such as Campa and Kedia (2002) and Villalonga (2004a),

suggest that, once the endogeneity of diversification status is accounted for, diversification is not detrimental to firm value. Nevertheless, the fact that a staggering number of U.S. firms

have divested unrelated businesses to refocus on the areas of their core strength indicates that managing a large, complex, diversified firm is not an easy task.1 The wave of corporate refocusing hit the other side of the Pacific in the late 1990s. Nowadays, the business press is full of articles reporting Japanese companies, even those generally considered the bluest of blue, actively slashing unprofitable businesses to increase their industry focus. This is rather puzzling in light of the fact that the diversification For instance, in

strategy of Japanese firms has been long espoused as economically sound.

an article having exerted a powerful influence on today’s management thinking, Prahalad and Hamel (1990) draw heavily on Japanese companies, such as Honda, Canon, and NEC, to illustrate the importance of nurturing the firm’s “core competence” to sustain growth without risking corporate health. Why then the huge wave of refocusing? What went wrong? Unfortunately,

researchers do not have a ready answer to these questions due to the paucity of research on the recent fact about corporate diversification in Japan. In this article, we want to fill the

1

The trend of corporate refocusing in the United States is reported in Comment and Jarrell (1995) and Markides (1995) among others.

1

gap by studying the diversification behavior and performance of large established Japanese manufacturers in 1973-98. Our study’s important predecessor is Yoshihara et al. (1981) that We take the same firms as our

studied 118 firms in great depth for the period of 1958-1973. sample and study their diversification in a more recent period.

The goal of this article, however, is not to replicate Yoshihara et al. (1981) with a different study period. They cover a number of issues we do not address in this study, such as the strategy-organization fit, which occupies a large place in the diversification literature (Chandler, 1962; Williamson, 1975). Rather, we focus our attention to the evolution of the

firm’s diversification posture, especially the relatedness of its constituent businesses. It is often argued that, for diversification to create value, the firm must keep the relatedness (coherence) of its businesses high. Shleifer and Vishny (1991) contend that the root cause of U.S. corporate refocusing in the 1980s is a decline in firm coherence brought about by the conglomerate merger wave in the 1960s and 70s. In this study, we devise multiple measures of relatedness based on the Input-Output (IO) table as in Lemelin (1982) and Fan and Lang (2000) and study their development in the long-run and impact on firm performance. Our results show that large established Japanese manufacturers increased their industry diversification in 1973-98, a trend continued from earlier periods. Despite this increase, however, the relatedness of their businesses remained high and virtually constant, supplying no evidence for a widespread decline in firm coherence. Regression results show

that firms pursuing the “linked diversification” exploiting inter-business links distant from the core industry segment tend to be less profitable than firms engaged in the “constrained diversification” exploiting relatedness links centered on the core. This result suggests that

tightly packing businesses around a firm’s area of core strength increases economic gains from diversification. However, results show that Japanese firms on average did not increase

the extent of linked diversification. We thus conclude that the recent refocusing activities of
2

Japanese firms are hard to explain purely from a coherence perspective, which seems to have guided the restructuring of U.S. firms importantly. The plan of this article is as follows. The next section briefly reviews the

economics and management literature of diversification and evidence on Japanese firms. The third section introduces data and methodology. The fourth section presents key findings.

The fifth section studies the effect of diversification on firm profitability and growth with a formal econometric analysis. The final section concludes.

2. Background 2.1. General overview Operating in multiple industries entails extra costs on the firm, which would be unnecessary when it is specialized in a single industry. The more heterogeneous a firm’s

activities become, the more overloaded its managers’ information processing capacity, hampering the quality and timeliness of their decisions. Though turning to a decentralized,

divisional form of organization (M-form organization) might mitigate the problem (Williamson, 1975), delegating authorities to division managers has its own problems. Entrenched managers might routinely engage in rent-seeking activities to increase their private benefits, wasting scarce firm resources (Milgrom and Roberts, 1988). It might be

difficult to maximize firm performance by designing and installing an incentive scheme, which aligns the varied interests of organizational subparts (Rotemberg and Saloner, 1994). Roberts (2004) provides a number of reasons to suspect that large complex economic organizations like diversified firms might underperform small rivals organized more simply. Of course, diversification has benefits as well as costs. Of the possible gains from

diversification, Montgomery (1994) points to the increased utilization of firm resources as the most important. Penrose (1959) posits that, as the firm grows in extant businesses, it
3

accumulates slack resources that can be used elsewhere at low cost.

Her view is consistent

with Panzar and Willig’s (1981) theory of economies of scope showing that the existence of a multi-product firm implies the existence of “quasi-public” inputs that can be shared across products. Such inputs might be physical (e.g. production facilities), but more likely intangible, such as technological know how, well-recognized brand name, and management practices (Teece, 1980). If the joint use of these resources confers a competitive advantage on the firm’s constituent businesses by lowering cost or increasing customer’s willingness to pay, diversification creates value as long as the gain is sufficiently large to offset the cost of organizational complexity. The possibility of resource sharing, however, is not limitless. Honda’s expertise in the internal combustion engine technology, historically nurtured in the motorcycle business, is potentially deployable in multiple industries, especially automobiles. Thus, the company entered the automobile industry very successfully, but, to the best of our knowledge, it never considered entering the semiconductor industry simply because the technological expertise required there was vastly different from Honda’s core strength. As this straightforward example illustrates, the effectiveness of resource sharing increases with the similarity (relatedness) of industries a firm has in its business portfolio. Many conglomerates failed to

create value (and were eventually dissolved) precisely because their business mix was such that there was little room for resource sharing and, hence, synergies (Shleifer and Vishny, 1991). It is commonly believed that a high degree of inter-business relatedness is critical for the performance of diversified firms. Rumelt’s (1974) pioneering study shows that the

profitability of unrelated diversifiers (i.e. conglomerates) is significantly lower than that of

4

firms pursuing related diversification.2

Studies, such as Wernerfelt and Montgomery (1988),

Berger and Ogek (1995), and Fan and Lang (2000), report that Tobin’s Q of diversified U.S. firms is increasing in the relatedness of their businesses. Villalonga (2004b) contends that

the diversification discount widely reported in the finance literature should be seen as a discount for unrelated diversification. This is because companies tend to tack highly related Consequently, when diversification is measured

businesses into a single reporting segment.

based on financial reporting data, one underestimates the extent of related diversification. She shows that, when this bias is corrected for, diversified firms enjoy a valuation premium.

2.2 Evidence on Japanese firms Evidence on the diversification of Japanese firms is limited in supply.3 documents the long-term trend in the diversification of Japanese firms. Goto (1981)

Based on

disaggregated sales data of 124 largest industrial firms, he reports that the average extent of diversification measured by the Herfindahl index steadily increased during 1963-75. The same trend is reported in Yoshihara et al. (1981) studying 118 firms for a closely matched period, a result which is unsurprising given that their sample is essentially identical to that of Goto (1981). Based on a similar sample of large manufacturing firms, Yasuki (1995) shows

that the trend of increasing diversification continued through 1990. With regard to more qualitative aspects of diversification, such as relatedness, systematic evidence is even more limited. as a milestone. In this respect, Yoshihara et al. (1981) stands out

Following a scheme originated by Rumelt (1974), they categorized the

sample firms’ diversification strategy based on a combination of objective and subjective criteria.
2

They report a number of interesting findings.

For instance, they demonstrate that,

There is a related stream of research studying the effect of the extent of diversification on firm profitability. Results tend to show that the effect is neutral or weakly negative (Montgomery, 1994). 3 See Itoh (2003) for a more detailed review of available evidence.

5

despite the general increase in diversification, firms pursuing unrelated diversification are limited in number and not increasing. firms. This is in contrast to Rumelt’s (1974) finding for U.S.

They also show that, in firms engaged in related diversification, those pursuing the

linked diversification as opposed to the constrained diversification are increasing. In essence, the constrained diversification refers to the case where the firm diversifies into industries that are all related through its core business serving as the main provider of sharable resources. Honda’s diversification from motorcycles into automobiles On the other

and other products, such marine engines and generators, is a prime example.

hand, in the linked diversification, a firm increases its industry scope sequentially, leveraging resources acquired through the process of diversification, not necessarily those the firm had in the core at the onset of diversification. Canon is a good example.4 After establishing a

strong foothold in the camera industry, it entered the photocopier industry. The company then successfully diversified into laser printers by leveraging the expertise in electronics it acquired in the copier and other businesses outside its historical core (camera). illustrates the distinction of the two strategies. The finding that the linked diversification is increasing implies that firms are growing in areas, which are not unrelated one another but increasingly distant from their core area of strength. This strategy might be a double-edged sword. On the one hand, because Figure 1

the area of diversification is not confined to the vicinity of the core, it may promote fast growth. If executed successfully, it would enrich the firm’s repository of resources, Yoshihara et al. (1981) show that, conditional

increasing growth opportunities even more.

on the extent of diversification, firms pursuing the linked diversification grow faster than those engaged in the constrained diversification in terms of revenue. On the other hand, it
4

Yoshihara et al. (1981) classifies Canon as an unrelated diversifier. However, in a later article, Yoshihara (1993) cites the company as representing the linked diversification strategy. Parhalad and Hamel (1990) take a similar view.

6

might pose a formidable challenge on management because the intra-firm flow of resources is likely to be very complex as well as organizational tasks supporting it.5 Failures in one nod

of the link can easily translate into failures in other nods. Yoshihara et al. (1981) again show that, on average, the linked diversifiers perform poorly relative to the constrained diversifiers in terms of profitability. Rumelt (1974) reports the same pattern for U.S. firms. Yoshihara et al. (1981) predicted that, by the turn of the 20th century, firms pursuing the linked diversification would dominate the universe of large Japanese firms. This is an

intriguing prediction that may shed light on the cause of recent trembling of large diversified firms. Unfortunately, due to the paucity of research, it remains untested. In what follows,

we study the diversification of large manufacturing firms in 1973-98 with particular attention to the nature of their inter-business links.

3. Data and methodology Our research sample is the same 118 firms studied by Yoshihara et al. (1981). They

are basically the largest manufacturing firms in terms of capital and/or sales at the time of 1970. Though the time of sample selection is rather old, these firms still dominate Japan’s Performing the same

industrial landscape today (see Appendix for a list of sample firms).

sample selection procedure for a more recent period yields a set of firms that overlaps the present sample substantially. Studying these firms therefore allows us to develop a very

long-term view of their diversification strategy by ensuring continuity from a past leading study without sacrificing the representativeness of research sample (in the class of large influential manufacturing firms).

5

This implies that the corporate headquarters (CHQ) of linked diversifiers play more active planning and coordination roles than those of constrained diversifiers. Kagono et al. (1985) report that product divisions of Japanese diversified firms are generally less autonomous than U.S. counterparts. A recent survey by Collis et al. (2003) shows that the CHQ of Japanese firms are significantly larger than those of U.S. and European firms.

7

Following Yoshihara et al. (1981), we observe the sample firms’ diversification posture every five years. The first year of observation is 1973, which is the last year in their study period. The other years of observation are 1978, 83, 88, 93, and 98. In contrast to

earlier periods, these firms experienced a chronic decline in profitability in our study period. Figure 2 shows that the average ROA of sample firms deteriorated substantially in the last three decades. phenomenon. Our primary data source is the Yukashoken Hokokusho, Japanese equivalent of 10-K, supplying the breakdown of sales by product segments. Since firms adopt different policies We would like to know whether diversification is more or less related to this

in grouping products into a segment, a cross-sectional comparison of raw segment data may not be very meaningful. common standard. To facilitate comparisons, we must adjust segments according to a

For this task, we use the 4-digit classification of the 1985 IO table

containing 183 industries, of which 106 are in the manufacturing sector. Japanese IO table is updated every five years. our study period. SIC 4-digit. We adopt the 1985 version because 1985 is the middle year of

Ideally, one would like to use a finer industry classification, such as the

For many firms, however, uniquely matching their reporting segments to a In what follows, we use the term “industry

finer-level industry unambiguously is infeasible.

segment” to refer to the regrouped segment, not the original segment appearing in the Yukashoken Hokokusho. An important caveat is that our data is built on the unconsolidated (parent only) financial statement as in Yoshihara et al. (1981), implying that diversification through subsidiaries is set aside in the following analysis. This is due to the fact that Japanese legal and accounting practices had long centered on unconsolidated statement until recently. In our study period, many firms do not disclose the breakdown of consolidated sales by segment and, when they do, it is often too crude for our research purpose.
8

As such, all variables used

in this research are measured at the parent level, not the firm as a whole. We gauge the extent of a firm’s diversification by three measures: the number of IO 4-digit industry segments, Herfindahl index, and core sales ratio. is widely used to gauge diversification. size of segments. The simple segment count

It, however, does not pay attention to the relative

As Berry (1971) points out, this is a major drawback because the segment The Herfindahl index corrects for this

size distribution in a firm is usually highly skewed. deficiency. It is defined in the familiar fashion as:
H it =

j∈J it

∑S

2 ijt

,

(1)

where Jit is the set of firm i’s industry segments in year t and S is a segment’s share in total sales.6 The core sales ratio, commonly known as the specialization ratio, takes the firm’s

largest industry segment as its core and measures its share in total sales. Needless to say, the Herfindahl index and core sales ratio take one if a firm is specialized in a single industry and decrease as it diversifies. Concerning the measurement of relatedness, we adopt a new method. Rumelt

(1974) and Yoshihara et al. (1981) categorized firms into distinct strategy types based on the judgment of the extent and nature of relatedness among businesses. The basis of judgment

includes the similarities in product and process technologies and marketing activities. This approach has an advantage that one can take into account numerous factors, which are hard to quantify but potentially affecting relatedness. It also has drawbacks. For instance, changes

in the firm’s diversification are captured only as a large discrete jump from one strategy to another, though, in practice, they usually proceed gradually and continually. Also, it

6

Following Berry (1971), it is customary to use the complement of Herfindahl index (i.e. one minus Hit) rather than the index itself to gauge diversification. We break apart from this convention, but this distinction is of course nominal.

9

assumes away the inter-firm heterogeneity within a category. In this study, we adopt a different method to overcome these limitations by devising continuous measures of relatedness. The IO table recording the transactions among industries provides a natural basis for developing such measures.7 Lemelin (1982) is the first to use the IO table in quantifying relatedness. Denoting industry i’s purchase from industry k per the value of i’s output as aik, he proposes to measure the relatedness of industries i and j by ρij, the correlation coefficient of aik and ajk across all k.

ρ therefore measures the relatedness of an industry pair as the similarity of their input use.8
For a diversified firm having its core in industry c, we measure the relatedness of its non-core segments to the core (NCC) in a manner akin to Fan and Lang’s (2000):

NCC it = ∑ j ≠c

S ijt 1 − S ict

ρ cj .

(2)

The correlation coefficient ρ does not have the time subscript t because we use ρs computed from the 1985 IO table for all years. This is tantamount to assuming that ρ does not vary much over time, which we believe is not an unreasonable assumption.9 Because NCC measures the non-core segments’ collective relatedness to the core, it fits nicely with the concept of constrained diversification. However, it does not represent

the firm’s overall coherence because it does not pay attention to the relative weight of the core.
7

It is natural to posit that the firm is most coherent when its activities are confined to a

Yet another way to gauge relatedness is to use an industry classification system such as SIC. For instance, industries sharing the same 3-digit SIC code are classified as closely related and those only sharing the 2-digit code are regarded as weakly related. Many authors have questioned the validity of SIC-based relatedness measures. See, for instance, Fan and Lang (2000) and Robins and Wiersema (2002) for a critique. 8 Here, the “inputs” include not only goods and services supplied by upstream industries but also productive services provided by downstream firms, such as distributors and advertising agencies. 9 Fan and Lang (2000) computed a similar measure for U.S. firms using constant ρs based on a single year’s IO table and updated ρs computed from IO tables in multiple years. Their results show that the two methods return very similar value.

10

single industry (Sict=1). As such, we measure the firm’s total relatedness around the core (TRC) as follows:

TRCit = 1 = Sict + (1− Sict) NCCit

if

Sict =1 (specialized) (3)

otherwise (diversified).

TRC is set to one for specialized firms because ρcc=1.

These expressions are equivalent to

the share-weighted sum of ρs for all j including the core (i.e. TRC = Σj∈J Sjρcj). Our second relatedness measure recognizes the fact that, when a firm is pursuing the linked diversification strategy, a segment is not necessarily most closely related to the core. In this event, a more appropriate measure of non-core relatedness would be:

NCM it = ∑ j ≠c

Sijt 1 − Sict

max{ρ hj } . h≠ j

(4)

This variable takes a large value even if non-core segments are distant from the core but if they are closely related among themselves. An example will be illustrative. NCC of

Yamaha, Japan’s leading musical instrument producer, in 1998 is 0.160, rather a low value, because its fastest growing segment, electronic equipment parts (IO3431), is remotely located from the core, musical instruments (IO3919), in terms of ρ. However, its NCM in that year

is as high as 0.772 because the growing segment is closely related to the company’s another important segment, household electric appliances (IO3211) including audio equipment. Thus, depending on which measure to look at, one will have very different impressions about Yamaha’s diversification. In fact, NCC and NCM inform us of the nature of the company’s

diversification strategy most clearly when they are viewed jointly. Based on NCM, a firm’s coherence (total relatedness) is defined in a way analogous

11

to (3):

TRMit = 1 = Sict + (1− Sict) NCMit

if Sict =1 (specialized) otherwise (diversified). (5)

Note that, by construction, NCM (TRM) ≥ NCC (TRC).

Equality holds when a firm has no

more than two industry segments or it is engaged in the constrained diversification in the most strict sense (i.e. max{ρhj}=ρcj for all h excluding j).

4. Diversification patterns of large Japanese firms Table 1 reports the extent of sample firms’ diversification in 1973-98. The sample

size decreases over years due to the existence of firms acquired by or merged with another firm.10 firms. The top panel shows the number of IO 4-digit industry segments operated by sample The mean segment count was 3.6 in 1973. It increased steadily to 4.0 by 1993, but

declined after that year. The decrease is mainly due to the down-scoping by already highly diversified firms operating in more than five IO 4-digit industries. Meanwhile, the mean

Herfindahl index decreased from 0.57 in 1978 to 0.49 in 1998, showing a consistent trend toward greater diversification.11 The Herfindhal index of 0.49 roughly corresponds to the Contrasting this number to

case where a firm operates two equally sized industry segments.

the mean segment count of 3.9 in 1998 reveals the skewed nature of segment size distribution. Increasing diversification is evident also with the core sales ratio reported in the bottom panel. In 1973, a firm’s largest segment accounted for a 69% of total sales on average. By 1998,

10

There is also one firm omitted from the 1998 sample because it radically altered segment reporting to defy matching with a standard industry classification. 11 To facilitate a comparison with other studies, we also computed the Herfindahl index based on the 3-digit Japanese Standard Industrial Classification (JSIC). When JSIC is used in lieu, the index declined monotonically from 0.60 in 1973 to 0.53 in 1998. Our JSIC-based Herfindahl index is very close to that reported in Yasuki (1995) studying a similar set of firms.

12

the average share of the largest segments declined to 60%. Quartile statistics reveal that the trend of increasing diversification is pervasive in a wide spectrum of firms. Yoshihara et al.

(1981) and Goto (1981) report a steady increase in diversification for the same set of firms in the 1960s and early 70s. It is therefore safe to conclude that large Japanese manufacturing This is in

firms continued diversification for a very long time at least until the late 1990s.

contrast to the case of U.S. firms that generally decreased the extent of diversification in the 1980s by refocusing on the core (Comment and Jarrell, 1995; Markides, 1995). Table 2 reports four IO-based relatedness measures.12 The top panel shows the

non-core segments’ collective relatedness to the core (NCC) for diversified firms (i.e. firms operating in no less than two IO industries). The mean NCC in 1973 was 0.43. If a firm

with its core in industry c diversifies randomly into other industries, the expected value of NCC would be ρcj averaged over all j excluding c. evaluate observed NCCs. firms is 0.06. This value serves as a benchmark to

For 1973, the benchmark value averaged across 106 diversified

A cross-sectional t-test rejects the null of random diversification (i.e. NCC=

benchmark value) at the 0.1% significance level. Thus, we have evidence that, in deciding where to diversify and expand, firms are choosing target industries with attention to their relatedness to the core industry segment.13 Yet another important observation is that the average NCC remained virtually constant throughout the study period despite the general increase in diversification. This finding indicates that, consistent with Yoshihara et al. (1981), large manufacturing firms generally stayed away from unrelated diversification. The second panel shows the sample

firms’ total relatedness (TRC) as defined in (3) (in parentheses are figures only for diversified firms).
12

The average TRC declined slightly over years due to the decrease in core sales ratio

We also tried an alternative definition of the core segment in computing relatedness measures by setting the largest industry segment in 1973 as the core. Results are virtually identical to those reported in Table 2. 13 Teece et al. (1994) provide similar evidence for U.S. firms based on a different relatedness measure.

13

(Sict).

Nevertheless, because sample firms generally expanded in areas that are related to

their core, the magnitude of decline is modest. The third and bottom panels show the summary statistics of non-core segments’ relatedness (NCM) and firm coherence (TRM) as defined in (4) and (5), respectively. Two observations are noteworthy. First, very much like NCC, the average NCM stayed constant over the study period. This indicates that, in the sense of linked diversification as well, the sample firms’ diversification posture basically remained unchanged. average larger than NCC by about 0.1. Second, NCM is on

This implies that, for many firms, there is a group of

non-core segments that are more closely related among themselves than to the core segment. In other words, the element of linked diversification exists to some extent in many firms’ diversification strategy. We now examine how our IO-based relatedness measures relate to the conventional strategy classification based on more qualitative criteria by focusing on 1973, which Yoshihara et al. (1981) also studied. For 1973, they classify 76 firms as diversified firms, of which eight firms are unrelated diversifiers, 30 are constrained diversifiers, and 38 are linked diversifiers.14 The average NCC of unrelated diversifiers in 1973 is 0.27, which, based on a The average

t-test, is significantly lower than that of other diversified firms at the 5% level.

NCC (NCM) for firms engaged in the constrained and linked diversification is 0.56 (0.60) and 0.43 (0.57), respectively. Accordingly, the mean difference of NCM over NCC is 0.04 for constrained diversifiers and 0.14 for linked diversifiers. This inter-group difference is statistically significant at the 1% level. Therefore, in terms of our IO-based related measures, we can characterize unrelated diversifiers as firms with a low NCC.
14

Constrained diversifiers are those with a high NCC

The remaining 42 firms are specialized and vertically integrated firms according to their classification. They further divide the constrained and linked diversifiers into finer categories according to the relative weight of the core segment. We neglect this classification because NCC and NCM are adjusted for the core segment’s share.

14

and a NCM that is very close to NCC, implying that these firms’ non-core segments usually have the closet link with the core. Linked diversifiers are firms with a moderately high NCC and a NCM that is significantly greater than NCC. Our relatedness measures therefore

indicate that both constrained and linked diversifiers have a high degree of coherence in terms of NCM, but the structure through which their coherence is achieved is different in a manner consistent with Figure 1. We conclude that our relatedness measures are reasonably

successful in capturing the gist of strategy classification employed by earlier authors. Overall, the main findings of this section can be summarized as follows. First, large Japanese manufacturing firms continued diversification in 1973-98, though there is a sign that highly diversified firms started refocusing in the late 1990s. Second, despite the

increase in diversification, their diversification posture in terms of inter-business relatedness stayed essentially constant. For the sample firms as a whole, there is no evidence for a large

discontinuity in their diversification behavior. Their coherence was high in the early 1970s when they enjoyed superior performance in today’s standard and remained high until recently. This finding suggests that, if there is a problem with the Japanese firm’s diversification, it is not generally the lack of or decrease in coherence but something else. issue in the concluding section. We will return to this

5. Econometric analysis 5.1. Methodology This section explores the effect of diversification on firm profitability and growth. The last section’s analysis reveals that, in aggregate, the diversification behavior of sample firms stayed stable during our study period. However, there are of course large inter-firm We therefore study how the

differences in the extent and nature of industry diversification.

diversification posture of individual firms affects their performance in an econometric
15

framework. For this type of analysis, our observation window is relatively long, covering more than two decades. We, however, observe the sample firms’ diversification only at five year intervals. As such, we split the study period into five non-overlapping spells. A spell

begins with year t+1, where t = 1973, 78, 83, 88, and 93 and ends in t+5. The last spell therefore covers 1994-98. We then analyze how the performance in (t, t+5] is affected by

firm attributes including diversification in t as well as their contemporaneous change over the spell. In the analysis of profitability, our dependent variable is the simple average ROA during a spell, where ROA is defined as the ratio of operating income to total assets. In the analysis of firm growth, the dependent variable is the annual compound growth rate in assets over a spell. Firms that ceased to exist as an independently listed firm in the course of a spell due to M&A are omitted from estimations. Because our diversification measures are all sales-based, we also performed analysis using returns on sales and growth in sales as the dependent variable. reported). Research on the diversification-performance link suggests several pitfalls we should try to avoid. One is the effect of firm’s industry affiliation on diversification. It is known Results are qualitatively identical to those reported below (thus not

that firms in different industries have different propensities to diversify due to various industry traits (e.g. Lemelin, 1982; Montgomery and Hariharan, 1991). Montgomery (1985)

shows that such industry-level heterogeneity in diversification behavior obscures the effects of diversification and industry on performance. In addition, research on the diversification

discount, such as Campa and Kedia (2002) and Villalonga (2004a), demonstrates that treating diversification as exogenous in the analysis of firm value is misleading because there are factors affecting firm value and diversification simultaneously. that our analysis is immune to this endogeneity problem.
16

There is no reason to expect

To mitigate these and other problems, we employ the following estimation strategy. First, we adjust the dependent variables for the time and industry effects by subtracting the industry-time-specific mean. For instance, instead of a firm’s raw ROA in (t, t+5], we use

its difference from the mean ROA in the spell of firms operating in the same industry. This adjustment controls for the time as well as industry effect on performance. Removing the

time effect is important because Japan experienced a sharp decline in economic growth in our study period, which undoubtedly impacted the performance of sample firms. We do not

want diversification and other firm variables to pick up the effect of external shocks that are clearly out of control of individual firms. A critical question here is how to assign a firm to

a particular industry. This is obviously not straightforward because most sample firms are diversified across industries. An alternative would be to use a firm-specific imputed value, such as the segment-share-weighted sum of industry means, rather than an industry mean, a method popular in the corporate finance literature. This is infeasible for our study simply because necessary data (e.g. industry average ROA according to the IO 4-digit classification) is unavailable. Therefore, for the adjustment purpose, we define the industry rather broadly This classification

following the industry classification made by Yoshihara et al. (1981).

roughly corresponds to the 2-digit classification in a standard industry classification and contains 16 manufacturing industries. Second, we specify our regression model as follows and estimate it based on the fixed effects (FE) estimation method by pooling the five spells:

yi(t,t+5] = αi + β1 xit + β2 ∆xi[t,t+5] + γ1 dit + γ2 ∆di[t,t+5] + εi (t, t+5],

(6)

where y is the time- and industry-adjusted dependent variable, x is a vector of control variables, and d is a vector of diversification variables. ∆x and ∆d are the vector of changes
17

in x and d over a spell, respectively. ε is a disturbance term.

We make x relatively The variables in d vary

parsimonious and include the log of assets and firm leverage in it. by specification.

Table 3 provides a list of dependent and independent variables with their

summary statistics and descriptions. The firm-specific constant term α is included because research on the profit persistence shows that inter-firm differences in performance persist (Mueller, 1986). Based on a sample of large Japanese manufacturing firms, Maruyama and We expect Three

Odagiri (2002) show that “the persistent of profits persists” for a very long time.

that factors causing the performance persistence are not orthogonal to diversification.

prominent perspectives on diversification as summarized by Montgomery (1994), the market power, agency theoretic, and resource-based views, all imply that factors underlying α (i.e. market power/agency costs/idiosyncratic resources) also affect diversification. The FE specification is employed because of its robustness to the endogeneity problem.

5.2. Results Table 4 reports regression results for the profitability model. The effect of change

in firm size (logged assets) is positive and highly significant across all specifications, indicating that firms growing fast tend to achieve a high profit rate as well. leverage and its change are negatively associated with profitability. Meanwhile,

Turning to the effect of

diversification variables, Column (1) shows that the extent of diversification represented by the Herfindahl index is not strongly related to firm profitability. Herfindahl index change is significantly positive. therefore improves profitability. Nevertheless, the effect of

An increase in industry specialization

As Column (2) reveals, however, this positive effect of

industry focus becomes ambiguous when firm relatedness measured by TRC is introduced. In Column (2), the coefficients of TRC and its change are positive and significant, implying that firms concentrating on industries that are closely related to the core and firms increasing
18

coherence in this regard attain superior performance even if they are highly diversified. Interestingly, a different implication emerges from Column (3) where firm coherence is gauged more comprehensively by TRM. In this specification, coherence is negatively

associated with profitability, though the effect is only weakly and imprecisely estimated. Viewing Columns (2) and (3) together suggest that the constrained and linked diversification may have different implications for profitability. In Column (4), TRC and TRM enter the model jointly. The result lends strong

support to the differential impact of coherence mode: the effects of TRC and its change are positive and significant, while the effect of TRM is significantly negative. In reading this result, it is important to note that TRC and TRM do not always move independently because of the fact that TRM ≥ TRC. In particular, for firms engaged in the constrained

diversification in the strict sense, TRM always equals to TRC. In this case, the positive effect of a greater TRC is mitigated by the negative effect of accompanying TRM increase. Nevertheless, the result indicates that, for a given value of TRM, profitability is maximized when every non-core segment has the closest link with the core (TRM=TRC), and adjusting diversification toward this posture increases profitability. Our results therefore support the

finding by Rumelt (1974) and Yoshihara et al. (1981) that the linked diversifiers generally underperform constrained diversifiers in terms of profitability. making up of TRC and TRM enter the model separately. In Column (5), the variables

The Herfindahl index is omitted Though this

from this specification because of its high correlation with the core sales ratio.

specification is estimable only for diversified firms, the result strongly echoes Column (4), implying that an active pursuit of the linked diversification tends to penalize profitability. Table 5 reports regression results where the dependent variable is the growth rate of assets. In this set of estimations, the change in firm size is omitted from regressors because the dependent variable itself is firm growth. In every specification, the effect of firm size is
19

significantly negative, showing that larger firms grow slower.

This departure from Gibrat’s

Law is widely reported in the recent studies of firm growth reviewed by Sutton (1997) and Caves (1998). The effect of leverage is also significantly negative across all specifications, The negative effect

though the effect of its change is not significantly different from zero.

of leverage on firm growth is also reported in Lang, Ofek, and Stulz (1996) for U.S. firms. Turning to the effect of diversification, Column (6) suggests that firm growth is largely independent from the extent of industry focus. The other specifications in Table 5

show that firm growth is also largely independent from relatedness variables with one notable exception. In contrast to Yoshihara et al. (1981), the results reported in Columns (8)-(10) strongly suggest that an increase in the extent of linked diversification retards growth. Two explanations are conceivable for this result. First, because the difficulty of managing complex inter-business links in the linked diversification is such that increasing its extent restricts the firm’s ability to grow at least in the short run. Alternatively, it could be that it is firms facing low growth opportunities that actively exploit diversification opportunities distant from the core. In any event, we do not observe that linked diversifiers grow faster Our results rather suggest the opposite.

than constrained diversifiers.

Overall, this section’s analysis shows that the firm’s diversification posture in terms of inter-business links affects profitability. In particular, an active pursuit of linked

diversification strategy generally lowers profitability without the benefit of faster growth. This study therefore highlights the cost of linked diversification more clearly than Yoshihara et al. (1981). Their results imply that the linked diversification depresses profitability but In contrast, results The linked

the effect on profits is ambiguous because it also accelerates growth.

here showing a negative growth effect imply that profit itself is likely to decline.

diversification is an ambitious growth strategy exploring new resources in new places to feed further growth. Our results suggest, however, that successfully executing this strategy is in
20

practice difficult.

The cost of managing complex business links seems to easily outweigh its

benefits. This finding points to an important source of performance differentials among diversified firms. It also provides a partial explanation for the cutback in business scope

large Japanese manufacturing firms are currently undertaking, because the element of linked diversification exists in many firms’ diversification strategy to some extent. We note,

however, that, in aggregate, they have not increased the extent of linked diversification since the time when they prospered as we saw in the last section. The widespread restructuring involving the divestiture and spin-off of businesses appears difficult to explain purely from a coherence perspective.

6. Conclusion Large Japanese manufacturers experienced a severe performance decline in the 1990s. Because these firms operate in many industries, this article studied the long-term

development of their industry diversification with particular attention to inter-business relatedness. 1973-98. We found that the diversification of large manufacturers increased steadily in

Despite this increase, the coherence of their businesses stayed virtually the same

since the early 1970s when they enjoyed healthy growth and profitability. It is perhaps misleading to claim that large diversified manufacturers failed to sustain profitability because of their having expanded in areas that are too remote from their core strength. This observation, however, does not imply that industry diversification of Japanese firms is problem free. Arguably, a high degree of industry-level relatedness is a necessary

but not sufficient condition for the success of corporate diversification. If firms in the same industry diversify into different yet related industries in a herding fashion, these firms’ coherence as measured in this study remains high, but only firms endowed with a unique synergistic advantage would attain superior performance due to increased competition.
21

Porter and Sakakibara (2004) suggest that “me-too entry” is indeed a salient feature of the diversification behavior of large Japanese firms. Even if a firm has a unique potential for

synergies, the corporate socialism distorting the intra-firm allocation of resources might undermine their realization (Scharfstein and Stein, 2000). In this regard, Itoh (2003) presents an interesting scenario that a high degree of relatedness exacerbates rather than mitigates the allocative inefficiency. The link between the Japanese firm’s diversification

and performance clearly demands more research. In a broader context, this article showed that the distinction of constrained and linked diversification matters in studying the effect of firm coherence on performance. These two

concepts are not new in the management literature but have been under-visited by economists presumably because they are hard to quantify. We showed that, by extending IO-based relatedness measures introduced by Lemelin (1982) and Fan and Lang (2000), they can be accommodated to econometric research based on a large sample of firms. Their applications

to non-Japanese firms might yield fresh insights into their diversification as well.

22

Appendix: List of sample firms
Food products Snow Brand Milk Products Morinaga Milk Industry Meiji Dairies Nisshin Flour Milling Takara Shuzo Asahi Breweries Kirin Brewery Sapporo Breweries Nichirei Nippon Suisan Taiyo Fishery (Maruha) Nichiro Textile products Asahi Kasei Kuraray Teijin Mitsubishi Rayon Unitika Toray Industries Kanebo Toyobo Paper and pulp Jujo Paper Oji Paper Honshu Paper Daishowa Paper Manufacturing Sanyo-Kokusaku Pulp Chemicals Showa Denko Denki Kagaku Kogyo Mitsui Toatsu Chemicals Shin-Etsu Chemical Japan Synthetic Rubber Ube Industries Sumitomo Chemical Mitsubishi Chemical Kureha Chemical Industry Toyo Soda Manufacturing (Tosoh) Mitsui Chemicals Other chemicals Tanabe Seiyaku Shionogi Takeda Chemical Industries

Dainippon Ink and Chemicals Fuji Photo Films Shiseido Petroleum refinery Daikyo Oil (Cosmo Oil) Mitsubishi Oil Maruzen Oil Showa Shell Sekiyu Tonen Corporation Nippon Oil Rubber products Bridgestone Yokohama Rubber Stone, clay and glass products Nihon Cement Onoda Cement Sumitomo Cement (Sumitomo Osaka Cement) Asahi Glass Nippon Sheet Glass Iron and steel Kawasaki Steel NKK Nisshin Steel Kobe Steel Sumitomo Metals Nippon Steel Mitsubishi Steel Daido Steel Hitachi Metals Japan Steel Works Nonferrous metals Nippon Mining (Japan Energy) Mitsubishi Metal (Mitsubishi Materials) Mitsui Kinzoku Sumitomo Metal Mining Dowa Mining Nippon Light Metal Sumitomo Light Metal Sumitomo Electric Furukawa Electric Hitachi Cable Fujikura Electric Wire Machineries Niigata Engineering Kubota Komatsu

Daikin Industries NSK Koyo Seiki Sumitomo Heavy Industries Electric machineries Toshiba Mitsubishi Electric Hitachi Fuji Electric Fujitsu Oki Electric NEC Matsushita Electric Works Sony Sharp Sanyo Electric JVC Matsushita Electric Industrial Transportation equipment IHI Hitachi Zosen Mitsui Engineering and Shipbuilding Kawasaki Heavy Industries Mitsubishi Heavy Industries Automobiles Fuji Heavy Industries Mazda Motor Nissan Motor Hino Motors Daihatsu Motor Isuzu Motors Toyota Motor Honda Motor Suzuki Motor Precision instruments Shimadzu Canon Ricoh Other manufacturing Yamaha Dai Nippon Printing Toppan Printing

Note: Company names are as of 1973 in principle.

23

References: Berger, P. G., and Ofek, E. (1995). Diversification’s effect on firm value, Journal of Financial Economics 37, 39-65. Berry, C. H. (1971). Corporate growth and diversification, Journal of Law and Economics 14, 371-83. Campa, J. M., and Kedia, S. (2002). Explaining the diversification discount, Journal of Finance 57, 1731-62. Caves, R. (1998). Industrial organization and new findings on the turnover and mobility of firms, Journal of Economic Literature 36, 1947-1982. Chandler, A. D. Jr. (1962). “Strategy and Structure: Chapters in the History of the American Industrial Enterprise,” MIT Press, Cambridge, MA. Collis, D. J., Young, D., and Goold, M. (2003). The size, structure, and performance of corporate headquarters, Harvard Business School working paper. Comment, R., and Jarrell, G. A. (1995). Corporate focus and stock returns, Journal of Financial Economics 37, 67-87. Fan, J. P. H., and Lang, L. H. P. (2000). The measurement of relatedness: An application to corporate diversification, Journal of Business 73, 629-60. Goto, A. (1981). Statistical evidence on the diversification of large Japanese firms, Journal of Industrial Economics 29, 271-78. Itoh, H. (2003). Corporate restructuring in Japan Part I: Can M-form organization manage diverse businesses? Japanese Economic Review 54, 49-73. Kagono, T., Nonaka, I., Sakakibara, K., and Okumura, A. (1985) “Strategic vs. Evolutionary Management: A U.S.-Japan Comparison of Strategy and Organization,” North-Holland, Amsterdam.

24

Lang, L., Ofek, E., and Stulz, R. (1996). Financial Economics 40, 3-30.

Leverage, investment, and firm growth, Journal of

Lang, L., and Stulz, R. (1994). Tobin’s q, corporate diversification, and firm performance, Journal of Political Economy 102, 1248-1280. Lemelin, A. (1982). Relatedness in the patterns of industrial diversification, Review of Economics and Statistics 64, 646-57. Markides, C. C. (1995). “Diversification, refocusing, and economic performance,” MIT Press, Cambridge, MA. Martin, J. D., and Sayrak, A. (2003). Corporate diversification and shareholder value: A survey of recent literature, Journal of Corporate Finance 9, 37-57. Maruyama, N., and Odagiri, H. (2002). Does the ‘persistence of profits’ persist? A study of company profits in Japan, 1964-97, International Journal of Industrial Organization 20, 1513-33. Milgrom, P., and Roberts, J. (1988). An economic approach to influence activities in organizations, American Journal of Sociology 94, S154-79. Montgomery, C. A. (1985). Product-market diversification and market power, Academy of Management Journal 28, 789-98. Montgomery, C. A. (1994). Corporate diversification, Journal of Economic Perspectives 8, 163-78. Montgomery, C. A., and Hariharan, S. (1991). Diversified expansion in large established manufacturers, Journal of Economic Behavior and Organization 15, 71-89. Mueller, D. C. (1986). “Profits in the Long Run,” Cambridge University Press, Cambridge. Panzar, J. C., and Willig, R. D. (1981). Economies of scope, American Economic Review 71, 268-72. Penrose, E. T. (1959). “The Theory of the Growth of the Firm,” Basil Blackwell, Oxford.

25

Porter, M. E., and Sakakibara, M. (2004). Competition in Japan, Journal of Economic Perspectives 18, 27-50. Prahalad, C. K., and Hamel, G. (1990). The core competence of the corporation, Harvard Business Review 66 (May/June), 79-93. Roberts, J. (2004). “The Modern Firm: Organizational Design for Performance and Growth,” Oxford University Press, Oxford. Robins, J. A., and Wiersema, M. F. (2002). The measurement of corporate portfolio strategy: Analysis of the content validity of related diversification indexes, Strategic Management Journal 24, 39-59. Rotemberg, J., and Saloner, G. (1994). The benefits of narrow business strategies, American Economic Review 84, 1330-49. Rumelt, R. P. (1974). “Strategy, Structure, and Economic Performance,” Harvard Business School Press, Boston, MA. Scharfstein, D. S., and Stein J. C. (2000). The dark side of internal capital markets: Divisional rent-seeking and inefficient investment, Journal of Finance 55, 2537-64. Shleifer, A., and Vishny, R. W. (1991). Takeovers in the ‘60s and the ‘80s: Evidence and implications, Strategic Management Journal 12, 51-60. Sutton, J. (1997). Gibrat’s legacy, Journal of Economic Literature 35, 40-59. Teece, D. J. (1980). Economies of scope and the scope of the enterprise, Journal of Economic Behavior and Organization 1, 223-47. Teece, D. J., Rumelt, R., Dosi, G., and Winter, S. (1994). Understanding corporate coherence: Theory and evidence, Journal of Economic Behavior and Organization 23, 1-30. Villalonga, B. (2004a). Does diversification cause the “diversification discount”? Financial Management 33, 5-27.

26

Villalonga, B. (2004b). Diversification discount or premium? New evidence from the Business Information Tracking Series, Journal of Finance 59, 479-506. Wernerfelt, B., and Montgomery, C. A. (1988). Tobin’s q and the importance of focus in firm performance, American Economic Review 78, 246-50. Williamson, O. E. (1975). “Markets and Hierarchies,” Free Press, New York. Yasuki, H. (1995). “Big Business in the Contemporary Japan: Firm Behavior and Industrial Organization (in Japanese),” Nihon Hyoron Sha, Tokyo. Yoshihara, H. (1993). Diversification and dynamic synergy (in Japanese), in “The Japanese Enterprise System: Organization and Strategy (H. Itami, T. Kagono, and M. Itoh eds.),” Yuhihaku, Tokyo. Yoshihara, H., Sakuma, A., Itami, H., and Kagono, T. (1981). “The Diversification Strategy of the Japanese Firm (in Japanese),” Nihon Keizai Shinbun Sha, Tokyo.

27

Figure 1: Constrained vs. linked diversification
Constrained diversification Linked diversification

Core segment

Non-core segment

Closest link

Note: Adapted from Yoshihara et al. (1981) and modified by the authors.

28

Figure 2: The average ROA of sample firms, 1970-2002
0.10

0.08

0.06

0.04

0.02

0.00 1970 1975 1980 1985 1990 1995 2000

29

Table 1: Diversification of large Japanese manufacturing firms
Year Number of IO 4-digit industry segments 1973 1978 1983 1988 1993 1998 1973 1978 1983 1988 1993 1998 1973 1978 1983 1988 1993 1998 n 118 118 117 117 116 112 118 118 117 117 116 112 118 118 117 117 116 112 Mean 3.64 3.65 3.74 3.90 3.97 3.86 0.57 0.56 0.53 0.52 0.50 0.49 0.69 0.67 0.64 0.63 0.61 0.60 SD 1.95 1.84 1.83 1.71 1.78 1.65 0.24 0.24 0.24 0.24 0.24 0.24 0.21 0.21 0.22 0.21 0.22 0.22 Median 3 3 3 4 4 4 0.55 0.52 0.47 0.46 0.44 0.44 0.70 0.67 0.64 0.63 0.59 0.58 Q1 2 2 2 3 3 3 0.39 0.37 0.35 0.35 0.31 0.31 0.54 0.51 0.47 0.46 0.44 0.42 Q3 5 5 5 5 5 5 0.72 0.73 0.67 0.64 0.61 0.60 0.84 0.84 0.79 0.78 0.76 0.75

Herfindahl index (H it )

Core sales ratio (S ict )

30

Table 2: Relatedness measures of sample firms
Year NCC 1973 1978 1983 1988 1993 1998 1973 1978 1983 1988 1993 1998 1973 1978 1983 1988 1993 1998 1973 1978 1983 1988 1993 1998 n 106 107 107 108 107 103 118 118 117 117 116 112 (106) (107) (107) (108) (107) (103) Mean 0.43 0.44 0.43 0.43 0.42 0.43 0.81 0.81 0.79 0.79 0.77 0.77 (0.79) (0.79) (0.77) (0.77) (0.75) (0.75) SD 0.31 0.30 0.30 0.29 0.29 0.29 0.18 0.18 0.18 0.18 0.19 0.19 (0.17) (0.18) (0.18) (0.17) (0.18) (0.18) Median 0.40 0.44 0.41 0.43 0.40 0.42 0.87 0.85 0.83 0.81 0.79 0.80 (0.85) (0.82) (0.80) (0.79) (0.79) (0.77) Q1 0.15 0.16 0.15 0.15 0.18 0.16 0.67 0.70 0.66 0.69 0.67 0.66 (0.66) (0.68) (0.63) (0.68) (0.64) (0.66) Q3 0.65 0.64 0.62 0.61 0.59 0.59 0.97 0.96 0.95 0.94 0.93 0.93 (0.94) (0.95) (0.93) (0.92) (0.90) (0.91)

TRC

NCM

106 107 107 108 107 103 118 118 117 117 116 112 (106) (107) (107) (108) (107) (103)

0.54 0.55 0.55 0.55 0.56 0.55 0.86 0.85 0.84 0.84 0.83 0.83 (0.84) (0.84) (0.83) (0.83) (0.82) (0.81)

0.29 0.29 0.28 0.28 0.26 0.26 0.14 0.14 0.14 0.14 0.14 0.14 (0.14) (0.14) (0.14) (0.14) (0.14) (0.14)

0.57 0.57 0.59 0.59 0.56 0.56 0.89 0.89 0.88 0.88 0.85 0.85 (0.87) (0.88) (0.87) (0.87) (0.84) (0.84)

0.33 0.32 0.32 0.31 0.35 0.37 0.78 0.77 0.77 0.77 0.75 0.75 (0.77) (0.77) (0.75) (0.76) (0.74) (0.73)

0.76 0.75 0.76 0.75 0.74 0.76 0.97 0.96 0.95 0.95 0.95 0.94 (0.96) (0.95) (0.95) (0.94) (0.93) (0.93)

TRM

Note: in parentheses are values only for diversified firms.

31

Table 3: Summary statistics and description of variables n roa astg ln (ast ) 580 580 580 580 580 580 580 580 580 580 580 580 580 580 529 529 Mean 0.000 0.000 12.77 0.266 0.741 0.536 0.648 0.794 0.847 0.163 0.007 0.220 0.010 SD 0.020 0.043 0.959 0.263 0.144 0.239 0.216 0.181 0.141 0.133 0.052 0.155 0.064 Min Max Description ROA adjusted for the industry and time effects Annual growth rate of assets (time- and industry-adjusted) Log of assets in million yen Five year difference in log of assets Leverage defined as the ratio of liabilities to assets Five year difference in leverage Herfindahl index of diversification Five year difference in Herfindahl index Sales share of the core (largest) segment Five year difference in core sales ratio Firm coherence as defined in (3) Five year difference in TRC Firm coherence as defined in (5) Five year difference in TRM NCC as defined in (2) multiplied by one minus Sc Five year difference in (1-Sc )NCC NCM as defined in (4) multiplied by one minus Sc Five year difference in (1-Sc )NCM -0.064 0.095 -0.301 0.211 10.26 0.361 0.148 0.208 0.272 0.403 15.66 0.997 1.000 1.000 1.000 1.000 -0.885 1.276 -0.285 0.150 -0.653 0.272 -0.581 0.298 -0.295 0.290 -0.281 0.219 -0.009 0.554 -0.207 0.303 0.000 0.618 -0.231 0.343

∆ ln (ast ) lev ∆ lev
H

-0.029 0.061 -0.017 0.081 -0.017 0.083 -0.010 0.059 -0.007 0.053

∆H
Sc

∆ Sc
TRC

∆ TRC
TRM

∆ TRM
(1-Sc )NCC (1-Sc )NCM

∆ (1-Sc )NCC 529 ∆ (1-Sc )NCM 529

32

Table 4: FE estimation results of the profitability model
Dep. Variable ln (ast ) (1) roa -0.002 (0.002) 0.009** (0.003) -0.015 (0.011) -0.086*** (0.012) 0.007 (0.010) 0.027*** (0.009) (2) roa -0.003 (0.002) 0.008** (0.003) -0.023** (0.012) -0.090*** (0.012) -0.016 (0.014) 0.011 (0.012) 0.049** (0.021) 0.035** (0.017) (3) roa -0.002 (0.002) 0.009** (0.003) -0.015 (0.011) -0.086*** (0.012) 0.014 (0.013) 0.029** (0.011) (4) roa -0.003 (0.002) 0.008*** (0.003) -0.026** (0.012) -0.092*** (0.012) -0.007 (0.014) 0.015 (0.013) 0.077*** (0.024) 0.043** (0.019) -0.056** (0.024) -0.023 (0.019) (5) roa -0.003 (0.002) 0.012*** (0.004) -0.029** (0.012) -0.096*** (0.013)

∆ ln (ast ) lev ∆ lev
H

∆H
TRC

∆ TRC
TRM

∆ TRM
Sc

-0.018 (0.021) -0.004 (0.018)

∆ Sc
(1-Sc )NCC

∆ (1-Sc )NCC
(1-Sc )NCM

∆ (1-Sc )NCM
R 2 (within) n 0.130 580 0.142 580 0.132 580 0.153 580

0.019 (0.017) 0.040** (0.016) 0.093*** (0.025) 0.049** (0.021) -0.047** (0.024) -0.016 (0.019) 0.167 529

Note: In parentheses are standard errors. * significant at the 10% level. ** significant at the 5% level. *** significant at the 1% level.

33

Table 5: FE estimation results of the growth model
Dep. Variable ln (ast ) lev (6) astg -0.017*** (0.004) -0.104*** (0.028) 0.030 (0.030) -0.023 (0.024) 0.001 (0.023) (7) astg -0.018*** (0.004) -0.109*** (0.029) 0.030 (0.030) -0.046 (0.034) 0.009 (0.031) 0.051 (0.052) -0.011 (0.044) (8) astg -0.017*** (0.004) -0.101*** (0.028) 0.034 (0.030) -0.012 (0.032) 0.038 (0.028) (9) astg -0.018*** (0.004) -0.112*** (0.029) 0.028 (0.030) -0.033 (0.036) 0.028 (0.032) 0.078 (0.060) 0.033 (0.047) -0.071 (0.061) -0.113** (0.048) (10) astg -0.017*** (0.004) -0.103*** (0.030) 0.015 (0.031)

∆ lev
H

∆H
TRC

∆ TRC
TRM

∆ TRM
Sc

-0.031 (0.053) -0.099** (0.044)

∆ Sc
(1-Sc )NCC

∆ (1-Sc )NCC
(1-Sc )NCM

∆ (1-Sc )NCM

-0.039 (0.042) -0.059 (0.038) 0.068 (0.060) 0.028 (0.051) -0.079 (0.058) -0.111** (0.046)

0.079 0.082 0.090 0.094 0.085 R 2(within) n 580 580 580 580 529 Note: In parentheses are standard errors. * significant at the 10% level. ** significant at the 5% level. *** significant at the 1% level.

34