Journal of Asian Economics 15 (2004) 321–353
Foreign ownership and plant productivity in the
Thai automobile industry in 1996 and 1998:
a conditional quantile analysis
Keiko Ito*
The International Centre for the Study of East Asian Development (ICSEAD),
11-4, Otemachi, Kokurakita-ku, Kitakyushu, Japan
Received 3 February 2002; received in revised form 26 December 2003; accepted 25 February 2004
Abstract
This paper investigates the productivity differentials between foreign and local plants in the Thai
automobile industry using plant-level data for 1996 and 1998. The findings show that labor
productivity is higher at foreign-affiliated plants than at local plants. However, foreign plants in
the motor vehicle bodies and the motor vehicle parts industries tend to have a lower capital
productivity than local plants. Moreover, comparisons of TFP levels reveal no evidence that foreign
plants have higher TFP that can be related to their ownership-specific advantages.
# 2004 Elsevier Inc. All rights reserved.
JEL classification: D24; F23; L11; L62
Keywords: Productivity; Automobile industry; Thailand; Multinational corporations; FDI
1. Introduction
In the traditional theory of multinational corporations (MNCs), foreign direct
investment (FDI) by MNCs is regarded as the movement of managerial resources
(in other words, the intangible assets related to technological knowledge in production
and marketing or managerial know-how). A large body of literature on MNCs suggests
that a firm becomes internationally active in order to exploit the firm-specific
advantages embodied in its managerial resources (e.g. Caves, 1996; Dunning,
*
Tel.: þ81-93-583-6202; fax: þ81-93-583-4602.
E-mail address: [email protected] (K. Ito).
1049-0078/$ – see front matter # 2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.asieco.2004.02.005
322
K. Ito / Journal of Asian Economics 15 (2004) 321–353
1988; Markusen, 1991). Therefore, especially in developing countries, we would
expect MNCs to possess larger amounts of intangible assets than local firms and to
differ in terms of productivity or efficiency. Moreover, host countries’ attempts to
attract FDI are based on the expectation of technology transfer or technology spillover
effects from MNCs to local firms.
Recognizing the expected roles of MNCs, many researchers have investigated technology transfer from MNCs to local firms and provided comprehensive analyses based on
interviews and questionnaires. However, hardly any persuasive quantitative studies have
been conducted, mainly because of the limited availability of suitable data. However, in the
past few years, firm-level or establishment-level data have become partly available in some
countries, including the ASEAN countries. The few quantitative examinations that have
been conducted yield ambiguous results: Some of them suggest that, contrary to economic
theory, foreign MNCs are not significantly more productive than local firms. For example,
Okamoto (1999) analyzed the impact of Japanese FDI on the productivity of the US auto
parts industry using establishment-level data, and found that Japanese-affiliated firms were
less productive than their US counterparts in 1992. Ramstetter (2001), analyzing labor
productivity of manufacturing plants in Thailand in 1996 and 1998, found that labor
productivity differentials across plants mostly derived from differences in factor intensities
such as the capital–labor ratio and the ratio of the number of non-production workers to the
number of production workers. He concluded that after controlling for factor intensities,
there was no evidence suggesting that foreign plants enjoy systematically higher labor
productivity than local firms.1
Therefore, the so-called ‘‘Ownership advantage’’ in the theory of MNCs does not always
seem to operate in practice or, put differently, MNCs do not always exploit their firmspecific advantages in terms of productivity. However, such previous studies have not
satisfactorily investigated why foreign MNCs were not able to perform well in some host
countries, and what characteristics of the MNCs or host countries prevented the MNCs
from exploiting their ownership advantages in the host countries.
Using establishment-level data for the Thai automobile industries in 1996 and 1998,
Ito (2002a) examined the following two questions. (1) Are foreign plants more productive than local plants? (2) What characteristics determine the productivity of plants? In
order to answer these questions, the study tested if there was a statistically significant
difference in productivity between local and foreign plants; and it explored which
factors determined plant productivity and whether the difference in ownership was an
important determinant of productivity. Scale, vintage, learning and trade effects as well
1
Okamoto and Sjöholm (2000), examining productivity performance and its dynamics in the Indonesian
automobile industry between 1990 and 1995, concluded that although foreign establishments tended to show a
better performance than local firms, the spillover effect of foreign MNCs did not seem to have been strong. In Ito
(2002b), I conducted a more detailed analysis of productivity in the Indonesian automobile industry for the
period from 1990 to 1999. While the study confirmed that foreign plants tended to display higher labor
productivity, productivity growth for foreign and local plants alike was mostly explained by the scale effect
rather than any technological change. Using panel data on Venezuelan plants, Aitken and Harrison (1999) found
that plant productivity is positively correlated with foreign participation. However, in their test for spillover
effects from foreign-owned plants to local plants, they found that foreign investment negatively affected the
productivity of domestically owned plants.
K. Ito / Journal of Asian Economics 15 (2004) 321–353
323
as ownership or managerial ability were discussed and tested, mainly using two measures
of plant productivity: labor productivity (value added per hour for production workers)
and relative total factor productivity (TFP). Ito (2002a) concluded that productivity
differentials between foreign and local plants were not pervasive in the Thai automobile
industry. However, the paper suffered from one statistical drawback: the ordinary least
squares (OLS) regression employed there does not yield coefficient estimates that are
representative of the distribution of automobile plants when the observed productivity is
not normally distributed.
This paper, therefore, tries to obtain more robust results, employing the quantile
regression technique in order to take into account any possible bias due to the nonnormality and unobserved heterogeneity among plants.
The remainder of the paper is organized as follows. Section 2 provides a brief overview
of the development of the Thai automobile industry. Using plant-level data, Section 3
compares various productivity measures of local and foreign establishments, and examines
whether the differences of the average productivity measures are statistically significant
using the mean comparison test (t-test). Section 4 discusses the theoretical principles of the
determinants of productivity levels, explains the methodology of quantile regressions and
reports the results of the regression analysis. Section 5 concludes the paper by considering
the implications of the findings.
2. Overview of the ASEAN and Thai automobile industry
Many developing countries have targeted the automobile sector as a major focus of
their industrialization drive. A number of characteristics explain why this industry has
been considered more conducive to development efforts than other sectors. These
particular industry characteristics include the following. The automobile sector is: (1)
a large-scale industry requiring a huge amount of capital; (2) a synthetic industry
which has wide-ranging related industries and where significant technology transfer
effects on related industries are expected; (3) a key industry the output of which makes
up a significant proportion of gross domestic product (GDP) due to its large scale,
when the conditions of market size, infrastructure, and technology levels are satisfied.2
Thailand is no exception in this attempt to foster an automobile industry within the
country, and in response to these efforts, major foreign automakers started operating
there in the 1960s and 1970s. Moreover, since the late 1980s, many Japanese auto
parts suppliers have striven to set up reliable production and marketing networks in
Thailand.
The Thai government first began to provide investment incentives by enacting the 1960
Industrial Promotion Act as part of its import substitution strategy. In response to the
investment promotion policy, the first automobile assembly plant in Thailand, the AngloThai Motor Company, started operating in 1961. After that, at least nine assembly plants
2
See Waseda Daigaku Shogakubu [Faculty of Commerce, Waseda University, Japan] and Zaidan-Hojin Keizai
Kouhou Center (Eds.) (1995).
324
K. Ito / Journal of Asian Economics 15 (2004) 321–353
were set up during 1961–1969. In the 1970s and early 1980s, due to the strengthening of
local content requirements, many Japanese automobile parts suppliers established
production companies or concluded license agreements with local manufacturers, while
some of the European and US automakers discontinued local assembly because of the
intensified competition.3 Throughout this process of automobile industry promotion, the
Thai government protected the domestic sector by imposing high tariffs on completely
built-up automobiles (CBU) and lured foreign automakers by offering them various
incentives for investing in the country. The intention was to achieve import substitution
by promoting joint-ventures between foreign automakers and local firms. In addition, in
an attempt to promote the localization of parts and components, the government set local
content requirements for automakers. This policy encouraged Japanese automakers to
bring in their parts suppliers to Thailand. In the early 1990s, the government reversed
this policy and reduced tariff rates and removed the restriction limiting foreign ownership to 49%. Due to these liberalization policies and rapid economic growth, the Thai
automobile market and FDI inflows into the automobile industry in Thailand grew
rapidly in the first half of the 1990s. However, annual automobile production and sales in
Thailand at their peak reached no more than about 600 000 units.4 This is far lower than
in Japan, where approximately 10 million vehicles are produced annually and 6–7
million a year sold domestically. The small size of developing countries’ domestic
market, which prevents the establishment of efficient-sized plants, has been highlighted
by many researchers as a major obstacle to the development of a domestic auto parts
industry.5,6
In 1999, there were thirteen major foreign automakers with assembly plants in Thailand
that had a total annual production capacity of 976 000 units.7 Most of the plants are located
3
As a result, most of the automobiles sold in the country have been made by Japanese automakers. The market
share of Japanese automobile manufacturers in Thailand in terms of unit sales was more than 90% as of 1996,
and in 2001 Japanese automobile manufacturers still held a market share of 80%.
4
Automobile production and sales in Thailand hit a peak in 1996. After the economic crisis in 1997, however,
sales dropped to 363 156 units from 589 126 units in 1996. In 1998, automobile sales continued to tumble,
declining to 144 065 units, which was a drop of 76% from 1996 sales (Poapongsakorn & Wangdee, 2000).
5
For the case of Thailand, these issues have been discussed, for example, in Buranathanung (1996),
Maruhashi (1995), Terdudomtham (1997), and Yahata and Mizuno (1988).
6
To take advantage of economies of scale, the AICO (ASEAN Industrial Cooperation) scheme was launched
in November 1996. This is designed to encourage cooperative industrial production in the ASEAN countries and
permits participants to gain early access to preferential tariff rates, while normal tariff rates remain high.
However, the member governments of ASEAN have tended to require a degree of reciprocity in terms of
balanced trade flows under the agreement (Farrell & Findlay, 2001).
7
The thirteen automakers were as follows: Isuzu (49% owned by General Motors), Ford Motor, Mazda
(33.4% owned by Ford Motor), Volvo (100% owned by Ford Motor), DaimlerChrysler (Mercedes-Benz),
Mitsubishi (34% owned by DaimlerChrysler), Toyota, Hino (33.8% owned by Toyota Motor), Nissan (36.8%
owned by Renault), Nissan Diesel (22.5% owned by Renault and 22.5% owned by Nissan), Peugeot and Citroen,
Honda, and BMW. Out of the total production capacity of 976 000 units in Thailand, the capacity of the Toyota
Group (Toyota and Hino) is 260 000 units, thus accounting for about 27% of total capacity. Each of the other
four major groups, the GM–Isuzu group, the Ford–Mazda–Volvo group, the DaimlerChrysler–Mitsubishi group,
and the Renault–Nissan group, had a share of about 15% of total production capacity. Moreover, the production
capacity in Thailand accounts for about 40% of total capacity in the ASEAN-4 countries. Even in Thailand,
however, the capacity utilization ratio remained at only around 34% in 1999 (FOURIN, 2000).
K. Ito / Journal of Asian Economics 15 (2004) 321–353
325
in Bangkok, the Vicinity of Bangkok area, and the Eastern Region. Major local and
foreign-owned auto parts plants are also concentrated in these regions.
3. Differences in the economic performance of local and foreign plants
In the remainder of this paper, I investigate the differences in various productivity
measures between local and foreign plants for the Thai automobile industry using the plantlevel data underlying National Statistical Office (1999, 2001). Although plant-level data
are available for 1996 and 1998, the analysis in this paper mainly relies on the 1996 data
since the data for 1998 are not very appropriate for a productivity analysis as most
automobile plants in Thailand were still suffering from the slump following the 1997 East
Asian crisis.8
Table 1 summarizes the plant-level data used in the analysis.9 In Thailand, about 20%
of all plants in the manufacturing sector as a whole are owned by foreigners. Moreover,
nearly half of all workers in manufacturing are employed in foreign plants, and foreign
plants account for more than half of gross output or value added in manufacturing as a
whole accrues. In the motor vehicle industry, more than half of all workers are employed
in foreign plants, and approximately 90% of gross output or value added accrues to
foreign plants. Looking at the size of the motor vehicle industry at a more detailedindustry level, 42.3% of workers in the motor vehicle industry as a whole were engaged in
the motor vehicle assembly industry in 1996 (the upper panel of Table 1). 77.2% of gross
output and 79.8% of value added in the motor vehicle industry as a whole came from the
motor vehicle assembly industry in the same year. Therefore, most of the gross output or
value added in the Thai automobile industry comes from foreign-owned assembly plants.
Although a fairly large number of auto body manufacturing plants or auto parts plants
exist and employ a large number of workers, the output or value added generated by them
is rather small.
Given the significance of foreign-owned plants in the Thai automobile industry,
let us now explore differences in the productivity of local and foreign plants, using the
plant-level data for 1996 and 1998 underlying National Statistical Office (1999,
2001).10
Fig. 1 shows the distribution of labor productivity (in logarithmic form) measured by
value added per hour for production workers for the year 1996 for the motor vehicle
industry as a whole and the three detailed industries: motor vehicle assembly (hereafter
8
In addition, although the motor vehicle industry is divided into the three detailed industries (i.e. the motor
vehicle assembly, the motor vehicle bodies and trailers, and the motor vehicle parts and accessories) in the 1996
data, it is divided into only two detailed industries in the 1998 data: motor vehicle assembly and parts and motor
vehicle bodies and trailers.
9
The figures in Table 1 are compiled using only reliable samples as explained in Ramstetter (2001).
Ramstetter (2001) found that the original plant-level data collected by the National Statistical Office of Thailand
contain several duplicate or near-duplicate records. He checked the duplicates and eliminated the duplicate
records from the data set for his analysis. Therefore, the figures in Table 1 are not equal to those in the National
Statistical Office’s publications (National Statistical Office, 1999, 2001).
10
For the methodologies of calculating productivity measures, see Appendix A.
326
K. Ito / Journal of Asian Economics 15 (2004) 321–353
Table 1
Summary of the plant-level data underlying the Thai industrial census
Number of
plants/firms
Value
Number of
workers
Share (%) Value
Gross output
(million baht)
Share (%) Value
Year 1996
Manufacturing sector total (shares in manufacturing sector
Total
8,952
(100.0)
1,669,504 (100.0)
Local
7,214
(80.6)
948,990 (56.8)
Foreign 1,738
(19.4)
720,514 (43.2)
Value added
(million baht)
Share (%) Value
total in parentheses)
2,717,842 (100.0)
1,145,301 (42.1)
1,572,541 (57.9)
Share (%)
809,575
378,400
431,175
(100.0)
(46.7)
(53.3)
Motor vehicles sector total (shares in motor vehicles sector total in parentheses)
Total
406
(100.0)
78,253 (100.0)
420,434 (100.0)
Local
330
(81.3)
32,808 (41.9)
43,362 (10.3)
Foreign
76
(18.7)
45,445 (58.1)
377,072 (89.7)
129,401
13,744
115,657
(100.0)
(10.6)
(89.4)
Motor vehicle
Total
Local
Foreign
103,294
1,509
101,785
(79.8)
(1.2)
(78.7)
assembly
46
28
15
(shares in motor vehicles sector total in parentheses)
(11.3)
33,062 (42.3)
324,676 (77.2)
(6.9)
3,433 (4.4)
7,799 (1.9)
(3.7)
29,629 (37.9)
316,877 (75.4)
Motor vehicle bodies and trailers (shares in motor vehicles sector total
Total
186
(45.8)
15,425 (19.7)
50,657
Local
178
(43.8)
10,918 (14.0)
15,009
Foreign
8
(2.0)
4,507 (5.8)
35,648
Motor vehicle
Total
Local
Foreign
parts and accessories (shares in motor vehicles
177
(43.6)
29,766 (38.0)
124
(30.5)
18,457 (23.6)
53
(13.1)
11,309 (14.5)
in parentheses)
(12.0)
11,987
(3.6)
4,403
(8.5)
7,584
sector total in parentheses)
45,101 (10.7)
14,120
20,553 (4.9)
7,832
24,547 (5.8)
6,288
Year 1998
Manufacturing sector total (shares in manufacturing sector total in parentheses)
Total
3,974
(100.0)
853,697 (100.0)
1,328,983 (100.0)
Local
3,026
(76.1)
461,593 (54.1)
559,402 (42.1)
Foreign
948
(23.9)
392,104 (45.9)
769,581 (57.9)
Motor vehicles sector total (shares in motor vehicles sector total in parentheses)
Total
134
(100.0)
27,164 (100.0)
105,952 (100.0)
Local
81
(60.4)
8,389 (30.9)
8,053 (7.6)
Foreign
53
(39.6)
18,775 (69.1)
97,899 (92.4)
Motor vehicle
Total
Local
Foreign
(9.3)
(3.4)
(5.9)
(10.9)
(6.1)
(4.9)
408,017
164,787
243,230
(100.0)
(40.4)
(59.6)
35,044
1,901
33,143
(100.0)
(5.4)
(94.6)
assembly, parts and accessories (shares in motor vehicles sector total in parentheses)
106
(79.1)
25,507 (93.9)
104,718 (98.8)
34,543 (98.6)
53
(39.6)
6,732 (24.8)
6,818 (6.4)
1,401 (4.0)
53
(39.6)
18,775 (69.1)
97,899 (92.4)
33,143 (94.6)
Motor vehicle bodies and trailers (shares in motor vehicles sector total in parentheses)
Local
28
(20.9)
1,657 (6.1)
1,235 (1.2)
500
(1.4)
No foreign-owned establishments were classified in the motor vehicle bodies and accessories industry in 1998.
Sources: Appendix Tables A.1–A.3, A6, and A.7 in Ramstetter (2001), and compiled from plant-level data
underlying National Statistical Office (1999, 2001).
K. Ito / Journal of Asian Economics 15 (2004) 321–353
327
% of establishments
(a) Motor vehicle assembly
25
20
15
Foreign
10
Local
5
0
0
2
4
6
ln (labor productivity)
% of establishments
(b) Motor vehicle bodies and trailers
30
25
20
15
10
5
0
Foreign
Local
0
2
4
6
ln (labor productivity)
% of establishments
(c) Motor vehicle parts and accessories
25
20
15
Foreign
10
Local
5
0
0
2
4
6
% of establishments
ln (labor productivity)
(d) Motor vehicles: total
30
25
20
15
10
5
0
Foreign
Local
0
2
4
6
ln (labor productivity)
Fig. 1. Labor productivity distribution in 1996 (logarithm of value added per hour for production workers) (local
vs. foreign). Compiled from plant-level data underlying National Statistical Office (1999).
simply: assembly industry), motor vehicle bodies and trailers (hereafter: bodies and
trailers industry), and motor vehicle parts and accessories (hereafter: parts industry).
The number of establishments is calculated at each level of productivity for the two
groups, foreign plants and local plants. Fig. 1 presents the share of the number of
establishments at each level in the total number of establishments. It is found that in
the assembly and the bodies and trailers industries the majority of foreign plants
enjoyed a higher level of productivity, while local ones registered lower levels of
328
K. Ito / Journal of Asian Economics 15 (2004) 321–353
productivity (panels (a) and (b)). In contrast, in the parts industry and in the motor
vehicle industry as a whole, although foreign plants dominate the higher levels of
productivity, they also hold a significant share at the lower levels of productivity
(panels (c) and (d)).
% of establishments
(a) Motor vehicle assembly
35
30
25
20
15
10
5
0
Foreign
Local
-5
-2
1
4
% of establishments
ln (capital productivity)
(b) Motor vehicle bodies and trailers
20
15
Foreign
10
Local
5
0
-5
-2
1
4
% of establishments
ln (capital productivity)
(c) Motor vehicle parts and accessories
20
15
Foreign
10
Local
5
0
-5
-2
1
4
% of establishments
ln (capital productivity)
(d) Motor vehicles: total
20
15
Foreign
10
Local
5
0
-5
-2
1
4
ln (capital productivity)
Fig. 2. Capital productivity distribution in 1996 (logarithm of value added per baht of fixed assets) (local vs.
foreign). Compiled from plant-level data underlying National Statistical Office (1999).
K. Ito / Journal of Asian Economics 15 (2004) 321–353
329
Looking at capital productivity (in logs) measured by value added per baht of fixed
assets in Fig. 2, many foreign plants tend to have lower capital productivity than local
plants, particularly in the parts industry (panel (c)). In the assembly industry,
although many foreign plants tend to display higher capital productivity, some
% of establishments
(a) Motor vehicle assembly
30
25
20
15
10
5
0
Foreign
Local
-5
-3
-1
1
ln (relative TFP)
% of establishments
(b) Motor vehicle bodies and trailers
25
20
Foreign
15
Local
10
5
0
-5
-3
-1
1
ln (relative TFP)
% of establishments
(c) Motor vehicle parts and accessories
25
20
15
Foreign
10
Local
5
0
-5
-3
-1
1
ln (relative TFP)
% of establishments
(d) Motor vehicles: Total
25
20
Foreign
15
Local
10
5
0
-5
-3
-1
1
ln (relative TFP)
Fig. 3. Distribution of relative total factor productivity (logarithm) in 1996. Compiled from plant-level data
underlying National Statistical Office (1999).
330
K. Ito / Journal of Asian Economics 15 (2004) 321–353
foreign plants are registered low levels of capital productivity (panel (a)). In the
motor vehicle industry as a whole, it seems that more foreign plants are found at lower
levels (panel (d)).
In addition, the relative TFP level of each plant is calculated, and its distribution in 1996
is shown in Fig. 3.11 This also indicates that foreign plants seem to exhibit higher relative
TFP in the assembly industry, while there is no clear tendency in the other two industries
(panels (a), (b), and (c)). Looking at the distribution in the motor vehicle industry as a
whole, again no clear tendency can be observed (panel (d)).12
The remainder of this section tests the hypothesis suggested by MNC theory that firmspecific managerial advantages enable foreign-owned plants to do better than local ones in
terms of productivity or efficiency. The simple mean comparison test (t-test) is applied to
the analysis in this section, although the distributions of various productivity measures
presented in Figs. 1–3 seem to have thicker right tails and imply non-normal distributions.13
Table 2 compares the various productivity measures of local and foreign plants in
1996 for each of the three detailed industries. In all three industries, the average
employment size at foreign plant is larger than at local ones, and the difference is
statistically significant in the assembly and the bodies and trailers industries. Although
average employment at foreign plants is approximately 15 times greater than at local
ones in the assembly industry, this may be because many small tuk-tuk repair shops are
classified in this industry.14 As for years in operation, foreign plants are significantly
older than local ones in the assembly industry, while they are newer than local ones in
the other two industries, statistically significantly so in the parts industry. The amount
of registered capital and the capital–labor ratio are much larger for foreign plants and
the difference is statistically significant in all three industries. For each of the four labor
productivity measures, foreign plants are significantly more productive than local ones
in all three industries. However, as for capital productivity and the relative TFP level,
the comparison between local and foreign plants shows different results for the
industries. In the assembly industry, capital productivity and the relative TFP level
are significantly higher for foreign plants. On the other hand, capital productivity is
higher in local plants than in foreign plants in the other two industries and the
difference is statistically significant in most cases. As for relative TFP, this tends to
11
Relative TFP is defined as the TFP level for each plant relative to a hypothetical plant whose value added,
labor input and capital input are the industry average values. The formula for calculating relative TFP is
presented in the Appendix A.
12
The performance distribution for the year 1998 is also investigated and the results are shown in Fig. B.1 in
the Appendix B. For the year 1998, the distribution displays similar tendencies as in 1996.
13
It would also be interesting to examine the differentials among foreign plants by nationality. However, in
the automobile industry, the majority of foreign plants are Japanese. In the assembly industry, 14 out of 15
foreign plants are Japanese. In the bodies and trailers industry, 7 out of 8 foreign plants are Japanese. In the
parts industry, 37 out of 53 foreign plants are Japanese, one is from Korea, four are from Taiwan, two from
the United States, two from Europe, and one is from China. The nationality of the rest is unknown. Given
such dominance of Japanese plants, we do not distinguish foreign plants by nationality in the analyses in
this paper.
14
A tuk-tuk is an auto three-wheeler, which is very popular in Thailand and other Asian developing countries.
Table 2
Comparison of economic performance: by ownership
Assembly (1996)
Simple average
t-Test
Foreign
28
3,433
123
10
50
605
15
29,629
1,975
18
1,442
1,732
539
102
Number of establishments
Total number of employees
Employment per establishment
Years in operation
Registered capital (billion baht)
Capital–labor ratio (1,000 baht)
Productivity measures
Output per hour for production
workers (baht)
Value added per hour for
production workers (baht)
Output per employee
(1,000 baht)
Value added per employee
(1,000 baht)
Output per 1 baht of fixed assets
Value added per 1 baht of
fixed assets
Relative TFP
Parts and accessories (1996)
Simple average
Simple average
t-Test
Local
Foreign
–
–
178
10918
61
9.5
14.0
385
8
4507
563
8.4
248.4
3,078
–
–
5,490
411.3
1,032
818.7
11,857.6
186.9
2,884.1
10.147
3.739
0.403
Local
Total (1996)
Total (1998)
t-Test Simple average
Foreign
t-Test
Local
Foreign
76
45,445
598.0
9.6
388.1
1,645.6
–
–
330
32,808
99.4
10.2
19.6
413.6
2,056
t-Test
Local
Foreign
53
18,775
354.2
9.8
290.9
2,249.0
–
–
81
8,389
103.6
12.6
27.4
633.0
373.2
1,402.8
124
18457
149
11.2
21.4
412
53
11309
213
7.4
110.9
1,405
3303.4
355.8
896.1
401.3
128.3
793.1
124.7
317.7
124.7
508.8
124.4
473.2
738.9
6,984.7
759.0
1,967.0
753.2
4,447.2
528.8
1,906.3
244.1
1,603.8
271.4
654.9
249.5
1,194.8
196.2
624.8
40.102 7.011 1.275 38.231
15.934
9.577 2.555 0.146
0.377
26.145
11.715
0.296
–
–
Simple average
8.025 3.091
0.292
(1) The t-tests are performed based on the assumption of unequal variances. Significant at 10% level, significant at 5% level,
each year. Source: compiled from plant-level data underlying National Statistical Office (1999, 2001).
31.307
13.314
14.519 3.808 11.224
4.768
9.197
2.679 0.224
0.088 0.367
0.115 K. Ito / Journal of Asian Economics 15 (2004) 321–353
Local
Bodies and trailers (1996)
significant at 1% level (two-tailed test). (2) Figures are at market prices for
331
332
K. Ito / Journal of Asian Economics 15 (2004) 321–353
be higher for foreign plants in these two industries, but the difference is not statistically
significant.15
Table 2 compares the difference in economic performance between foreign and local
plants in the whole motor vehicle industry for both 1996 and 1998.16 For both years, it is
found that foreign plants display higher values in terms of employment per plant, registered
capital, the capital–labor ratio, labor productivity, and relative TFP. However, foreign
plants tend to suffer from significantly lower capital productivity for both years, although
the decline in average capital productivity from 1996 to 1998 is larger in local plants. It
should be noted that the comparison between 1996 and 1998 should be viewed with
caution, because the sample plants in each data set are not the same.
The results in Table 2 show that the labor productivity of foreign plants is significantly
higher than that of local ones. However, it should be taken into account that foreign plants
tend to be larger in terms of employment per plant and capital intensity.17
4. Determinants of productivity levels in the Thai automobile industry
As seen in the previous section, foreign plants exhibit significantly higher labor
productivity than their local counterparts. However, as pointed out by Ito (2002a) and
Ramstetter (2001), the higher labor productivity of foreign plants might be explained by the
higher capital–labor ratio. This section examines whether the higher productivity is
explained by advantages in the managerial resources of MNCs after controlling for factor
intensities. In order to obtain more robust results than Ito (2002a), the determinants of
productivity are investigated using conditional quantile regression analysis.
15
A comparison of other economic indicators such as inventory ratios, price-cost margins, wages, and so on, is
presented in Table B.1 in the Appendix B. Wages at foreign plants are significantly higher than at local ones. The
results summarized in Table B.1 imply that foreign plants in the assembly industry are relatively successful in
their inventory management, while foreign plants in the other two industries still suffered from holding large
amounts of inventory in 1996. This is probably because of the lack of lower-tiered parts suppliers in Thailand,
forcing foreign parts suppliers to rely to a considerable extent on imported materials. However, in 1998, the
inventory ratios tended to be higher for local plants. These figures of inventory ratios suggest that the demand
shock caused by the 1997 East Asian crisis was more moderate for foreign plants, or that foreign plants were
more successful in adjusting their inventories by exporting to other countries. Another indication hinting at
foreign plants’ faster recovery from the crisis is that the price-cost margin was significantly higher for foreign
plants in 1998. Although the average price-cost margin of local plants declined from 1996 to 1998, that of
foreign plants increased.
16
Although each plant in the motor vehicle industry is classified in the three detailed industries for the original
1996 data, for the 1998 original data, it is classified in only two detailed industries, motor vehicle assembly and
parts, and motor vehicle bodies and trailers. Moreover, as the plant code number is different for the two data sets
and the number cannot be matched, we could not conduct a panel-data analysis.
17
Conducting the same analysis for large plants only produces similar results. Following Ramstetter (2001), a
large plant is defined as a plant whose output exceeds 25 million baht. The results for large plants are shown in
Table B.2 in the Appendix B. Even with the limited sample, the difference between foreign and local plants
shows a similar tendency for most of the economic indicators as the results for the full sample. Foreign firms
show significantly higher labor productivity, capital–labor ratios, and wages. However, capital productivity is
lower in foreign plants, and there is almost no difference in the relative TFP level between local and foreign
plants.
K. Ito / Journal of Asian Economics 15 (2004) 321–353
333
4.1. Analytical principles
Two measures of productivity are used: labor productivity calculated as value added per
hour for production workers (in log form) and the relative TFP level of plant i (in log
form).18 These productivity measures are estimated directly as a function of plant
characteristics as follows:
VAi
Ki
ENi
(1)
ln
¼ a0 þ a1 ln
þ a2 ln
þ a3 Doldi þ a4 Dboii þ a5 Zi
EPi
EPi
EPi
lnðRLTFPi Þ ¼ b0 þ b01 Sizei þ b2 Doldi þ b3 Dboii þ b4 Zi
(2)
where VAi: value added of plant i; RLTFPi: relative TFP level of plant i; EPi: hours worked
by production workers of plant i; ENi: hours worked by non-production workers of plant i;
Ki: average book value of fixed assets of plant i; Doldi: a dummy variable that takes 1 if
plant i started its operations in 1986 or earlier, otherwise 0; Dboii: a dummy variable that
takes 1 if plant i is BOI (Board of Investment)-promoted, otherwise 0;19 Sizei: a vector of
variables representing the size of plant i; (therefore, b01 is the transpose of the parameter
vectors) Zi: other characteristics of plant i, such as ownership, openness to international
trade, and industry dummies.
Further details on the definitions and sources of the variables are provided in the
Appendix A20.
Following Baily, Hulten, and Campbell (1992), five factors which should determine the
plant productivity level are considered: (1) factor intensity, (2) differences in managerial
resources and other plant fixed effects, (3) returns to scale, (4) learning by doing, and (5)
vintage capital.
(1) Factor intensity
As derived from the Cobb–Douglas production function, the coefficients of factor
intensities in equation (1) will be positive.
(2) Differences in managerial resources and other plant fixed effects
The distribution of productivity at a point in time may be a reflection of plant
heterogeneity. Heterogeneity is considered to be related to managerial abilities,
ownership, BOI status, openness to international trade, products of the plant, or
geographical characteristics of the region where the plant is located. According to
18
The measure of labor productivity ignores the substitution of labor input for other inputs and differences in
technical efficiency and input composition at different scales of production. Therefore, labor productivity is a
more restrictive measure with respect to these points. On the other hand, the measure of the relative TFP level,
which is based on the Törnqvist–Theil-translog index, accommodates differential substitution across inputs and
allows for scale and regulatory biases, without restrictive assumptions about the shape of the technological
relationships. However, we should note that the construction of the relative TFP measure would often contain
measurement errors due to the difficulties in calculating the real value of inputs, especially in the case of capital
inputs. Particularly in developing countries, constructing comprehensive capital stock data is very difficult
because of the lack of deflators. For the exact translog index numbers, see Caves, Christensen, and Diewert
(1982).
19
The correlation matrix of the variables is shown in Table B.3 in the Appendix B.
20
The 1998 data do not contain information on BOI promotion.
334
K. Ito / Journal of Asian Economics 15 (2004) 321–353
traditional theory, MNCs possess advantages in managerial resources that make them
more productive or efficient than local firms. As BOI-promoted plants tend to have
more advanced technologies or management know-how, one would expect them to be
more productive. As for openness to international trade, it may be expected that
exporting plants are more exposed to international competition and should therefore
have higher productivity than non-exporters. If plants use more imported materials
which embody high technology or are of a higher quality, such plants might be
expected to be more productive than those that do not use imported materials. On the
other hand, however, plants using imported materials might be less efficient as they
incur extra costs such as import tariffs and transportation costs.
(3) Returns to scale
The empirical literature suggests that there may be increasing returns to scale at the
plant level. If scale is an important determinant of productivity, larger plants should
exhibit higher productivity.
(4) Learning by doing
Following the idea first proposed by Arrow (1962) that knowledge and productivity
gains derive from investment and production, experience would have a positive effect
on productivity. In order to examine this hypothesis, a dummy variable for the age of a
plant is employed; older plants are expected to exhibit higher productivity.
(5) The vintage capital model
It is assumed that when a plant is built it embodies a particular technological
vintage. Therefore, a measure of the vintage of plant i at time t, vit, is included in the
production function as follows:
Qit ¼ FðKit ; Lit ; vit Þ
If the vintage model does provide a correct explanation of plant-level productivity,
older plants are expected to have lower productivity. Therefore, in contrast to the
learning-by-doing effect, the relative productivity level of plants is expected to
decline over time. However, as there are many old plants that may have been reequipped, it is possible that plant age is unimportant if technological change is
being correctly captured in the capital price deflators. Since there are no reliable
capital deflators in many developing countries, it is necessary to test the
assumption.
Most previous studies on this kind of micro-level productivity employ standard OLS or
GMM techniques which concentrate on the conditional mean function of the dependent
variable. When dealing with cross-sections of firms or plants like in this study, however,
OLS or GMM estimates may not be representative of the entire distribution of the
dependent variable if this is not identically distributed across firms or plants. In the
presence of large and persistent heterogeneity across plants, the usual assumption in the
literature of normally distributed productivity may not hold. Table 3 shows detailed
summary statistics and P-values for two tests of normality for the logarithms of labor
productivity and relative TFP. In most cases, the normality assumption is rejected at a high
significance level. Figs. 1 and 3 also indicated that the distributions of productivity
measures for 1996 have relatively thicker right tails as well as long left tails and depart
Value added per hour for production workers (in logarithm)
Relative TFP level (in logarithm)
1996
1996
Total
Mean
S.D.
10th quantile
25th quantile
Median
75th quantile
90th quantile
Skewness
Kurtosis
Number of observations
Test 1 (P-value)
Test 2 (P-value)
4.500
1.197
3.154
3.797
4.346
5.140
6.301
0.208
3.889
406
0.000
0.004
1998
Local
4.292
1.018
3.115
3.717
4.291
4.843
5.543
0.214
4.988
330
0.000
0.000
Foreign
5.404
1.474
3.577
4.231
5.385
6.819
7.213
0.267
1.999
76
0.014
0.005
Total
4.693
1.319
3.316
4.009
4.748
5.468
6.225
0.487
4.516
134
0.001
0.000
Local
4.155
1.240
2.295
3.702
4.355
4.821
5.501
0.837
4.122
81
0.000
0.000
Foreign
5.516
0.971
4.447
4.824
5.324
6.159
6.739
0.842
4.105
53
0.516
0.350
Total
0.166
1.044
1.336
0.764
0.162
0.389
1.155
0.453
5.007
406
0.000
0.000
1998
Local
0.224
0.977
1.357
0.762
0.183
0.303
0.896
0.578
5.954
330
0.000
0.000
Foreign
0.088
1.272
1.202
0.861
0.048
1.181
1.743
0.455
3.093
76
0.017
0.208
Total
0.173
1.138
1.747
0.797
0.020
0.572
1.204
0.788
3.977
132
0.001
0.000
Local
0.367
1.233
2.353
1.002
0.083
0.474
1.039
0.792
3.518
79
0.008
0.009
Foreign
0.115
0.916
1.149
0.412
0.068
0.750
1.285
0.154
3.047
53
0.691
0.780
K. Ito / Journal of Asian Economics 15 (2004) 321–353
Table 3
Summary statistics of labor productivity (auto industry as a whole)
Test 1: test for normality (Shapiro & Francia, 1972). Test 2: skewness and kurtosis test for normality (D’Agostino, Balanger, & D’Agostino Jr., 1990).
335
336
K. Ito / Journal of Asian Economics 15 (2004) 321–353
from normality.21 Such distribution characteristics require special econometric treatment
that provides a more robust and efficient alternative to least squares estimators.
Thus, the quantile regression technique developed by Koenker and Bassett (1978) is
employed in this paper.22 Denoting the dependent variables by Y and the vector of
regressors by X, the quantile regression model can be written as:
Yi ¼ Xi0 gy þ eyi ;
Quanty ðYi jXi Þ ¼ Xi0 gi
(3)
where gy is the vector of parameters to be estimated for a given value of the distribution’s
quantile y in (0, 1); Quanty ðYi jXi Þ denotes the yth quantile of the conditional distribution of
Yi given the known vector of regressors Xi. The quantile regression model therefore
provides predictions of a specific quantile y of the conditional distribution of Yi. The
estimation of the yth quantile parameters gy has been implemented by solving the following
minimization problem:
(
)
X
0
jYi Xi gjhi
min
(4)
g
i
with
hi ¼
2y
2ð1 yÞ
if Yi Xi0 g > 0
if otherwise
Here jYi Xi0 gj is the absolute value of the residual, and the weight, hi, is derived from the
definition of the yth sample quantile. For y ¼ 0:5, the median is obtained and problem (4) is
equivalent to the problem of minimum absolute deviations. For the estimation of quantiles
other than the median, the residuals are weighted appropriately, depending on whether they
are positive or negative.23 This contrasts with the OLS approach, which weighs negative
and positive deviations equally and assigns a disproportionate weight to large deviations in
the calculation. Therefore, compared to the OLS method, the quantile regression method
assigns a smaller weight to larger errors in order to reduce the distortion caused by outliers.
Using the method suggested by Koenker and Bassett (1978), the estimated standard errors
are underestimated in cases of heteroscedastic errors. For this reason, robust standard errors
were obtained by using the option of bootstrapping procedures introduced by Gould (1992,
1997). By increasing y from 0 to 1, one can trace the entire conditional distribution of plantlevel productivity, conditional on the set of regressors. Therefore, we can focus attention on
specific parts of the productivity distribution and see differences in the effects exerted on
productivity by independent variables at various quantiles. This is a particularly important
advantage of quantile regressions, which allows us to analyze different responses of plants
at different points of the productivity distribution. This study considers regression
estimates at five different quantiles, i.e. the 10th, 25th, 50th (median), 75th, 90th
21
Fig. B.1 in the Appendix B shows that the productivity distributions for 1998 also depart from normality.
Dimelis and Louri (2001) and Grima and Görg (2002) use this technique to analyze firm-level or plant-level
productivity and spillovers from FDI.
23
This is set up as a linear programming problem and solved via linear programming techniques as suggested
by Armstrong, Frome, and Kung (1979).
22
K. Ito / Journal of Asian Economics 15 (2004) 321–353
337
percentiles of the productivity distribution as well as estimates of the conditional mean.
The estimations of the conditional mean are equivalent to the OLS estimation.
4.2. Empirical results
The model specification of equation (1) in Section 4.1 is derived from the Cobb–Douglas
production function under restrictive assumptions: (1) there are constant returns to scale,
and (2) differences in ownership or other plant-specific characteristics only affect labor
productivity. The estimation results of equation (1) for the year 1996 are presented in
Table 4. Panel A shows the estimation results without industry dummies, and the OLS
estimation (column 1) in panel A of Table 4 is exactly the same as that conducted by
Ramstetter (2001) in his Appendix Table C12. Columns 2–6 in panel A of Table 4 present
the estimation results at various quantiles. Panel B of Table 4 shows the estimation results
when the industry is controlled for by dummy variables (Dassy and Dbody).24 Overall,
according to the results in Table 4, differentials in the capital–labor ratio (ln(K/EP)) and the
ratio of non-production workers to production workers (ln(EN/EP)) are significantly
positively correlated with the differentials in labor productivity in most cases. The dummy
for large plants (Dlarge) tends to have a significantly positive coefficient, suggesting that
large plants tend to show higher labor productivity. Moreover, plants with BOI promotion
(Dboi) achieve significantly higher productivity. As for the openness to international trade
(Dx, Dm), the coefficients do not show a significant correlation with labor productivity in
most cases. Coefficients on the old-plant dummy variable (Dold) are not significant,
suggesting that there are neither learning nor vintage capital effects, or that the two effects
offset each other. The coefficients on the foreign ownership dummy (Df) are positive but
not statistically significant in many cases, which does not suggest a strong tendency for
foreign plants to show higher labor productivity than local plants as a result of their
advantages in managerial resources. However, according to the results of the quantile
regressions, the coefficients on Df are positive and statistically significant at the 75th and
the 90th quantiles. This can be interpreted as evidence that foreign ownership does matter
only among the relatively more productive plants. It is only in the upper-productivity range
that foreign-owned plants confirm their superior efficiency by causing a productivity shift.
This result might imply that foreign-owned plants can exploit their ownership-specific
advantages only when some level of productivity is achieved.
In addition, since the two industry dummies (Dassy and Dbody) have significantly
positive coefficients in most cases (panel B of Table 4), it can be concluded that plants in
the parts industry display lower labor productivity than those in the other two industries.
Table 5 shows the estimation results of equation (1) for the year 1998. Contrary to the
results in Table 4, the coefficients on the foreign ownership dummy (Df) are significantly
positive in most cases, although factor intensities (ln(K/EP) and ln(EN/EP)) hardly have
significant coefficients. Because most automobile plants in Thailand were still suffering
from the slump in 1998 due to the 1997 East Asian crisis, the data for 1998 are not very
24
In the case of controlling for both the industry and the region (BOI-Zones 1, 2, and 3), the result was mostly
consistent with that in the case of controlling only for the industry. Moreover, the region dummies were not
statistically significant.
338
Table 4
Regression results for labor productivity for the year 1996 (auto industry as a whole)
OLS Estimates
Quantile regression estimates
mean (1)
10th quantile (2)
Age
Dold
Foreign plants
Df
Dboi
Dx
Dm
_cons
Number of observations
Pseudo R-square
F
Adj. R-square
0.176 (1.60)
0.244 (1.25)
0.592 (2.78)
0.160 (1.00)
0.036 (0.33)
3.905 (20.90)
406
25th quantile (3)
Median (4)
75th quantile (5)
90th quantile (6)
0.234 (2.53)
0.154 (0.96)
0.145 (2.62)
0.240 (2.83)
0.171 (3.55)
0.201 (2.67)
0.211 (7.01)
0.206 (2.76)
0.276 (5.50)
0.047 (0.44)
0.059 (0.27)
0.228 (1.37)
0.135 (1.04)
0.129 (1.05)
0.218 (0.97)
0.287 (0.93)
0.162 (0.39)
0.370 (1.53)
0.310 (1.81)
2.817 (6.15)
0.048 (0.24)
0.779 (2.65)
0.185 (0.87)
0.260 (1.51)
3.809 (13.48)
0.135 (0.53)
1.018 (3.92)
0.053 (0.31)
0.117 (0.99)
4.046 (18.48)
0.890 (2.86)
0.649 (2.09)
0.049 (0.24)
0.008 (0.08)
4.452 (26.79)
0.665 (2.12)
0.806 (1.98)
0.306 (0.92)
1.055 (0.45)
4.554 (16.60)
406
0.0822
406
0.0874
406
0.1504
406
0.2628
406
0.3135
24.9
0.287
(Panel B) Industry controlled: year 1996
Factor intensity
ln(K/EP)
0.186 (5.25)
ln(EN/EP)
0.209 (3.42)
0.138 (1.08)
0.251 (1.47)
0.202 (2.99)
0.216 (2.87)
0.168 (4.07)
0.184 (2.68)
Age
Dold
0.066 (0.64)
0.018 (0.07)
0.126 (0.85)
0.028 (0.24)
Foreign plants
Df
Dboi
Dx
0.168 (0.87)
0.412 (2.02)
0.000 (0.00)
0.313 (0.98)
0.003 (0.01)
0.385 (1.36)
0.004 (0.02)
0.274 (1.15)
0.232 (1.07)
0.168 (0.59)
0.816 (3.06)
0.196 (1.07)
0.165 (4.33)
0.137 (1.67)
0.046 (0.35)
0.671 (2.44)
0.612 (2.37)
0.432 (2.25)
0.194 (4.41)
0.129 (1.63)
0.234 (1.01)
0.717 (2.38)
0.034 (0.11)
0.367 (1.51)
K. Ito / Journal of Asian Economics 15 (2004) 321–353
(Panel A) Industry not controlled: year 1996
Factor intensity
ln(K/EP)
0.225 (6.10)
ln(EN/EP)
0.209 (3.15)
Dm
Dlarge
Industry
Dassy
Dbody
_cons
0.603 (3.84)
0.466 (4.01)
3.485 (17.45)
406
0.250 (1.11)
0.463 (2.09)
0.568 (1.30)
0.518 (2.40)
2.812 (4.76)
406
0.1170
0.092 (0.57)
0.503 (4.11)
0.660 (3.32)
0.530 (2.83)
3.067 (10.07)
406
0.1383
0.093 (0.90)
0.538 (4.40)
0.102
0.996 (5.39)
0.282 (0.87)
1.176 (4.81)
0.373 (1.69)
0.418 (2.86)
3.658 (18.16)
0.393 (2.15)
0.435 (3.07)
3.970 (18.91)
0.429 (2.60)
0.425 (2.13)
4.211 (14.53)
406
0.2049
406
0.3215
406
0.3974
26.83
0.3756
The figures in parentheses are t-statistics. t-Statistics for the OLS estimation (column 1) are based on White’s robust standard errors (White, 1980). Significant at 10%
level, significant at 5% level, significant at 1% level (two-tailed test). Dependent variable: value added per hour for production workers (ln(VA/EP)). Source:
author’s calculations.
K. Ito / Journal of Asian Economics 15 (2004) 321–353
Number of observations
Pseudo R-square
F
Adj. R-square
0.009 (0.08)
0.790 (6.71)
339
340
Table 5
Regression results for labor productivity for the year 1998 (auto industry as a whole)
Quantile regression estimates
Mean (1)
10th quantile (2)
25th quantile (3)
Median (4)
75th quantile (5)
90th quantile (6)
0.136 (1.77)
0.028 (0.26)
0.400 (1.90)
0.077 (0.43)
0.062 (0.39)
0.072 (0.56)
0.096 (1.53)
0.021 (0.17)
0.196 (1.76)
0.048 (0.34)
0.095 (0.78)
0.187 (0.97)
Age
Dold
0.112 (0.42)
0.505 (0.75)
0.202 (0.66)
0.257 (1.00)
0.053 (0.15)
0.355 (0.77)
Foreign plants
Df
Dx
Dm
_cons
1.035 (4.41)
0.269 (1.08)
0.282 (1.11)
4.850 (13.60)
1.288 (2.50)
0.154 (0.21)
0.508 (0.76)
4.764 (5.17)
0.904 (2.19)
0.322 (0.69)
0.418 (1.21)
4.113 (7.31)
0.808 (3.98)
0.362 (1.11)
0.053 (0.19)
4.655 (11.65)
0.699 (1.74)
0.447 (1.21)
0.298 (0.84)
5.746 (11.52)
132
0.2206
132
0.163
132
0.1595
132
0.1777
Number of observations 132
Pseudo R-square
F
9.51
Adj. R-square
0.3068
0.909 (2.02)
0.288 (0.77)
0.094 (0.22)
5.662 (8.43)
132
0.1954
The figures in parentheses are t-statistics. t-Statistics for the OLS estimation (column 1) are based on White’s robust standard errors (White, 1980). Significant at 10%
level, significant at 5% level, significant at 1% level (two-tailed test). Dependent variable: value added per hour for production workers (ln(VA/EP)). Source:
author’s calculations.
K. Ito / Journal of Asian Economics 15 (2004) 321–353
Factor intensity
ln(K/EP)
ln(EN/EP)
OLS estimates
Table 6
Regression results for relative TFP Level for the year 1996 (auto industry as a whole)
OLS estimates
Quantile regression estimates
Mean (1)
10th quantile (2)
0.106 (1.12)
0.142 (0.85)
25th quantile (3)
0.197 (3.28)
0.110 (0.85)
Median (4)
0.268 (3.80)
0.034 (0.39)
75th quantile (5)
90th quantile(6)
0.332 (6.47)
0.033 (0.41)
0.377 (7.36)
0.092 (0.86)
0.155 (0.96)
Age
Dold
0.018 (0.18)
0.045 (0.23)
0.026 (0.17)
0.197 (1.68)
0.112 (0.91)
Foreign plants
Df
Dboi
Dx
Dm
1.391
0.099
0.015
0.119
2.525 (0.79)
0.641 (1.44)
0.480 (1.40)
0.455 (1.98)
2.206
0.227
0.198
0.086
1.025
0.044
0.003
0.083
0.475 (0.30)
0.173 (0.83)
0.368 (2.01)
0.047 (0.50)
1.512 (0.73)
0.059 (0.20)
0.586 (3.12)
0.048 (0.34)
Industry
Dassy
Dbody
_cons
0.104 (0.61)
0.306 (2.68)
4.558 (6.42)
0.142 (0.36)
0.423 (1.44)
3.175 (2.00)
0.186 (0.61)
0.445 (2.96)
4.215 (4.57)
0.267 (1.44)
0.369 (2.47)
4.780 (4.07)
0.087 (0.55)
0.379 (2.60)
5.454 (6.08)
0.359 (1.79)
0.075 (0.45)
5.487 (6.02)
406
406
0.0615
406
0.0935
406
0.1249
406
0.1972
406
0.2573
Number of observations
Pseudo R-square
F
Adj. R-square
(1.08)
(0.49)
(0.10)
(1.11)
(0.96)
(0.66)
(1.19)
(0.49)
(0.67)
(0.21)
(0.02)
(0.70)
12.21
0.2099
Panel B: Industry controlled: year 1996
Size
LOUT Dassy
0.144 (1.37)
LOUT Dbody
0.307 (5.55)
LOUT Dparts
0.225 (3.98)
Df LOUT Dassy
0.098 (1.03)
Df LOUT Dbody
0.032 (0.32)
0.073 (0.34)
0.287 (2.44)
0.080 (0.93)
0.094 (0.54)
0.044 (0.26)
0.129 (0.69)
0.186 (2.85)
0.212 (3.55)
0.024 (0.13)
0.031 (0.16)
0.166 (0.92)
0.352 (3.40)
0.248 (3.18)
0.122 (1.09)
0.049 (0.43)
0.117 (0.84)
0.340 (7.05)
0.282 (3.66)
0.129 (1.08)
0.021 (0.18)
K. Ito / Journal of Asian Economics 15 (2004) 321–353
Panel A: Industry controlled: year 1996
Size
LOUT
0.250 (5.85)
Df LOUT
0.059 (0.87)
0.163 (1.13)
0.387 (5.76)
0.367 (4.55)
0.037 (0.31)
0.098 (0.78)
341
342
Table 6 (Continued )
Df LOUT Dparts
OLS estimates
Quantile regression estimates
Mean (1)
10th quantile (2)
25th quantile (3)
Median (4)
75th quantile (5)
90th quantile(6)
0.075 (0.39)
0.011 (0.06)
0.107 (0.84)
0.097 (0.70)
0.106 (0.75)
0.012 (0.12)
0.018 (0.09)
0.005 (0.03)
0.210 (1.73)
0.090 (0.74)
0.203 (1.31)
Foreign plants
Df
Dboi
Dx
Dm
1.421
0.071
0.017
0.117
0.022
0.162
0.223
0.130
2.262
0.027
0.029
0.018
1.907
0.183
0.282
0.032
Industry
Dassy
Dbody
_cons
1.375 (0.72)
1.064 (0.84)
4.131 (4.36)
0.120 (0.03)
3.025 (1.21)
2.720 (1.90)
1.499 (0.49)
0.931 (0.65)
4.489 (4.87)
1.626 (0.53)
1.328 (0.63)
4.489 (3.40)
2.837 (1.17)
0.516 (0.33)
4.690 (3.56)
2.873 (1.07)
0.323 (0.18)
5.321 (3.78)
406
0.0797
10.01
0.2182
406
0.0989
406
0.1344
406
0.2120
406
406
0.2647
Number of observations
Pseudo R-square
F
Adj. R-square
(0.70)
(0.35)
(0.12)
(1.09)
1.064
0.764
0.271
0.348
(0.30)
(1.51)
(0.99)
(1.33)
(0.01)
(0.51)
(1.41)
(0.70)
(1.00)
(0.12)
(0.23)
(0.12)
(0.76)
(0.71)
(1.58)
(0.29)
1.495 (0.55)
0.427 (1.46)
0.588 (3.24)
0.071 (0.50)
a
The figures in parentheses are t-statistics based on White’s robust standard errors (White, 1980). Significant at 10% level, significant at 5% level, significant at 1%
level (two-tailed test). Dependent variable: relative TFP level (ln(RLTFP)). Source: author’s calculations.
K. Ito / Journal of Asian Economics 15 (2004) 321–353
0.032 (0.56)
Age
Dold
Table 7
Regression results for relative TFP level for the year 1998 (auto industry as a whole)
Quantile regression estimates
Mean (1)
10th quantile (2)
25th quantile (3)
Median (4)
75th quantile (5)
90th quantile (6)
Size
LOUT
Df LOUT
0.293 (2.19)
0.052 (0.35)
0.177 (0.59)
0.159 (0.32)
0.298 (1.41)
0.116 (0.47)
0.339 (3.57)
0.071 (0.44)
0.303 (2.80)
0.119 (0.65)
0.369 (2.21)
0.105 (0.60)
Age
Dold
0.183 (0.87)
0.352 (0.48)
0.115 (0.31)
0.282 (1.24)
0.340 (1.55)
0.252 (0.91)
Foreign plants
Df
Dx
Dm
_cons
0.893 (0.34)
0.002 (0.01)
0.201 (0.89)
5.151 (2.32)
2.497
0.050
0.061
4.829
2.075
0.023
0.065
5.820
0.830 (0.28)
0.135 (0.37)
0.015 (0.08)
5.701 (3.59)
1.888 (0.57)
0.025 (0.10)
0.180 (1.18)
4.589 (2.51)
1.660 (0.53)
0.337 (0.77)
0.224 (0.75)
5.399 (1.98)
Number of observations
Pseudo R-square
F
Adj. R-square
132
0.1252
5.69
0.1754
132
0.102
132
0.1213
132
0.1433
132
(0.28)
(0.09)
(0.14)
(0.93)
(0.48)
(0.05)
(0.26)
(1.66)
132
0.1109
The figures in parentheses are t-statistics. t-Statistics for the OLS estimation (column 1) are based on White’s robust standard errors (White, 1980). Significant at 10%
level, significant at 5% level, significant at 1% level (two-tailed test). Dependent variable: relative TFP level (ln(RLTFP)). Source: author’s calculations.
K. Ito / Journal of Asian Economics 15 (2004) 321–353
OLS estimates
343
344
K. Ito / Journal of Asian Economics 15 (2004) 321–353
appropriate for the productivity analysis. Moreover, the detailed industries cannot be
controlled for because the 1998 data do not provide adequate information. Therefore, even
though the productivity level is significantly higher for foreign-owned plants in the year
1998, the results should be interpreted with caution. The fact that labor productivity is
higher for foreign-owned plants after the Asian economic crisis might be interpreted not as
higher efficiency but as evidence that foreign-owned plants may have received some form
of support from parent firms.
The estimation results of equation (2) for the year 1996 are shown in Table 6. The
estimates indicate no strong evidence suggesting that foreign plants have a significantly
higher relative TFP (Df). In order to investigate whether the importance of scale economies
differs among the industries or between foreign and local plants, a variety of cross-terms is
introduced in panels A and B in Table 6. The results of panel A indicate that scale
economies (LOUT) have significantly positive effects on TFP levels but scale economies
for foreign plants (Df LOUT) are not significantly larger than those for local plants.
However, when controlling for the industries (panel B), it is found that the effects of scale
economies are significantly larger in the bodies and trailers and the parts industries
(LOUT Dbody, LOUT Dparts). Nevertheless, no differences in the effects of scale
economies can be observed when comparing foreign and local plants (Df LOUT Dassy, Df LOUT Dbody, Df LOUT Dparts). As for other plant-specific characteristics, the age dummy variable (Dold), BOI-promotion (Dboi), and trade orientation
variables (Dx and Dm), all do not have any significant impact on relative TFP levels.
In addition, the estimation results of equation (2) for the year 1998, which are shown in
Table 7, also indicate that there is no significant difference in TFP between foreign and
local plants. For 1998 as well as for 1996, plant size is the only significant determinant of
relative TFP levels.
5. Concluding remarks
In this paper, two questions were examined. One asked whether or not foreign plants are
more productive than local plants. The other was what characteristics determine the
productivity of plants.
Comparing the simple mean of various productivity measures of foreign and local plants,
it was found that foreign plants have significantly higher labor productivity and capital–
labor ratios. Capital productivity is significantly higher for foreign than local plants in the
assembly industry. However, it is significantly lower for foreign plants in the bodies and
trailers and the parts industries. The relative TFP level is significantly higher for foreign
plants only in the assembly industry. Therefore, the various productivity measures suggest
that although foreign plants in the bodies and trailers and the parts industries have a higher
capital–labor ratio than local plants, the high capital intensity probably has not yet led to
productivity improvements.
The results of the regression analyses provided only weak evidence that foreign plants
possess higher labor productivity after controlling for factor intensities. However, the
quantile regression results on the determinants of labor productivity imply that foreignowned plants have a superior efficiency in the upper-productivity range and that foreign
K. Ito / Journal of Asian Economics 15 (2004) 321–353
345
ownership does matter among the relatively more productive plants. In the regression
results on the determinants of relative TFP, again, there was no strong evidence that foreign
plants had relatively higher TFP as a result of advantages in managerial resources. On the
other hand, strongly significant effects of scale economies were found both in the labor
productivity equations and the relative TFP equations. The effects of scale economies are
significantly larger in the bodies and trailers and the parts industries.
Two problems explicitly raised by the analysis in this paper are the following: first,
foreign plants in the bodies and trailers and the parts industries display low productivity
despite of their high capital intensity. Many foreign plants in these industries were
relatively newly established compared with the foreign plants in the assembly industry.
Due to the lack of experience in operation, workers in foreign plants may not have
been skilled enough yet to utilize state-of-the-art equipment. Another reason might be
one suggested by Cusumano and Takeishi (1991) in their study on Japanese auto
parts suppliers operating in the United States, which showed that the shortage of
maintenance workers made it difficult for Japanese-owned firms to effectively utilize
newly purchased equipment. In addition, foreign-owned plants tend to pay higher
wages in order to retain their workers, which might cause the low TFP for foreignowned plants.
The second problem raised is that economies of scale are an important determinant of
plant productivity, particularly in the bodies and trailers and the parts industries.
Although many foreign plants in these two industries are equipped with a lot of
machinery and equipment, their production scale may be far below the minimum
efficient scale. This may be the reason for the low TFP for plants in these industries.
These results also imply that participating in the AICO (ASEAN Industrial Cooperation)
scheme in order to achieve economies of scale would raise the productivity of Thai
automobile plants.
In line with the major findings in Ito (2002a) and Ramstetter (2001, 2002), the results
of this paper suggest that productivity differentials between foreign and local plants are
not pervasive in the Thai automobile industry. The results also suggest that the low level
of productivity and the high capital–labor ratio of foreign plants could be partly
explained by the low productivity of capital and by the small scale of production.
Ramstetter (2001) provides two possible interpretations of his results: (1) foreign plants
are not that productive in Thai manufacturing, or (2) local plants have been successful in
reaching productivity levels that do not differ pervasively from those in generally
efficient foreign MNCs. In order to answer which interpretation applies to the Thai
automobile industry, further investigations are necessary regarding issues such as
productivity differentials between MNCs’ parent plants and affiliated plants, the determinants of the differentials between them, and technology spillover effects from MNCaffiliated plants to local plants.
Acknowledgements
This paper was completed as a part of the ICSEAD (International Centre for the Study of
East Asian Development) project on ‘‘Foreign Multinational Corporations and Host-
346
K. Ito / Journal of Asian Economics 15 (2004) 321–353
Country Labor Markets in Asia’’ and the author would like to thank Profs. Kyoji Fukao and
Hiroyuki Odagiri of Hitotsubashi University, Prof. Yumiko Okamoto of Nagoya University, and Prof. Shinichi Ichimura, Dr. Eric Ramstetter, and other research staff at
ICSEAD for advice on this study. This paper has also benefited from the constructive
comments of two anonymous referees and the editor of the journal on an earlier version.
Financial support from the Setsutaro Kobayashi Memorial Fund of Fuji Xerox Company is
gratefully acknowledged. All remaining errors are mine.
Appendix A
A.1. The data
The data used in this paper are plant-level data for 1996 and 1998, compilations of the
1997 Industrial Census and the 1999 Industrial Survey collected by the National Statistical
Office, Office of the Prime Minister, Thailand. The original data sets contain several
duplicate or near-duplicate records. On the methodology of eliminating these duplicate
records, see Ramstetter (2001). Moreover, a large number of small plants (10–19 employees) are included in the original data, but they were dropped from the analysis for the
following reasons: (1) I do not think it is very meaningful to compare very small local
establishments with foreign MNCs, (2) there are very few foreign establishments that are
this small, and (3) it is very difficult to check the duplicate information for smaller
establishments. Moreover, I only included in my analysis the data for plants which have
positive observations for the number of non-production workers, the number of production
workers, intermediate inputs, and value added.
A.2. Calculation of relative total factor productivity (RLTFP)
Our measure of the relative TFP level for each plant is the one suggested by Caves,
Christensen, and Diewert (1982): relative TFP is calculated by relating the deviation of
plant output from the industry mean to the deviations of the factor inputs from the industry
mean. The specification is:
X
1
1
ln RLTFPi ¼ ln VAi ln VA aLPi ðln EPi ln EPÞ
aEPi þ
2
n
X
1
1
1
aENi ðln ENi ln ENÞ aENi þ
2
n
2
X 1
aKi ðln Ki ln KÞ
a Ki þ
n
1
P
1
P
1
P
where ln
VA
¼
,
ln
EP
¼
,
ln
EN
¼
ln
VA
ln
EP
ln ENi , and
i
i
n
n
n
P
ln Ki .VAi is gross output of plant i, and EPi, ENi, and Ki are production
ln K ¼ 1n
labor inputs, non-production labor inputs, and capital inputs of plant i. aEPi ; aENi ; aKi are the
factor cost shares of production labor, non-production labor, and capital inputs of plant i.
For labor inputs, EPi and ENi, total hours worked by production workers and total hours
K. Ito / Journal of Asian Economics 15 (2004) 321–353
347
worked by non-production workers are used, respectively. As a proxy variable for real
capital stocks, the book value of fixed assets is used. In the calculation of capital costs, the
rental rate of capital (pk) is estimated as follows:
dqk
pk ¼ qk r þ dK qk
where qk is the price of the investment goods, r is the rate of return on all capital, and dk is
the rate of depreciation of the investment good. Finally, dqk/qk is the rate of capital gain on
that good (Jorgenson & Griliches, 1967). Then the implicit cost of capital is obtained by
multiplying the book value of fixed assets with the rental rate per baht of capital. The
producer price index for capital equipment is used for the price of the investment goods,
and the prime lending rates (commercial banks, minimum loan rates) are used for the rate
of return on capital. Both data were taken from Bank of Thailand, Quarterly Bulletin. The
depreciation rate is set at 10%.
A.3. Definitions and descriptions of variables used in the regressions
Ownership dummies:
Df:
1 if the foreign ownership share is 1% or greater, 0 otherwise.
We should note that in the Census, each plant is labeled either (1) 100% foreignowned, (2) 50–99% foreign-owned, or (3) 1–49% foreign owned. Out of the 76
foreign plants in the sample for the year 1996, only three were 100% foreign
owned, 19 were 50–99% foreign owned, and 54 were 1–49% foreign owned.
Although I tried to conduct some OLS regression analyses including three kinds
of foreign dummies (Df-100, Df-majority, Df-minority), all the dummy variables
did not have significant coefficients. Refer to Ito (2002a).
Trade orientation dummies:
Dx:
Dm:
1 if the plant exports, 0 otherwise;
1 if the plant imports, 0 otherwise.
Industry dummies:
Dassy:
Dbody:
Dparts:
1 if the plant is classified in the motor vehicle assembly industry, 0 otherwise;
1 if the plant is classified in the motor vehicle bodies and trailers industry, 0
otherwise;
1 if the plant is classified in the motor vehicle parts and accessories industry,
0 otherwise.
Plant size:
Dlarge:
LOUT:
1 if the output is 25 million baht or more, 0 otherwise;
logarithm of the output of the plant.
348
K. Ito / Journal of Asian Economics 15 (2004) 321–353
Appendix B
Additional Figure and Tables
% of establishments
(a) Labor productivity (logarithm) distribution
20
15
Foreign
10
Local
5
0
1
3
5
7
ln (labor productivity)
% of establishments
(b) Relative TFP (logarithm) distribution
20
15
Foreign
10
Local
5
0
-4
-2
0
1
ln (relative TFP)
Fig. B.1. Productivity distribution in the whole motor vehicle industry in 1998 (local vs. foreign). Compiled
from plant-level data underlying National Statistical Office (2001).
Table B.1
Comparison of various economic indicators: by ownership
Assembly
Inventory ratios (%)
Total inventory (%)
Final goods inventory (%)
Work-in-process inventory (%)
Raw materials inventory (%)
Other indicators
Share of non-production
workers (%)
Price-cost margin (%)
Production worker wages
(1000 baht)
Non-production worker
wages (1000 baht)
Total worker wages
(1000 baht)
Capital utilization (%)
Local
Foreign
11.7
0.6
3.2
7.9
6.7
1.1
0.5
5.1
19.3
18.6
22.2
63.5
26.1
161.8
95.9
t-Test
Parts and accessories
Total (1996)
Simple average
Simple average
Simple average
Local
t-Test
t-Test
Foreign
Local
Foreign
Local
Foreign
9.1
1.0
2.2
6.0
9.4
0.9
1.3
7.2
17.2
4.3
2.2
10.7
20.9
8.6
3.2
9.1
12.4
2.2
2.3
7.9
16.9
6.3
2.5
8.1
16.5
17.8
15.8
20.4
16.5
21.9
77.6
21.1
143.1
17.8
75.2
22.3
103.9
158.3
102.7
226.2
141.9
302.4
68.8
160.7
81.1
166.1
80.7
133.4
27.5
38.6
27.6
35.4
31.9
38.5
Total (1998)
t-Test
Simple average
t-Test
Local
Foreign
5.7
14.2
26.7
6.8
3.8
13.1
24.4
7.5
19.8
26.6
34.2
20.4
75.5
22.9
119.5
14.6
99.6
24.0
145.3
116.9
265.9
136.3
322.3
79.9
142.3
87.9
154.1
29.2
38.2
26.9
29.3
K. Ito / Journal of Asian Economics 15 (2004) 321–353
Simple average
Bodies and trailers
(1) The t-tests are performed based on the assumption of unequal variances. Significant at 10% level, significant at 5% level, significant at 1% level (two-tailed
test). (2) Figures are at market prices for each year. (3) Definition of indicators are as follows. Average inventory ¼ 1=2 ðinventory Jan 1 þ inventory Dec 1); pricecost margin ¼ ðvalue added wage income of all employees social security payments of all employees)/output; capital utilization ¼ ðhours per day days per year)/
(24 366). Source: compiled from plant-level data underlying National Statistical Office (1999, 2001).
349
350
K. Ito / Journal of Asian Economics 15 (2004) 321–353
Table B.2
Comparison of economic performance: by ownership (motor vehicles total, large plants, 1996 and 1998)
Total (1996)
Total (1998)
Simple average
Number of establishments
Total number of employees
Employment per establishment
Years in operation
Registered capital (billion baht)
Productivity measures
Output per hour for production
workers (baht)
Value added per hour for production
workers (baht)
Output per employee (1,000 baht)
Value added per employee (1,000 baht)
Output per 1 baht of fixed assets
Value added per 1 baht of fixed assets
Relative TFP
Inventory ratios (%)
Total inventory (%)
Final goods inventory (%)
Work-in-process inventory (%)
Raw materials inventory (%)
Other indicators
Share of non-production workers (%)
Capital–labor ratio (1,000 baht)
Price-cost margin (%)
Production worker wages (1,000 baht)
Non-production worker wages
(1,000 baht)
Total worker wages (1,000 baht)
Capital utilization (%)
t-Test
Local
Foreign
117
25,210
215
12.6
44.8
70
45,183
645
9.8
420.1
824.6
212.7
Simple average
t-Test
Local
Foreign
–
–
29
6,049
208.6
15.8
58.9
50
18,671
373.4
10.0
308.6
2220.0
699.9
1,470.5
548.9
197.6
494.0
932.0
314.1
11.670
5.027
0.120
1,522.7
4,803.9
426.8
1,289.2
37.230
14.532
14.109
3.631
0.123
0.194
12.5
2.2
2.7
7.6
14.3
3.7
2.5
8.1
16.9
618.1
20.1
90.4
164.2
19.6
1,745.7
24.1
124.9
279.1
97.3
30.9
148.7
38.8
1992.9
649.7
9.612
2.777
0.138
23.8
4.4
3.7
13.1
24.0
6.9
4.0
30.5
783.3
23.4
143.7
128.3
34.6
2,335.2
23.5
151.0
337.3
115.5
27.2
160.0
29.5
15.6
–
–
(1) The t-tests are performed based on the assumption of unequal variances. Significant at 10% level,
significant at 5% level, significant at 1% level (two-tailed test). (2) Figures are at market prices for each
year. (3) Definition of indicators are as follows. Average inventory ¼ 1=2 ðinventory Jan 1 þ inventory Dec
1); price-cost margin ¼ ðvalue added wage income of all employees social security payments of all
employees)/output; capital utilization ¼ ðhours per day days per year)/(24 366). Source: compiled from
plant-level data underlying National Statistical Office (1999, 2001).
K. Ito / Journal of Asian Economics 15 (2004) 321–353
351
Table B.3
Correlation matrix
ln(VA/EP) ln(RLTFP) ln(K/EP) ln(EN/EP) LOUT
Panel A: Year 1996 (observations ¼ 406)
ln(VA/EP)
1
ln(RLTFP)
0.8237
1
ln(K/EP)
0.4552
0.091
1
ln(EN/EP)
0.2378
0.035
0.209
1
LOUT
0.6434
0.4037
0.5047
0.1139
Dold
0.0608
0.0813
0.0015
0.0083
Df
0.3628
0.1167
0.4435
0.1101
Dboi
0.3965
0.1576
0.432
0.1329
Dx
0.2632
0.1524
0.2611
0.0176
Dm
0.1971
0.0559
0.2574
0.1817
Dassy
0.1679
0.1148
0.1191
0.0512
Dbody
0.0755
0.0375
0.2218 0.0237
Dlarge
0.4402
0.2797
0.3663
0.0459
Dm
Dassy
Dbody
Dlarge
Dm
Dassy
Dbody
1
0.04
0.2751
0.2754
1
0.3165
0.0192
1
0.3438
Dlarge
Dzone1
Dzone2
1
0.0706
0.2686
1
0.6913
Dboi
Dx
Dlarge
1
Panel B: Year 1998 (observations ¼ 132)
ln(VA/EP)
1
ln(RLTFP)
0.8036
1
ln(K/EP)
0.4031
0.1334
1
ln(EN/EP)
0.1248
0.2364
0.118
1
LOUT
0.6242
0.4025
1
Dold
0.119
0.0518
0.0971 0.1265
Df
0.5196
0.2085
0.5452
0.1579
Dx
0.4087
0.2006
0.4364
0.3602
Dm
0.092
0.0093
0.1841
0.2203
Dlarge
0.548
0.323
0.7418
0.0033
Dzone1
0.0282
0.1789
0.1807 0.0345
Dzone2
0.1602
0.0028
0.2496
0.0364
Dzone1
Df
1
0.1725 1
0.5632 0.0732 1
0.5521 0.0436 0.6465 1
0.5346 0.1341 0.5011 0.4595 1
0.3749 0.0536 0.3227 0.3274 0.3708
0.2542 0.0773 0.1426 0.1689 0.1075
0.3068 0.0599 0.3399 0.2206 0.3297
0.7608 0.1323 0.4433 0.3859 0.4673
ln(VA/EP) ln(RLTFP) ln(K/EP) ln(EN/EP) LOUT
Dlarge
Dold
Dold
Df
Dx
Dm
0.0622 1
0.5643 0.1027 1
0.5298 0.1006 0.6587 1
0.2325 0.1338 0.3272 0.3299 1
0.5872 0.5076 0.2042
0.0439 0.2828 0.2269 0.0322 0.0715
0.1953 0.2231 0.2525 0.139 0.0762
Dzone2
1
References
Aitken, B. J., & Harrison, A. E. (1999). Do domestic firms benefit from direct foreign investment? Evidence
from Venezuela. American Economic Review, 89(3), 605–618.
Armstrong, R. D., Frome, E. L., & Kung, D. S. (1979). A revised simplex algorithm for the absolute deviation
curve fitting problem. Communications in Statistics Simulation and Computation, B8(2), 175–190.
352
K. Ito / Journal of Asian Economics 15 (2004) 321–353
Arrow, K. J. (1962). The economic implications of learning by doing. Review of Economic Studies, 29,
155–173.
Baily, M. N., Hulten, C., & Campbell, D. (1992). Productivity dynamics in manufacturing plants. Brookings
Papers on Economic Activity: Microeconomics, 187–267.
Buranathanung, N. (1996). The organization of parts procurement in the Thai automobile industry.
Chulalongkorn Journal of Economics, 8(1), 67–99.
Caves, D. W., Christensen, L. R., & Diewert, W. E. (1982). Multilateral comparisons of output, input and
productivity using superlative index numbers. Economic Journal, 92, 73–86.
Caves, R. E. (1996). Multinational enterprise and economic analysis (2nd ed.). Cambridge: Cambridge
University Press.
Cusumano, M. A., & Takeishi, A. (1991). Supplier relations and management: A survey of Japanese, Japanesetransplant, and US auto plants. Strategic Management Journal, 12, 563–588.
D’Agostino, R. B., Balanger, A., & D’Agostino, R.B., Jr. (1990). A suggestion for using powerful and
informative tests of normality. American Statistician, 44, 316–321.
Dimelis, S., & Louri, H. (2001). Foreign direct investment and efficiency benefits: A conditional quantile
analysis. Discussion paper no. 2868, June, Centre for Economic Policy Research, London.
Dunning, J. H. (1988). Explaining international production. London: Unwin Hyman.
Farrell, R., & Findlay, C. (2001). Japan and the ASEAN4 automotive industry: Developments and interrelationships in the regional automotive industry. Working Paper Series, vol. 2001-24, August, The
International Centre for the Study of East Asian Development, Kitakyushu.
FOURIN (2000). 2000 Asean, Taiwan Jidosha Buhin Sangyo [The auto parts industry in ASEAN and Taiwan,
2000]. Nagoya: FOURIN.
Gould, W. W. (1992). Quantile regression with bootstrapped standard errors. Stata Technical Bulletin, 9, 19–21.
Reprinted in Stata Technical Bulletin Reprints, 2, 137–139.
Gould, W. W. (1997). Interquantile and simultaneous-quantile regression. Stata Technical Bulletin, 38, 14–22.
Reprinted in Stata Technical Bulletin Reprints, 7, 167–176.
Grima, S., & Görg, H. (2002). Foreign direct investment, spillovers and absorptive capacity: Evidence from
quantile regressions. Paper presented at the International Trade and Finance Association’s 12th International
Conference, May 29–June 2, 2002, Ramkhamhaeng University, Bangkok.
Ito, K. (2002a). Are foreign multinationals more efficient? Plant productivity in the Thai automobile industry.
Working Paper Series, vol. 2002-19, July, The International Centre for the Study of East Asian Development,
Kitakyushu.
Ito, K. (2002b). Foreign ownership and productivity in the Indonesian automobile industry: Evidence from
establishment data for 1990–1999. Paper presented at the NBER 13th Annual East Asian Seminar on
Economics, Productivity, June 20–22, 2002, Melbourne, Australia.
Jorgenson, D. W., & Griliches, Z. (1967). The explanation of productivity change. Review of Economic Studies,
34, 249–283. Reprinted in Jorgenson, D. W. (1995). Productivity: Postwar US economic growth, vol. 1.
Cambridge, MA: The MIT Press.
Koenker, R., & Bassett, G., Jr. (1978). Regression quantiles. Econometrica, 46(1), 33–50.
Markusen, J. R. (1991). The theory of the multinational enterprise: A common analytical framework. In: E. D.
Ramstetter (Ed.), Direct foreign investment in Asia’s developing economies and structural change in the
Asia-Pacific region (pp. 11–32). Boulder: Westview Press.
Maruhashi, H. (1995). Japanese subcontracting system in Thailand: A case study of the Thai automobile industry
(sic). Master’s Thesis, Faculty of Economics, Thammasat University, Bangkok.
National Statistical Office (1999). Report of the 1997 Industrial Census: Whole Kingdom. National Statistical
Office, Office of the Prime Minister, Bangkok.
National Statistical Office (2001). Report of the 1999 Industrial Survey: Whole Kingdom. National Statistical
Office, Office of the Prime Minister, Bangkok.
Okamoto, Y. (1999). Multinationals, production efficiency and spillover effects: The case of the US auto parts
industry. Weltwirtschaftliches Archiv, 135, 241–260.
Okamoto, Y., & Sjöholm, F. (2000). Productivity in the Indonesian automotive industry. ASEAN Economic
Bulletin, 17(1), 60–73.
Poapongsakorn, N., & Wangdee, C. (2000). The Thai automotive industry. ASEAN Automobile Report,
available at http://www.asean-auto.org.
K. Ito / Journal of Asian Economics 15 (2004) 321–353
353
Ramstetter, E. D. (2001). Labor productivity in foreign multinationals and local plants in Thai manufacturing,
1996 and 1998. Working Paper Series, vol. 2001-13, The International Centre for the Study of East Asian
Economic Development, Kitakyushu.
Ramstetter, E. D. (2002). Does technology differ in local plants and foreign multinationals in Thai
manufacturing? Evidence from translog production functions for 1996 and 1998. Working Paper Series, vol.
2002-04, The International Centre for the Study of East Asian Economic Development, Kitakyushu.
Shapiro, S. S., & Francia, R. S. (1972). An approximate analysis of variance test for normality. Journal of the
American Statistical Association, 67, 215–216.
Terdudomtham, T. (1997). The automobile industry in Thailand. Bangkok: Thailand Development Research
Institute.
Waseda Daigaku Shougakubu [Faculty of Commerce, Waseda University, Japan], & Zaidan-Hojin Keizai
Kouhou Center (Eds.) (1995). Jidosha Sangyo no Gurobaru Senryaku: Chosen kara Kyosei e [The global
strategy of the automobile industry: From challenge to cohabitation]. Tokyo: Chuo Keizai-sha.
White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for
heteroskedasticity. Econometrica, 48, 813–838.
Yahata, S., & Mizuno, J. (1988). Nikkei Shinshutsu Kigyo to Genchi Kigyo to no Kigyo-kan Bungyo Kozo to
Gijutsu Iten: Tai no Jidosha Sangyo wo Jirei to shite [The division of labor and technology transfer among
Japanese affiliated firms and local firms: A case study on Thai automobile industry]. Tokyo: Institute of
Developing Economies.