Segmentation and Spillovers in
the Chinese Semi-conductor
Industry
Henry Chesbrough, Helen Liang
Haas School of Business, UC Berkeley
Aug 18, 2007
1
Outline
• Motivation, Background and Previous Findings
of China’s Semiconductor Industry
• Research Questions
• Existing Literature and Hypotheses
• Data, Measurement, and Empirical Strategy
• Empirical Results (To be finished)
– Semiconductor industry
– Other industries
• Conclusion and Discussion
2
Background and Motivation
• Globalization of semiconductor industry
– Expansion of production and R&D to
emerging market
– MNCs compete and cooperate with local firms
• Knowledge transfer
– Channels and effects of knowledge
transfer/spillovers
– Changing landscape of intellectual property
protection regime in developing countries
3
Previous Findings
– A Divided Industry
There are two very different industry segments
operating in the Chinese semiconductor
industry (Chesbrough 2005)
1. A Globally-oriented, globally competitive segment
that employs advanced technologies and is gaining market
share in world markets
2. A Domestically-oriented segment
that employs backward technologies, lacks access to
investment capital and management expertise, where the
government is the largest shareholder of each firm
4
Knowledge spillovers between key
segments of the
Chinese semiconductor industry
MNCs
Domestically- Globally-focused
Focused Chinese firms Chinese firms
5
Research Objective
• Identify the size and trend of development of the
two domestic segments
• Examine the effects of MNCs on the innovation
and productivity of domestic segments
• Examine the moderating factors of learning and
knowledge spillovers
6
Research Question
1. Size and trend
– How big are the two segments of the Chinese semiconductor industry, relative to each other? Is
one growing faster than the other? How profitable are the two segments?
– Is there a trend of converging of the two segments within the Chinese industry, or are they
diverging? Are they increasingly targeting the same market/customer base (within China or
internationally) or not? Is the domestic oriented segment improving its technology, productivity,
and profitability, relative to the globally oriented segment?
2. Knowledge spillover
– What are the comparative rates of spillover between the MNCs and the globally-oriented firms
and the domestically-oriented firms?
– Do the globally focused firms have higher absorptive capacity, and learn more from the MNCs?
– How is semiconductor industry compared with other industries, tech-intensive vs labor-intensive,
domestic-oriented vs exporters? Textile, machinery, food processing, etc.
3. R&D efforts and output
– What determines the R&D intensity at firm level?
– How does R&D input translate into output?
– How does R&D activity influence firm productivity and profitability?
7
Outline
• Motivation, Background and Previous Findings
of China’s Semiconductor Industry
• Research Questions
• Existing Literature and Hypotheses
• Data, Measurement, and Empirical Strategy
• Empirical Results (To be finished)
– Semiconductor industry
– Other industries
• Conclusion and Discussion
8
Existing Literature I
• Channels of spillovers: the role of industrial linkage
– Within industry (+/-)
• Imitation/Tech transfer (+)
• Personnel turnover and interaction (+/-)
• Negative externalities not related to tech spillovers (-) : input, labor,
land, etc
• Empirical findings on intra-industry spillovers:
– + in developed economies (U.S. and U.K.)
– 0/- in developing economies (Aikton& Harrison99, Javorcik04)
– Downstream Upstream industries (buyer supplier) (+)
• Customers provide support to suppliers: technology, mgmt skills
(Javorcik04)
• Export (Blalock and Gertler 04 & 05)
– Upstream Downstream industries (supplier buyer) (+)
• Better inputs and equipment higher productivity
9
Existing Literature II
• Moderating factors of spillovers
– Absorptive capacity
• Firms’ previous investment/experience in R&D and
personnel/skilled workers (Cohen&Levinthal, Blalock&Gertler)
– Ownership structure of
• knowledge source: wholly-owned MNC subsidiaries vs Joint
ventures
• knowledge recipient: incentive structure in state-owned vs
private firms
– Geographic distance to knowledge source
• Knowledge transfer is more effective at local level
– Jaffe et al. 93: patent citations in U.S.
– Keller 02: productivity benefit from other countries’ R&D
– Technology Gap between source and recipient
10
Existing Literature III
• R&D intensity and output
– R&D production function (Grilliches 1979)
– Chinese firms’ R&D Performance (Jefferson et al,
2002)
• A recursive 3-equation system:
– R&D intensity = previous performance
– R&D output = previous R&D intensity
– Performance/productivity = R&D output + L + K + M
11
Hypotheses
12
Outline
• Motivation, Background and Previous Findings
of China’s Semiconductor Industry
• Research Questions
• Existing Literature and Hypotheses
• Data, Measurement, and Empirical Strategy
• Empirical Results (To be finished)
– Semiconductor industry
– Other industries
• Conclusion and Discussion
13
Data
• Firm level data:
– Firms operating in China, Semiconductor
industry and other industries, 1998-2005
– Source: Enterprise Survey, National Bureau
of Statistics, China
• Variables:
– Input/output, R&D activities, patent
applications, new products, etc
14
Descriptive Statistics
• Size of the two segments vs. MNCs in semiconductor industry
(csic=415, 2002 version)
– Output (v207, current price), employment (v210)
– Export Ratio (v213/v209) to identify domestic-oriented and global-
oriented segments (ownership b10=110-190)
• Global-oriented segment: export ratio >=50%
• Domestic-oriented segment: export ratio <50%
– Compare with MNCs (ownership b10=310-340) and overseas Chinese-
invested enterprises (b10=210-240)
– Compare with other industries
• Trend of development:
– Growth rate of output, export ratio, and innovation output and input
– Are the two segments following the same trend?
• 2-sample t-test on firm level average growth rate 1998-2005
• Graph: average growth rate of each segment
15
Measurement
• Productivity and profitability
– TFP and labor productivity
– ROA
• Innovation activity
– Input of innovation: R&D funding (kj22), expenditure (kj28), personnel
(kj43), scientist and engineer (kj46)
– Output of innovation: # of patent application (kj77), patents granted
(kj79), # of S&T projects (kj85), R&D projects (kj87), new product value
(and as a proportion of output), new product value and export value
• Other moderating factors:
– Absorptive capacity: previous cumulated R&D investment and
personnel?
– Geographic distance to knowledge source
• Same city/province MNCs/global-oriented/domestic-oriented share of
output/labor in own industry sector and vertically related sectors
16
Model I: Productivity Spillovers
• A Translog Productivity Function with firm
fixed effects (Liang 2007)
ln Yijrt = α+ß1*lnKijrt + ß2*lnLijrt + ß3*lnMijrt
+ ß4*Horizontal_FDIjrt + ß5*Downstream_FDIjrt + ß6*
Upstream_FDIjrt
+ ß7*Controlijrt + αi + αt + ξijrt
17
Model I: Independent Variables --
Industrial Linkage
• Horizontal FDI share (+/-):
– The share of sector output by foreign affiliates
• Downstream FDI share (+):
– The share of sector output by foreign affiliates in the downstream
industries weighted by Input/Output coefficients, excluding own-
sector
• Upstream FDI share (+):
– Similar as downstream FDI share, excluding export in output
• Note:
– Input/Output coefficients at 2-digit China SIC level
– Obs. at 4-digit China SIC level
18
Model I: Control Variables
• Fixed effects: sector fixed effects, year dummies
• Export ratio and export-oriented sector dummy
• Olley-Pakes proxy for un-observed productivity
shocks
– Concern: FDI goes to locations with quick learners
– Use investment conditional on capital to identify un-observed
contemporaneous shocks
– A third order polynomial of investment and capital
19
Model I: Moderating Factors
• Moderating factors:
– Ownership of knowledge source and recipient
• Estimate wholly-owned subsidiaries and joint ventures’
impact separately
• Estimate different ownership of domestic segments
– Distance to knowledge source
• Compare city-level coefficients with province-level
coefficients (Liang 2007)
– Absorptive Capacity
• Compare/control firm-level R&D expenditure/personnel
• Proxy Tech gap with cumulative R&D expenditure
(alternatives: patent applications, # or projects) (Blalock and
Gertler)
20
Model I: Industry Analysis
• Does Semiconductor industry look like
other industries?
– Compare semiconductor industry result with
other industries
– Compare tech-intensive with labor-intensive
industries
21
Model II: Innovation Activities
• A Recursive 3-equation system (Jefferson
et al 2002) (this function form need to be thought through)
– R&D intensity = previous performance
– R&D output = previous R&D intensity
– Performance/productivity = R&D output + L + K + M
– R&D output: to be measured with patent application, new
product value, and export value.
22
Empirical Strategy
• Identification
– MNCs and entrants’ selection of industry and
location
• No good instruments so far
• Geographic distance absorbed in FE
• Dynamic panel? Using lagged dependent variables
as instruments
– Survival Bias
• Using balanced panel of firms introduce survival
bias
• Inverse Mills Ratio? Propensity Score?
23
Overview of the two segments in
China’s semiconductor industry
• Size
– The global-oriented segment has outgrown domestic-oriented segment in
number of firms, output, employment, and asset 1998-2005
• Productivity, Profitability, R&D, and patent activity
– Labor productivity (value-added/employment) in the two segments improve by
similar pace
– Domestic segment has higher R&D intensity in terms of expenditure and
personnel
– But Global segment has more patent application and granted patents
– Entrants in global segment are less profitable, while the opposite is true in
domestic segment
• Export
– The share of global segment in export increases from 43% to 92% 1998-2005
• Firm Age
– Firms are younger in global segment, and older in domestic segment
– Overtime firm age decreases in both segments – indicating exit of older firms and
entry
• Ownership composition
– Domestic segment is dominated by domestic-owned firms and global segment by
foreign-owned firms; in both segments domestic ownership declines overtime
24
Size of the two segments in China’s
semiconductor industry – Number of Firms 1998-
2005
Two Segments of China's Semiconductor Industry Two Segments of China's Semiconductor
1998-2005: Total Number of Firms Industry 1998-2005: Ratio of Firms
1.00
200
Total Number of Firms
0.90
0.80
150
Ratio of Firms
0.70
0.60
100 0.50
0.40
50 0.30
0.20
0.10
0
0.00
1998 1999 2000 2001 2002 2003 2004 2005 1998 1999 2000 2001 2002 2003 2004 2005
Year Year
Number of Firms -- Domestic Number of Firms -- Global Number of Firms -- Domestic Number of Firms -- Global
25
Size of the two segments in China’s
semiconductor industry -- output 1998-2005
Two Segments of China's Semiconductor Industry Two Segments of China's Semiconductor
1998-2005: Total Output (in Billion Yuan) Industry 1998-2005: Ratio of Output
1.00
200
0.90
0.80
Total Output
Ratio of Output
150 0.70
0.60
100 0.50
0.40
50 0.30
0.20
0.10
0
0.00
1998 1999 2000 2001 2002 2003 2004 2005 1998 1999 2000 2001 2002 2003 2004 2005
Year Year
Output -- Domestic Output -- Global Output Ratio -- Domestic Output Ratio -- Global
26
Size of the two segments in China’s
semiconductor industry -- Employment 1998-2005
Two Segments of China's Semiconductor
Two Segments of China's Semiconductor Industry
Industry 1998-2005: Ratio of Employment
1998-2005: Total Employment (in persons)
1.00
300,000 0.90
Ratio of Employment
Total Employment
250,000 0.80
0.70
200,000 0.60
150,000 0.50
0.40
100,000 0.30
50,000 0.20
0.10
0
0.00
1998 1999 2000 2001 2002 2003 2004 2005 1998 1999 2000 2001 2002 2003 2004 2005
Year Year
Employment -- Domestic Employment -- Global Employment Ratio-- Domestic Employment Ratio-- Global
27
Size of the two segments in China’s
semiconductor industry – Fixed Asset 1998-2005
Two Segments of China's Semiconductor
Two Segments of China's Semiconductor Industry
Industry 1998-2005: Ratio of Fixed Asset
1998-2005: Total Fixed Asset (in Billion Yuan)
1.00
90
0.90
80
Ratio of Fixed Asset
0.80
Total Fixed Asset
70
0.70
60
0.60
50
0.50
40
0.40
0.30 30
0.20 20
0.10 10
0.00 0
1998 1999 2000 2001 2002 2003 2004 2005 1998 1999 2000 2001 2002 2003 2004 2005
Year Year
Fixed Asset Ratio -- Domestic Fixed Asset Ratio -- Global Fixed Asset -- Domestic Fixed Asset -- Global
28
Productivity
Labor productivity (value-added/employment) in the two
segments improve by similar pace
Two Segments of China's Semiconductor Industry
1998-2005: Labor Productivity
(Value-added/em ploym ent, 1,000 Yuan/person, 1998 price)
200
Labor Productivity
150
100
50
0
1998 1999 2000 2001 2002 2003 2004 2005
Year
Labor Productivity -- Domestic Labor Productivity -- Global
29
R&D and patent activity
Domestic segment has higher R&D intensity in terms of
expenditure and personnel
Two Segments of China's Semiconductor Industry
1998-2005: R&D Expenditure/Sales, R&D
Staff/Employment
0.08
0.06
0.04
0.02
0.00
1998 1999 2000 2001 2002 2003 2004 2005
Year
R&D Expenditure/Sales -- Domestic R&D Staff/Employment -- Domestic
R&D Expenditure/Sales -- Global R&D Staff/Employment -- Global
30
R&D and patent activity
Global segment has more patent application and granted
patents
Two Segments of China's Semiconductor Industry
1998-2005: # of Patent Application and Granted
3.00
2.00
1.00
0.00
1998 1999 2000 2001 2002 2003 2004 2005
Year
Patent Application (invention) -- Domestic Patent Granted -- Domestic
Patent Application (invention) -- Global Patent Granted -- Global
31
Profitability
Entrants in global segment are less profitable, while the
opposite is true in domestic segment; profit plunged in
2001
Two Segments of China's Semiconductor Industry 1998- Two Segments of China's Semiconductor
2005: Ratio of Profit
Industry 1998-2005: Total Profit (in Billion Yuan)
1.00
6
0.80
0.60
Ratio of Profit
4
Total Profit
0.40
0.20 2
0.00
-0.20 1998 1999 2000 2001 2002 2003 2004 2005 0
-0.40 1998 1999 2000 2001 2002 2003 2004 2005
-0.60 -2
Year Year
Profit -- Domestic Profit -- Global Profit -- Domestic Profit -- Global
32
Profitability
Entrants in global segment are less profitable, while the
opposite is true in domestic segment;
exporting domestic firms are more profitable.
Profitable firm vs loss-making firms in the two Profitable firm vs loss-making firms in the two
segments segments -- Firm Age
Firm-year average (in million Yuan, persons)
25
1400 20
1200 15
1000 10
800 5
600
0
400 Firm Age
200
0
Number of firms Output Employment Export Profit
-200
Global -- profit Global -- loss Domestic -- profit Domestic -- loss
33
Profitability
Compare profiting and loss-making firms
Firm-Year Average (in million Yuan, persons)
Global -- Global -- Domestic - Domestic -
profit loss - profit - loss
Number of firm-year 482 201 705 243
Firm Age 8.75 5.72 16.41 21.19
Output 759.56 506.24 589.39 244.63
Employment 1,286.61 943.17 1,279.82 1,031.42
Export 665.83 426.00 105.68 40.36
Profit 48.28 -59.44 50.06 -35.99
Return on Asset 0.08 -0.07 0.06 -0.05
Export Ratio 0.87 0.89 0.12 0.09
34
Export
The share of global segment in export increases from 43%
to 92% 1998-2005
Two Segments of China's Semiconductor Industry 1998- Two Segments of China's Semiconductor
2005: Ratio of Export
Industry 1998-2005: Total Export (in Billion Yuan)
1.00
200
Ratio of Export
0.80
Total Export
150
0.60
100
0.40
50
0.20
0
0.00
1998 1999 2000 2001 2002 2003 2004 2005
1998 1999 2000 2001 2002 2003 2004 2005
Year
Year
Export -- Domestic Export -- Global
Export -- Domestic Export -- Global
35
Firm Age
Firms in global segment are younger, and in domestic
segment older; overtime firm age decreases in both
segments – indicating exit of older firms and entry
Two Segments of China's Semiconductor Industry 1998-2005:
Firm Age
30
Average Age of Firms
25
20
15
10
5
0
1998 1999 2000 2001 2002 2003 2004 2005
Year
Firm Age -- Domestic Firm Age -- Global
36
Ownership composition
Domestic segment dominated by domestic-owned firms
Global segment by foreign-owned firms
In both segments domestic ownership declines overtime
Two Segments of China's Semiconductor
Industry 1998-2005: Ownership Composition
1.00
0.80
0.60
0.40
0.20
0.00
1998 1999 2000 2001 2002 2003 2004 2005
Year
DomesticFirms -- Domestic OverseasChinese -- Domestic ForeignFirms -- Domestic
DomesticFirms -- Global OverseasChinese -- Global ForeignFirms -- Global
37