Comparing Productivity Impacts of Knowledge Spillovers
from Network and Arm’s Length Industries :
The Case of Business Groups in Korea
Keun LEE* 이근
and
Kineung CHOO** 추기능
*Professor of economics, Seoul National University,
e-mail: [email protected]
** Assistant Professor; Korea Naval Academy, Changwon, Korea.
Background
▪ Firms located in a spatial/conceptual neighborhood benefit from
other firms’ knowledge-creating activities
 Knowledge spillovers -> Productivity or performance
▪ Tacitness of knowledge tends to restrict knowledge flows in terms of
its transferability to, and learning, by other firms.
 Spillover and transfer among sister firms affiliated with the same
business group or conglomerate might be less subject to such
limitations.
=> bigger impact than in arm’s length relationship
Background
▪ However,
few studies analyze the spillover impacts of a knowledge pool of a
‘network’ consisting of affiliates in a business group, and
compare them with spillovers from arm’s length firms
 We deal with a new question of the relative size of spillovers from
networks vs. industries,
and also revisit the debate on the size of
intra- vs. inter-sector spillovers,
dividing the knowledge pool into those within and outside a sector
The Literature
▪ Knowledge accumulated by a firm tends to deepen and broaden other firms'
technology bases, without appropriate compensation and exhaustion (Glaeser
et al., 1992; Laursen & Meliciani, 2000).
▪ A firm’s production function depends on the level of knowledge
available in the economy, as well as on its own inputs (Jaffe, 1986; Medda
and Piga, 2007).
▪ Labor mobility of scientists and engineers is a conduit for knowledge
spillovers among firms (Dindaroglu, 2010; Kim and Marschke, 2005).
Higher mobility of skilled labor between affiliates may facilitate transfers
of knowledge better within the business group than within the market.
▪ A resource-based view of the firm stresses the sharing of resources among
affiliates within the group (Chang & Hong, 2000).
 By pooling resources at the business group level, and then sharing them
among affiliates, more efficient resource utilization is possible.
The Literature 2
▪ Studies have addressed the issue of the relative size of inter- and
intra-sector spillovers; for example, Bernstein (1988), Rouvinen
(2002), and Kafouros and Buckley (2008).
▪ However, no previous studies focus on the effects of inter- and
intra-sector spillovers within a business group, and compare
spillovers from affiliates in the different sectors and spillovers from
affiliates in the same sector.
Therefore,
▪ we propose that knowledge pools at the group level affect affiliates’
innovations and thus their productivities.
▪ Furthermore, influence of the spillover pool from the network may
be stronger than those from the market.
▪ Then, we compare the spillovers from the same and different sectors
within a business group.
Four Sources of Spillover Pools and Comparison Combinations
Part One: Four Sources of Spillover Pools
Network
Arm's Length Industry
Intra-sector
A: An affiliated firm in the same sector
C: Arm’s length firms in the same sector
Inter-Sector
B: An affiliated firm in the different sector
D: Arm’s length firms in the different sector
Part Two: Four Main Comparisons
Across sectors Within network
Spillover from A is bigger or smaller than B
Across boundaries within a sector
Spillover from A is bigger or smaller than C
Across sectors within Industry
Spillover from C is bigger or smaller than D
Across boundaries from different sectors
Spillover from B is bigger or smaller than D
Part Three: Two More Possible Comparisons
networked firms in same sector
Spillover from A is bigger or smaller than D
vs. other firms in different sector
networked firms in different sector
vs. other firms in same sector
Spillover from B is bigger or smaller than C
Data
▪ financial data for the Korean firms
▪ patent applications filed with the Korean Intellectual Property Office (KIPO) from
1989 to 1997.
▪ 79 groups and 417 firms were included in the sample. The number of firm-year
observations is 2,242 in the sample.
- The group definition was obtained from the study of Lee et al. (2007).
- broader than the typical top 30 chaebols that the Korea Fair Trade Commission
(FTC) had designated and monitored.
- to be included in the sample, the group needs to have more than two affiliates each
year during the seven-year sample period, 1991-1997.
- should have more than two patent applications, and continue to be in the KIS data
set all throughout the period, 1991-1997.
Variable Construction
▪ We sums up the number of patent which were applied for in periods T-2, T-1,
and T to obtain a proxy for a firm’s knowledge base for the year T.
▪ Then, concerned explanatory variables such as group and industry patents are
calculated
▪ We construct four variables representing additional knowledge bases of a firm
besides the firm's own patents based on these cumulative values.
▪ inter-sector-within-group spillover Group _ Spilloverint er   Pgj
j i
: group _patent(out)
▪ intra-sector-within-group spillover pool
Group _ Spilloverint ra  Pgi  pk
: group _patent(in)
▪ inter-sector spillover pool
: industry_patent(out)
▪ intra-sector spillover pool
: industry_patent(in)
Industry _ Spilloverint er   Pj  Group _ Spilloverint er
j i
Industry _ Spilloverint ra  Pi  Group _ Spilloverint ra
Summary of variables used in spillover estimation
Variables
sales_per_employee(million won)
firm patent
industry_patent(in)
industry_patent(out)
group_patent(in)
group_patent(out)
group_patent(in+out)
export_ratio(%)
market_share(%)
firm age
Mean
232
156
5,593
67,790
486
1,974
2,460
18.4
5.8
18.3
Std. Dev.
172
1,503
13,310
40,885
2,764
6,156
6,873
27.5
12.0
13.3
Min
37
0
0
6,954
0
0
0
0.0
0.0
0.9
Max
1,682
39,326
88,690
153,000
42,358
45,783
45,783
99.6
95.0
74.0
Obs
2,242
2,242
2,242
2,242
2,242
2,242
2,242
2,242
2,242
2,242
Regression Models
(1)
productivity    1  firm _ patent   2  industry _ patent (in )
 3 * industry _ patent (out )    Z  i  uit
(2)
productivity    1  firm _ patent   2  industry _ patent(in )
 3  industry _ patent(out )   4  group _ patent(in  out )    Z  i  uit
(3)
productivity    1  firm _ patent   2  industry _ patent (in )
  3  industry _ patent (out )   5  group _ patent (in )
  6  group _ patent (out )    Z  i  uit
Comparing the Impacts of the Intra- and Inter- sector Spillovers
variables / models
coef.
cons.
t(z)-value
coef.
firm patent
industry_
patent(in)
industry_
patent(out)
t(z)-value
model 1
RE
240.27
7.67
***
52.92
3.03
***
coef.
t(z)-value
model 2
RE
230.73
7.35
***
28.99
1.65
*
15.68
7.17
***
coef.
t(z)-value
coef.
export ratio
t(z)-value
coef.
market share
t(z)-value
coef.
firm age
t(z)-value
industry dummies
within
0.39
3.07
***
2.40
3.70
***
-10.02
-1.41
0.39
3.16
***
2.56
3.96
***
-7.52
-1.05
yes
0.015
yes
0.042
model 3
FE
134.30
17.62
***
39.19
2.41
**
7.90
20.54
***
0.28
2.25
**
6.60
5.83
***
model 4
FE
133.12
17.48
***
29.08
1.76
*
6.33
3.02
***
7.62
19.25
***
0.28
2.30
**
6.53
5.79
***
no
0.201
no
0.205
Comparing the Impacts of the Spillovers from the Network and the Industry
variables / models
coef.
cons.
t(z)-value
coe f.
firm pate nt
t(z)-value
coe f.
group_
pate nt(in+out) t(z)-value
industry_
pate nt(in)
industry_
pate nt(out)
coe f.
t(z)-value
model 1
RE
222.20
7.18
***
23.28
1.35
35.11
8.27
***
12.16
5.54
***
model 2
FE
133.53
17.61
***
33.56
2.07
**
19.53
4.51
***
t(z)-value
coe f.
industry_
pate nt(in+out) t(z)-value
export ratio
t(z)-value
coef.
market share
t(z)-value
coef.
firm age
industry dummies
t(z)-value
model 4
FE
132.70
17.50
***
26.31
1.59
17.21
3.95
***
17.80
4.05
***
4.86
2.29
**
7.27
18.02
***
7.45
18.81
***
coe f.
coef.
model 3
FE
132.88
17.53
***
23.05
1.42
0.36
2.93
***
2.27
3.56
***
-4.84
-0.69
0.26
2.16
**
6.49
5.77
***
7.11
18.94
***
0.27
2.23
**
6.41
5.71
***
yes
no
no
0.27
2.20
**
6.45
5.74
***
no
Spillovers from the Affiliates in the Same and Different Sectors
variables / models
coef.
cons.
t(z)-value
coef.
firm patent
industry_
patent(in)
industry_
patent(out)
group_
patent(in)
group_
patent(out)
t(z)-value
coef.
t(z)-value
coef.
coef.
t(z)-value
t(z)-value
coef.
market share
t(z)-value
model 3
RE
221.40
7.16
***
22.44
1.30
0.40
3.18
***
2.60
4.02
***
model 4
FE
132.91
17.51
***
33.85
2.09
**
12.67
5.62
***
30.37
3.22
***
coef.
t(z)-value
model 2
FE
134.48
17.69
***
35.46
2.18
**
7.61
19.26
***
t(z)-value
coef.
export ratio
model 1
RE
231.35
7.36
***
30.31
1.73
*
13.54
5.94
***
16.20
3.16
***
0.26
2.15
**
6.41
5.68
***
26.94
2.89
***
37.50
7.67
***
0.36
2.91
***
2.23
3.50
***
7.46
18.85
***
31.01
3.54
***
15.43
3.02
***
0.27
2.20
**
6.61
5.86
***
model 5
FE
132.39
17.44
***
27.30
1.65
*
4.31
1.96
**
7.30
18.04
***
25.36
2.75
***
15.36
3.01
***
0.27
2.22
**
6.53
5.80
***
Hypothesis test for the difference between the estimated coefficients
Hypothesis tested
industry patent(in)=industry patent(out)
The model 4
in table 7
F value
(Prob.>F)
The model 4
in table 8
F value
(Prob.>F)
The model 5
in table 9
F value
(Prob.>F)
0.33(0.563)
1.17(0.280)
1.66(0.197)
industry patent(in) =group patent(in)
4.34(0.038)**
industry patent(in)=group patent(out)
3.93(0.048)**
industry patent(in)=group patent(in+out)
6.21(0.013)**
industry patent(out)=group patent(in)
3.83(0.050)**
industry patent(out)=group patent(out)
2.39(0.122)+
industry patent(out)=group patent(in+out)
5.49(0.019)**
group patent(in)=group patent(out)
industry patent(in+out)=group
patent(in+out) (from model 3 in table8)
0.87(0.350)
5.09(0.024)**
The magnitude of the spillover estimated from the regressions
▪ One more patent applied for by sister firms in the same sector bring about 6.58
dollars increase in sales per employee for the firm affiliated to the same business
group.
=> 11,534 increase in sales
▪ A firm would experience a 1.12 dollars increase in labor productivity from one
patent by an unrelated firm in the same sector.
=>1,960 dollars increase in sales
▪ One more patent of the affiliated firms doing business in other sectors tend to
increases my firm’s sales per employee by 3.99 dollars.
=> 6,986 dollars increase in sales
▪ Additional patent in whatever an unrelated firm in other sectors produces 1.89
dollars increase in labor productivity of the concerned firm.
=> 3,320 dollars increase in sales
Regression Results
▪ Our results provide no evidence for dominance of either intra- or
inter-sector spillovers.
 This would be a support for the arguments that both intra- and inter- industry
spillovers matter (Hubert and Pain, 2001; Medda and Piga, 2007)
▪ The impacts of network-based spillovers are stronger than those of
arm’s length relationship-based spillovers.
 This finding implies that group-affiliated firms have additional
benefits from their sister firms in the viewpoint of spillovers.
We have the consistent results of bigger impacts from the network
than from arm’s length industry either in terms of intra- or intersector impacts.
Concluding Remarks : From Spillovers to Productivity
▪ Spillovers from the network are greater than those from arm’s
length industry
This should be a new contribution because the literature tends to focus on
spillover from arm’s length industry in general.
A business group is an effective organization to internalize knowledge
spillovers or to promote greater knowledge diffusion between affiliates.
 Owing to this benefit, affiliated firms would find it more possible to
broaden their technological capabilities and to achieve higher level of
productivity, other things being equal.
▪ There is no evidence for dominance of either intra- or inter-sector
spillovers, regardless of whether it is from arm’s length industry or
from networks.
 Both matters.
Thank you!