large diversified firms

Industrial and Corporate Change October 2003
Innovation, Technological Regimes and
Organizational Selection in Industry Evolution:
A "History Friendly Model" of the DRAM industry
Chang-Wook Kim* and Keun Lee**
* Samsung Economic Research Institute, Seoul, Korea. E-mail: [email protected]
** Economics Department, Seoul National University, Seoul 151-742 Korea.
E-mail: [email protected]. Fax 822-886-4231
1
1. Questions :
1)
What determines the evolution of an industry in terms of
market structure, distribution of the firms with entry and
exit, and selection among the heterogeneous organizational
forms of the firms?
2)
What would be the impact of technological innovation and
technological regimes of an industry on the evolution of
industry?
2
2. Purpose
** From appreciative theory (causal explanations of observed
patterns of economic phenomena) to 'history-friendly'
evolutionary economic modeling using a simulation method :
(Eg. long term evolution of the computer industry in Marlerba,
Nelson, Orsenigo, & Winter (1999)
** Purpose : a history-friendly model for the DRAM (dynamic
random access memory) chip industry; not only to replicate the
evolution of the industry but to analyze the complex relationship
among innovation, technological regimes, and selection of the
firms of different organizational forms in the evolution of an
industry.
3
3. Story of D-Ram :
From early entry and dominance by small specialized firms
to late entry and dominance by large diversified firms
Table 1. Classification of the Firms in DRAM Industry
Small Specialized Firms
AMD, Fairchild, Inmos, Intel, Intesil, Micron
Technology*, Signetics, Mostek, Zilog, National
Semiconductor*, STC(ITT), Eurotechnique*,
NMB**, Vitelic, Vangard
Large Diversified Firms
AT&T, IBM, Motorola, TI, Siemens, SGS-Ates,
Fujitsu, Hitachi, Matsushita, Mitsubishi, NEC,
Oki, Sanyo, Sharp, Toshiba, NSS (Nippon Steel
Semiconductor)**, Samsung, Hyundai, LG
*In what follows, for convenience, they are named as Micron, NS and Euro respectively.
**NMB started as a specialized firm, but merged into Nippon Steel in 1993.
4
Table 2. The Top 7’s in DRAM Industry
1975
1978
1981
1984
1987
1990
1993
1995
1
Intel
Mostek
Mostek
Hitachi
Toshiba
Toshiba
Samsung
Samsung
2
TI
TI
Fujitsu
NEC
NEC
Samsung
Hitachi
NEC
3
Mostek
NEC
NEC
Fujitsu
Mitsubishi
NEC
Toshiba
Hitachi
4
NEC
Intel
Hitachi
TI
TI
TI
NEC
Hyundai
5
Motorola
Motorola
TI
Mitsubishi
Hitachi
Hitachi
IBM
TI
6
Fairchild
Fujitsu
NS
Mostek
Fujitsu
Fujitsu
TI
Toshiba
7
NS
Hitachi
Motorola
Motorola
Samsung
Mitsubishi
Mitsubishi
LG
4
2
1
1
0
0
0
0
3
5
6
6
7
7
7
7
SS
Firms
LD
Firms
Source: Dataquest, DRAM Market Statistics, various years
5
Table 3. Trend of Entries and Exits in the D-RAM Industry
74
75
76
Total
3
9
11
SS
Firms
2
6
LD
Firms
1
Entries
Exits
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
13
16
16
17
18
20
21
22
21
19
18
19
20
19
19
17
19
19
20
7
8
8
7
7
7
9
9
8
8
7
6
5
5
4
4
2
2
2
3
3
4
5
8
9
10
11
11
12
14
13
12
12
14
15
15
15
15
17
17
17
0
6
2
2
4
1
1
2
2
1
2
1
1
0
2
1
0
0
0
2
0
1
0
0
0
0
1
1
0
1
0
0
1
2
3
1
1
0
1
0
2
0
0
0
Source: Dataquest, DRAM Market Statistics, various years
6
Figure 1. Life Spans of D-RAM Firms
74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
AMD
Fairchild
Inmos
Intel
Intersil
Micron
Mostek
NS
Signetics
STC(ITT)
Zilog
Euro
NMB
Vitelic
Vangard
AT&T
IBM
Motorola
TI
SGS-Ates
Siemens
Fujitsu
Hitachi
Matsushita
Mitsubishi
NEC
Oki
Sanyo
Sharp
Toshiba
NSS
Samsung
Hyundai
LG
source: Dataquest, DRAM Market Statistics, various years
7
Table 4. Average Capital and R&D Expenditure in the10
DRAM Firms ( Top 5 in the US and Top 5 in Japan)
Capital
R&D
Expenditure Expenditure
1979
1980
1981
1982
1983
1984
Capital
R&D
Expenditure Expenditure
1985
463
629
454
1986
651
681
992
2027
531
629
767
950
1987
1988
1989
1600
955
1213
1288
1230
1813
-
1338
1599
1791
Source: Meth'e (1991) p. 57-58
8
Organizational Difference between SS (small
specialized firms) vs. LD (large diversified firms) :
The LD firms tend to carry higher overhead costs per unit of
capital or output than the SS Firms:
unless the output reaches substantial amount, they cannot
expect any cost advantage over the SS firms.
9
Technological regime of the D-RAM Industry
1) Cumulativeness (degree that the productivity with the current
generation is related to the preceding generation chips): Low
2) Impact of the process innovation on productivity increase: high
Productivity in DRAM industry = number of chips per wafer as
well as the yield rate (the proportion of qualified chips)
; adopting a newer production facilities
-- substantial increase in the number of chips per wafer and the
yield rates.
10
4. Literature Review :
(1) Process Innovation Models
1) Flaherty(1980): Process Innovation  scale expansion and concentration
2) Klepper(1996):both process and product innovation + life cycle: similar results
3) Limits: homogeneous firms
(2) Cumulativeness Models
1) Nelson and Winter (1982):
Science-based regimes: Potential product determined by science
Cumulative regimes: Potential product determined by past product
; tend to be more concentrated
2) Winter (1996): introduced entry and exit; adjustment process for innovative and
imitative R&D expenditure
new product = weighted sum of present and future product
; weights = a measure of cumulativeness
similar results: increased concentration and decreasing no. of entry
11
3) Limit: still not allow heterogeneity of the firms
(3) Organizational forms models
1) Swann and Gill (1993):
complex conglomerates vs. Small specialized firms
production costs vs. adjustment costs
variations in tech competence ( eg. processing speed )
more variations = unpredictableness
--> bad for conglomerates
2) Critic: more variations anything relating to performance
--> bad for conglomerates
12
5. Model: Main feature
: cumulativeness + degree of productivity increase  on organizational selections
1) Firms' Production and Investment Behavior
Q t = i Q it
Pt = min [R/Qt, P max ] , where P max is the upper bound of price
Q it = hit  Kit
it = Pt hi t – ACK it – r it n - r it m
m* =  / ( - s i)
mit = Pt / (ACK i /hit)
Target Investment I*it = (  + ( (m it - m*it ) / m it )) Kit
if Pt  P max and s i  1
(8) I it = max [ 0, min(I*it, S + F) ]
(9) F = fit Kit
if i = SS
= fit Kit + Fl
if i = LD
(10) K i t+1 = ( 1 -  )K i t + Ii t
(1)
(2)
(3)
(4)
(5)
(6)
(7)
13
2) Heterogeneous Organizational Forms and Cost Differences
(average) variable costs (production costs)
: conglomerates < SMSE
(average) fixed (indirect ) costs : conglomerates > SMSE
(11) C i = VC i + FC i = VCK iKit + FC i
(where i = either L or S in terms of type)
(12) VCKL < VCK s and FCL > FC s .
(13) ACK i = VCK i + FC I / K i
(14) ACKL > ACK s
At same productivity levels,
AVC of conglomerates > AVC of SMSE
14
3) Innovation, Imitation and Cumulativeness
(15) Pr [ n = 1 ] = min [ an r n it Kit , 1 ]
(16) Pr [ m = 1 ] = min [ am r m It Kit , 1 ]
(17) productivity after innovation ; h nit = ( 1 +  ) hit
(where  = innovation impact parameter)
(18) productivity after imitation ; h m it = hit + ( 1 -  ) ht m
(where parameter  = cumulativeness)
(19) actual productivity in the next period ;
hi t+1 = max [hit , h n it , h m it ]
(20) X it = X it-1 + (1 - ) it ,
0  1
If X it < t , firms change their the innovation and imitation
activity parameters, r n it and r m with probability of 0.5:
(21) r n it+1 = ( 1 - ) r nit + r n i + n,
n  N ( 0, r n )
(22) r m it+1 = ( 1 - ) r m it + r mi + m
m  N ( 0, r m )
15
4) Exit and Entry
(23) Q i = hi  K i = 0 for all   t+1,
if X it < X min or Kit < K min
Entry by Imitation : (24) IM  P [ am E m ]
Exit Conditions :
(where total of external imitation expenditure = E m,
actual number of successful imitation cases = IM )
(25) he it = ht + ( 1 -  )ht m
(where ht is basic productivity of the industry
,htm productivity of target imitation firms)
Only when the expected average of the entry firms is lower
than the industrial average, actual entry will happen.
16
Figure 2A. Trends of Productivity and Unit Costs (Case 1)
(ρ = 0.00625 productivity jump, α = 0.05 cumulativeness)
(b) Trends of Unit Costs
(a) Trends of Productivity
0.18
Large Diversified Firm
0.175
Large Diversified Firm
1.1
0.17
1.05
0.165
0.16
10
20
30
40
0.95
0.155
10
20
30
40
0.9
17
Figure 3A. Trends of Productivity and Unit Costs (Case 2)
(ρ = 0.01, α = 0.05)
(b) Trends of Unit Costs
(a) Trends of Productivities
0.18
Diversified
LargeLarge
Diversified
FirmFirm
Large Diversified Firm
0.175
0.17
1.1
Large Diversified Firm
1.05
0.165
10
0.16
20
30
40
Large Diversified Firm
0.95
0.155
10
20
30
40
0.9
Large Diversified Firm
18
Figure 4: Entry by the LD firms at Different Combination
of ρ and α .
Case 1 : ρ = 0.00625, α = 0.05
Case 2: : ρ = 0.01, α = 0.05
1.05
25
25
0.95
1.1
1.1
1.1
50
75
100
125
150
Case 3: ρ = 0.01, α = 0.25
50
75
100
125
25
150
0.9
0.9
0.8
0.8
0.7
0.9
0.6
0.85
50
75
100
125
150
0.7
0.6
0.5
0.8
0.5
0.4
Solid line: the industrial average of the average unit cost of the Incumbent firms
Dotted line: the average unit cost of the potential entry firm (LD).
19
Table 5: Time Period when the AC of the LD firm getting below
that of the SS firms
ρ: productivity jump after innovation
α
0.00625
0.0075
0.01
0.015
0.025
0
46.9(36.9)
42.9(34.6)
28.8(27.1)
27.0(27.7)
19.4(21.6)
0.05
46.0(31.4)
37.2(29.2)
30.9(26.6)
28.0(31.1)
20.6(33.0)
0.1
49.1(37.2)
45.3(38.0)
29.6(28.8)
26.4(30.5)
19.9(26.0)
0.25
53.0(37.0)
43.5(36.6)
34.9(30.9)
29.4(28.4)
20.2(30.1)
Notes: The numbers in each cell indicate the average time periods after 160 simulation
runs, and the numbers in parentheses, standard deviations of the average. We run a large
number of runs due to the often large standard deviations .
20
Table 6: The average number of the newly entered LD firms
in 10 simulation runs per each case
ρ: productivity jump after innovation
α
0.00625
0.0075
0.01
0.015
0.025
0
1.6(1.35)
2.4(1.35)
2.5(1.20)
3.1(0.74)
3.2(0.42)
0.05
1.4(1.26)
2.0(1.33)
2.4(0.97)
2.5(1.08)
2.9(0.31)
0.1
1.3(1.06)
1.5(0.85)
2.0(1.05)
2.5(0.70)
2.2(0.63)
0.25
0.1(0.32)
0.6(0.84)
0.6(0.70)
1.1(0.74)
1.4(0.70)
Notes: The initial period has 5 SS firms and 0 LD firm. The numbers in
parentheses are the standard deviations of the average number of the LD firms.
21
Table 7A: The Average Share of the LD firms still Surviving at
the Final Period (160 period) in 10 simulations per each case
ρ
α
0.00625
0.0075
0
0.27(0.30)
0.49(0.35)
0.05
0.25(0.26)
0.1
0.25
0.01
0.015
0.025
0.54(0.33)
0.77(0.26)
0.88(0.21)
0.45(0.37)
0.54(0.30)
0.64(0.33)
0.82(0.19)
0.24(0.23)
0.28(0.22)
0.51(0.39)
0.65(0.25)
0.63(0.29)
0.02(0.05)
0.13(0.19)
0.14(0.21)
0.27(0.21)
0.50(0.33)
Notes: The share is the number of the LD firms divided by the total number of the
firms in the industry.
22
Table 7B: The Average Number of the All Firms still
Surviving at the Final Period (160 period) in 10
simulations per each case
ρ
α
0.00625
0.0075
0.01
0.015
0.025
0
7.8(2.15)
6.1(2.02)
5.6(1.78)
4.2(0.92)
3.9(1.20)
0.05
7.1(1.60)
5.6(1.84)
5.0(1.41)
4.5(1.58)
3.7(0.67)
0.1
6.6(1.43)
6.2(1.48)
5.1(1.85)
4.1(0.88)
3.9(1.10)
0.25
6.5(1.43)
6.3(1.64)
5.2(1.14)
4.1(1.10)
3.2(1.03)
23
Table 8: Summary Table of the Simulation Results
Technological Regimes
(causal factors)
Direction of
the Impacts
Organizational Selection
(aspects of the impacts)
More Productivity Jump after
innovation
+
Productivity difference among the
firms
More Productivity Jump after
innovation
+
Decrease of average costs of the LD
firms
Higher Cumulativeness
?
Decrease of average costs of the LD
firms
More Productivity Jump after
innovation
+
Possibility of Entry by the LD firms
Higher Cumulativeness
-
Possibility of Entry by the LD firms
Productivity Jump after innovation
+
-
Possibility of Leadership shift from
the SS firms to the LD firms
Higher Cumulativeness
Sources: Based on other tables .
24
6. Summary
1) Owing to the initial cost disadvantage, the LD firms were not
able to enter the industry but later entered as the process
innovation became dominant in market competition and it led to
increasing difference in productivity among the incumbent firms.
2) With less cumulativeness in production, the late entry was not
that much disadvantageous. And with the strong innovationproductivity links, the LD firms were able to expect greater
benefits from R&D and to put more R&D money, and became
the industry leader.
2) Technological regime featured by less cumulative technology and
bigger impact of innovation on productivity work more
advantageously toward large diversified firms than small
specialized firms.
25
7. Topics for Further Research ;
Technological Regimes and Economics of Catch-up
(ranks changes among the same types)
1) Easiness of exit: more difficult -- more R&D
2) Predictable-ness
3) speed of diffusion: quicker diffusion -- more catch-up
4) length of product life cycle
5) variations in productivity ( caused by degree of product jump)
(cf) speed of diffusion
6) the more cost difference among organizational forms > productivity difference,
the more importance of organizational selections
7) innovation probability < imitation probability --> more catch-up
higher external imitation probability --> more catch-up
8) financial capability: higher borrowing limit coefficient --> more catch-up
9) alternative cumulativeness: past innovator tend to innovate more
(cf) in this paper: cumulativeness = easiness of learning
26
10) process innovators vs. product innovators
Appendix table 1: Values of the Coefficients or Parameter at the
Initial Period
R (size of the market demand)
P max (upper bound of price)
hi0 (capital productivity at period 0) = basic productivity (ht )
K0 (capital stock)
VCK s (variable cost per unit of capital in SS firms)
VCK l (variable cost per unit of capital in LD firms)
FC s (fixed cost, SS firms)
FCL (fixed cost, LD firm)
(innovation expenditure ratio)
(imitation expenditure ratio)
(coefficient to adjust innovation expenditure)
(coefficient to adjust imitation expenditure)
E m (total external imitation expenditure)
 (depreciation ratio)
f (a coefficient of borrowing ability)
Fl, (LD firm's borrowing ability premium)
 (weight given to the lagged performance indicator)
 ( weight given to the current level of innovation/imitation expenditure)
r n (standard deviation of innovation expenditure ratio)
r m (standard deviation of imitation expenditure ratio)
Me (financial premium for entry firms)
M r (financial premium for firms changing their R&D expenditure)
X min (minimum level of performance not to be forced to exit)
K min (minimum level of capital not to be forced to exit)
K e,min (minimum entry requirement in terms of the size of capital stock)
64
2
0.15
64
0.16
0.13
0.1
4
0.005
0.002
1
0.4
0.2
0.03
2
1
0.75
0.2
0.002
0.0004
0.013
0.013
- 0.015
20
40
27