Chapter 5: Empirical results
5.3 Robustness test
5.3.3 Industry analysis
According to my interview, R&D period and performance is very different among different products. For example, it may only take one year to improve a memory chip, but 3-5 years may be required to invent large hardware. The length of the patent granting procedure is also influenced by institutional circumstances (Ernst 2001).
Finally, different industries face different technology complexity and product markets.
For instance, the semiconductor industry does not have products in the market place relative to the computer peripheral industry. Therefore, industry differences may play a very important role in the relationship between R&D cooperation, R&D investments, R&D outputs, and financial performance. To consider the disturbance arising from industry differences, this analysis separates the sample of high-technology industry into six sub-industries according to the characteristics of products: semiconductor, optoelectronics, telecommunications, computer component, computer peripheral, and system and equipment industries (including software, information system, and electronic equipment industry). Table 25 shows the descriptive statistics of these industries.
0.17***
-0.03 0.23***
0.20*** 0.25***
R&D Cooperation Intensity
R&D output R&D Investment
Financial Performancet+3
* significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level.
0.03
among six industries (28.5%), but has the least percentage of R&D cooperation (26.7%) and the least number of R&D cooperation events (0.622). On average, the semiconductor industry has the highest R&D investments, R&D outputs, absorptive capacity, and largest firm size. Most of the semiconductor firms are in an upstream industry (54.8%) while most of the firms of optoelectronics (46.8%), telecommunications (59.3%), computer component (52.3%), and computer peripheral (50%) industries are in a midstream segment.
Table 25: The descriptive statistics of sub-industries Variables Semi-
conductor
Optoelec-tronics
Telecom- municatio
ns
Computer component
Computer peripheral
System and equipment
N 84 80 59 172 125 84
R&D cooperation type .4756098 .5696203 .5423729 .2674419 .483871 .3809524 R&D cooperation intensity 3.073171 1.78481 2.20339 .622093 1.741935 1.142857 Profit (thousand) -320856.3 149797.7 1460391 288801.8 576063.9 149980.9 R&D investment 417988.2 159682.4 176350 52398.73 161489.6 38174.57 R&D output .4414037 -.0938835 -.1741875 -.0289392 -.0881625 -.2178195 Knowledge spillover 50.59756 57.24051 102.2542 61.06395 199.7419 79.34524 Absorptive capacity .1806442 .0925523 .0948646 .0369198 .0628343 .076099 Uncertainty .1587312 .2292901 -.2625014 -2.322589 -1.392393 .8818693 Sales growth 5.95 190.6486 82.41071 6.982353 14.225 187.0476 Capital structure 30.78049 37.82278 33.34483 40.11696 41.19512 33.11905
Firm size 1.31e+07 9175295 1.35e+07 3784093 7319260 1589802
Upstream .5487805 .2911392 .1016949 .2674419 .1612903 .3571429
Midstream .195122 .4683544 .5932203 .5232558 .5 .1904762
Downstream .2560976 .2405063 .3050847 .2093023 .3387097 .452381
The findings of the regression of R&D cooperation intensity on R&D investment are displayed in Table 26. It appears that R&D cooperation intensity of all sub-industries has a positive influence on R&D investments except for the computer component and the computer peripheral industry. This is perhaps because R&D cooperation in the computer component and the computer peripheral industry induces
a cost sharing effect and eliminates wasteful duplication, which in turn counteracts the encouraging effect of R&D cooperation on R&D investments.
In R&D output regression (Table 27), the model of system and equipment industry does not reach a significant level (F=0.53). Hence, I will not further discuss this model. On the other hand, semiconductor, optoelectronics, telecommunications, computer component, and computer peripheral industries all show positive signs for R&D cooperation intensity (t=5.57, p<0.01; t=3.17, p<0.01; t=2.47, p<0.05; t=12.55, p<0.01; t=5.62, p<0.01, respectively), which are consistent with the results of the Heckman two-step model.
With regard to financial performance, Table 28 reveals mixed results. The model of the optoelectronics industry does not reach a significant level (F=1.56). Therefore, I do not further discuss this model. The R&D cooperation intensity variable of semiconductor, telecommunications, computer component, and computer peripheral industries is significant and has a positive sign (t=2.38, p<0.05; t=2.43, p<0.05; t=5.97, p<0.01; t=5.12, p<0.01, respectively). However, the coefficient of R&D cooperation intensity in system and equipment industry is not significant (t=-0.77, p=0.446). The most probable reason is that R&D cooperation has a time-lagged effect on financial performance. Lin and Chen (2005) argue that the impact of R&D activity on performance often has a lag time of more than 2 to 3 years. Ernst (2001) also shows that patent applications lead to performance increases with a lag time of 2 to 3 years after the priority year. As mentioned earlier, the innovation process and innovation period are very different among products. In system and equipment industry, 80% of the firms are related to information systems and electronic equipment, which requires a longer R&D period to generate R&D outputs and profits, according to the interview.
I therefore test the time-lagged effect of R&D cooperation on future R&D outputs and profits for system and equipment industry. The results show that the coefficients of R&D cooperation intensity for system and equipment industry become positive and significant with a lag of one year (The estimated results of the lagged effect are not displayed here). Therefore, in summary, I find that the impact of R&D cooperation on R&D investments, R&D outputs, and financial performance differs across different industry sectors. In semiconductor, and telecommunication industries, R&D cooperation induces higher R&D investments and creates greater R&D outputs and
R&D cooperation creates synergy among R&D cooperative partners, and also leads to higher R&D outputs and profits. However, in system and equipment industry, the time-lagged effects influence the relationship between R&D cooperation, R&D outputs, and financial performance.
Table 26: R&D investment regression model for sub-industries Dependent variable: R&D output
Semicon- ductor
Optoelec- tronics
Telecom- munications
Computer component
Computer peripheral System and equipment Independent
variables Coef. t-value Coef. t-value Coef. t-value Coef. t-value Coef. t-value Coef. t-value R&D cooperation
intensity .1254901 5.68*** .1400886 2.88*** .123091 2.00** .0082681 0.04 .1520477 1.25 .5873397 1.84*
Sales growth -.012828
6 -0.12 -.190122
2 -1.31 .3019222 1.78* -.063139
1 -0.10 -.518137
1 -1.16 -.710414
1 -1.29 Capital structure
.0041585 0.59 .0041398 0.45 -.018338
8 -2.15** -.027912
7 -1.41 -.050797
4 -3.64*** -.071102
6 -2.40**
Firm size .5545549 6.70*** .5739332 4.89*** .6177863 3.83*** 1.034375 3.21*** 1.033247 3.96*** .4253183 0.77 Upstream
1.133748 4.06*** .523589 1.75* .7097205 1.55 -1.94916
1 -2.29** 2.762436 3.79*** -.079706
7 -0.09 Midstream
.2214048 0.76 -.304269
2 -1.03 .4761304 1.56 1.71017 2.16** 2.398168 4.49*** 2.036644 1.84*
_cons
2.001513 1.42 2.876642 1.65* .6917285 0.30 -5.80861
6 -1.04 -2.52036
7 -0.55 7.378075 0.91
N 72 70 51 170 116 80
Adj. R2 0.7778 0.6255 0.6405 0.2078 0.4113 0.1525
F-value 42.43*** 20.21*** 15.85*** 8.39*** 14.39*** 3.37***
Note:
1. * significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level 2. The figures in parentheses are the t value.
Table 27: R&D output regression model for sub-industries Dependent variable: R&D investment
Semicon- ductor
Optoelec- tronics
Telecom- munications
Computer component
Computer peripheral System and equipment Independent
variables Coef. t-value Coef. t-value Coef. t-value Coef. t-value Coef. t-value Coef. t-value R&D cooperation
intensity .1578392 5.57*** .0244593 3.17*** .0129333 2.47** .176439 12.55*** .0225754 5.62*** .0000429 0.08 Sales growth
.0337454 0.24 -.019295
6 -0.81 .0394195 2.60** .0356024 0.95 .016184 0.52 -.000201
8 -0.14 Capital structure -.002949
9 -0.34 .0014294 1.06 .0004256 0.59 .0009717 0.86 -.001087
2 -1.10 -.000053 -0.70 Firm size
.1287845 1.26 .0293714 1.59 -.015818 -1.23 -.015618
9 -0.83 .0380142 2.62** .0015214 1.07 Upstream -.278334
1 -0.85 .0399376 0.84 -.037130
8 -0.82 -.007071
9 -0.15 .0793668 1.48 .0024438 1.02 Midstream
.1047683 0.30 -.009364
7 -0.21 -.018215
3 -0.70 .0368225 0.81 .0348963 0.92 .0004218 0.15 _cons -2.04697
3 -1.18 -.599973
5 -2.13** -.170471
4 -0.91 -.263445
2 -0.80 -.788025
9 -2.81*** -.241982 1
-11.49**
*
N 80 72 54 168 118 81
Adj. R2 0.4945 0.3609 0.1435 0.5282 0.3459 -0.0369
F-value 13.88*** 7.68*** 2.48** 32.15*** 11.31*** 0.53
Note:
1. * significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level 2. The figures in parentheses are the t value.
Table 28: Profit regression model for sub-industries Dependent variable: Profit
Semicon- ductor
Optoelec- tronics
Telecom- munications
Computer component
Computer peripheral System and equipment Independent
variables Coef. t-value Coef. t-value Coef. t-value Coef. t-value Coef. t-value Coef. t-value R&D cooperation
intensity .0055714 2.38** .0014048 0.59 .0090026 2.43** .0080376 5.97*** .0092737 5.12*** -.000273
2 -0.77 Sales growth .0195461 1.79* .0055629 1.37 .0165818 1.72* .0106859 2.96*** .0292287 4.74*** .0034733 3.45***
Capital structure
-.001612 -2.47** -.000354
9 -1.47 -.000743
2 -1.61 -.000116
8 -1.08 -.000563
5 -2.88*** -.000052
2 -1.01 Firm size -.016301
9 -2.12** .0039564 1.04 .008013 0.96 .0099852 5.53*** .0194333 5.40*** .0082938 7.35***
Upstream -.027619
8 -1.11 -.007493
8 -0.92 .0187502 0.72 -.000971 -0.21 -.002739
3 -0.27 .0008071 0.49 Midstream -.016766
5 -0.62 -.011427
6 -1.49 .0012782 0.08 .003408 0.78 -.008058
3 -1.08 .0025269 1.26 _cons 16.85967 128.77**
* 16.59815 293.49**
* 16.4878 136.99**
* 16.47657 520.87**
* 16.28313 256.48**
* 16.533 1007.14***
N 76 66 54 168 117 81
Adj. R2 0.1620 0.0493 0.2271 04226 0.5996 0.4279
F-value 3.42*** 1.56 3.60*** 2.37** 29.96*** 10.97***
Note:
1. * significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level 2. The figures in parentheses are the t value.