Chapter 5: Empirical results
5.2 Hypotheses test
5.2.2 The impact of R&D cooperation on R&D investments, R&D
1. The impact of R&D cooperation intensity on R&D investments, R&D outputs, and financial performance
To avoid sample selection bias, I use the Heckman two-step model to test H5a, H5b, and H5c. The first step is to use probit model to examine the factors of R&D cooperation for the whole sample. Therefore, I test H1, H2, and H3 again. The model is as follows:
Model 1 (R&D cooperation type probit regression): Regressing R&D cooperation type on Absorptive capacity, Knowledge spillover, and Uncertainty:
ε
In the second step, I calculate inverse Mill’s ratio from the first step to be an adjusted item and use R&D cooperative companies as a research sample to examine the impact of R&D cooperation on R&D investments, R&D outputs, and financial performance. The models are as follows:
Model 2 (R&D investment regression): Regressing R&D investment on R&D cooperation intensity:
Model 3 (R&D output regression): Regressing R&D output on R&D cooperation intensity:
Model 4 (Profit regression): Regressing Profit on R&D cooperation intensity: capacity, knowledge spillover, and uncertainty are all positive and significant (Z=6.04, p<0.01l Z=1.73, p<0.1; Z=2.82, p<0.01, respectively). These results show that when absorptive capacity, knowledge spillovers, and uncertainty are higher, companies engage in R&D cooperation more frequently. The empirical results further support H1, H2. However, the result of H3 is contrary to the result of HLM. Because the variables of uncertainty is an industry-level variable and HLM is used to know how factors across different levels interact with one another and jointly determine the intensity of R&D cooperation. Hence, the result of HLM is more robust than that of Probit regression.
The impact of R&D cooperation intensity on R&D investments, R&D outputs, and financial performance is shown in Table 16. The coefficient of R&D cooperation intensity in Model 2 is positive and significant (t=2.49; p<0.05). This means that R&D cooperation has a positive impact on R&D investments, which supports H5a.
The coefficient of R&D cooperation intensity in Model 3 is also positive and significant (t=11.03; p<0.01). This means that R&D cooperation has a positive impact on R&D outputs, which supports H5b. Regarding financial performance, the coefficient of R&D cooperation intensity in the profit regression model (Model 4) is positive and significant (t=3.35; p<0.01). This means that R&D cooperation has a positive impact on profit. The result supports H5c.
Table 16: The impact of R&D cooperation intensity on R&D investments, R&D outputs, and financial performance—Heckman two-step model
Model 1 Model 2 Model 3 Model 4
Dependent variable:
R&D cooperation type
Dependent variable:
R&D investment
Dependent variable:
R&D output
Dependent variable:
profit
Independent variables Coef. Z-value Coef. t-value Coef. t-value Coef. t-value R&D cooperation intensity .1088185 2.49** .1493304 11.03*** .0048524 3.35***
Absorptive capacity 3.517169 6.04
Knowledge spillover .0014179 1.73
Uncertainty .1032128 2.82
Sales growth .1271074 0.66 .0184954 0.25 .0185296 1.90*
Capital structure -.0289435 -3.34*** .0012949 0.39 -.002108 -4.93***
Firm size .5917384 10.34 .1666954 0.69 .0989728 1.07 .0209127 1.70*
Upstream .5566342 1.48 -.0910848 -0.64 -.0032864 -0.18
Midstream 1.310687 3.99*** .06163 0.49 -.0324389 -1.98
Optoelectronics -.376435 -0.75 -.4788096 -2.47** .016557 0.66
Telecommunications -.2526581 -0.47 -.7351118 -3.54*** .0585871 2.13**
Computer component -1.940495 -3.55*** -.4905255 -2.37** .0591023 2.23**
Computer peripheral .1283916 0.27 -.4730079 -2.63*** .0879492 3.79***
System and equipment -1.261879 -2.40** -.5708449 -2.83*** .0339421 1.31
Inverse Mill’s Ratio -1.554908 -2.30** .3480231 1.35 .0156767 0.45
_cons -9.10099 -10.89 9.699255 2.38** -1.921345 -1.22 16.2967 78.21***
N 596 232 243 237
Psuedo R2 F-value
0.2400 0.4299 15.52***
0.5325 23.97***
0.2167 6.44***
Note: * significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level.
In order to compare the performance of R&D cooperation vs. R&D competition, I use treatment effects model and retain the entire sample for the second-step regression and treat the variable of R&D cooperation type as endogenous. The models are as follows:
Model 1 (R&D cooperation type probit regression): Regressing R&D cooperation type on Absorptive capacity, Knowledge spillover, and Uncertainty:
ε
Model 2 (R&D investment regression): Regressing R&D investment on R&D cooperation intensity and R&D cooperation type:
ε
Model 3 (R&D output regression): Regressing R&D output on R&D cooperation intensity and R&D cooperation type:
ε
Model 4 (Profit regression): Regressing Profit on R&D cooperation intensity and R&D cooperation type:
ε
Table 17 reports the treatment effects model that control for self-selection using full sample. The first-step probit model reports that higher absorptive capacity, knowledge spillovers, and uncertainty are likely to engage in R&D cooperation (Model 1). In the second step (Model 2, 3 and 4), the coefficient on R&D cooperation intensity is still positive and statistically significant which is consistent with the results of Heckman two-step model. However, the coefficient on R&D cooperation type is insignificant except Model 2 (R&D investment regression). This means that R&D cooperation is not enough for the firms to gain higher R&D outputs and financial performance. R&D cooperation intensity is the main drivers for improving R&D outputs and financial performance.
Table 17: The impact of R&D cooperation intensity on R&D investments, R&D outputs, and financial performance—Treatment effects model
Model 1 Model 2 Model 3 Model 4
Dependent variable:
R&D cooperation type
Dependent variable:
R&D investment
Dependent variable:
R&D output
Dependent variable:
profit
Independent variables Coef. Z-value Coef. t-value Coef. t-value Coef. t-value R&D cooperation intensity .1215369 1.93** .0559059 9.97*** .0032288 4.56***
R&D cooperation type 3.946434 3.08*** -.0656694 -0.65 .021107 1.23
Absorptive capacity 3.569024 6.24***
Knowledge spillover .002089 2.55**
Uncertainty .1076634 2.86***
Sales growth .0636759 0.28 .0070297 0.42 .0118905 4.12***
Capital structure -.0428587 -4.71*** .0009179 1.32 -.000607 -5.06***
Firm size .5752795 10.31*** .1011202 0.40 .0423711 2.15** .0079005 2.32**
Upstream .8093722 2.14** .0331422 1.14 .0015167 0.31
Midstream 2.043718 5.87*** .0288706 1.08 .000042 0.01
Optoelectronics .2357682 0.41 -.0699464 -1.60 .0170329 2.26**
Telecommunications -.0024564 -0.00 -.1211725 -2.53** .0211142 2.56**
Computer component -.9822309 -1.83* -.1061002 -2.57** .028062 3.97***
Computer peripheral .2146275 0.41 -.037088 -0.93 .0300923 4.37***
System and equipment -.3566659 -0.65 -.103397 -2.44** .0197959 2.74***
Inverse Mill’s Ratio -1.863144 -2.44** -.0266216 -0.45 -.0176462 -1.74*
_cons -9.09785 -11.09*** 6.720394 1.97** -.8201615 -3.12*** 16.4908 363.20***
N 598 581 576 569
Psuedo R2 F-value
0.2544 0.2522 16.05***
0.3354 23.32***
0.2110 12.68***
Note: * significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level.
2. The impact of different R&D cooperation types on R&D investments, R&D outputs, and financial performance
I further test the impact of different R&D cooperation types on R&D investments, R&D outputs, and financial performance. I also consider the impact of interaction between different R&D cooperation and knowledge spillovers. The models are as follows:
Model 1 (R&D investment regression): Regressing R&D investment on Vertical cooperation, Horizontal cooperation, and Generalized cooperation:33
ε
Model 2 (R&D investment regression): Regressing R&D investment on Vertical cooperation, Horizontal cooperation, Generalized cooperation, and the interaction with Knowledge spillover:
ε
Model 3 (R&D output regression): Regressing R&D output on Vertical cooperation, Horizontal cooperation, and Generalized cooperation:
ε
Model 4 (R&D output regression): Regressing R&D output on Vertical cooperation, Horizontal cooperation, Generalized cooperation, and the interaction with Knowledge spillover:
ε
Model 5 (Profit regression): Regressing Profit on Vertical cooperation, Horizontal cooperation, and Generalized cooperation:
ε
Model 6 (Profit regression): Regressing Profit on Vertical cooperation, Horizontal cooperation, Generalized cooperation, and the interaction with Knowledge spillover:
ε
Table 18 presents the impact of different R&D cooperation types on R&D investments, R&D outputs, and financial performance. Models 1, 3, and 5 show the estimation results with the base models. The results show that vertical cooperation and generalized cooperation significantly and positively affect R&D investments (t=3.14, t<0.01; t=2.25, t<0.05). In addition, generalized cooperation also leads to higher R&D
knowledge spillovers and the interaction between R&D cooperation and knowledge spillovers in the regressions. The findings provide very similar results as above. It is found that Vertical cooperation * Knowledge spillover and Generalized cooperation*Knowledge spillover have positive effects on R&D investments.
Although the coefficient of Vertical cooperation*Knowledge spillover (0.0125045) is larger than that of Generalized cooperation*Knowledge spillover (0.0090275), the difference is not significant (F=0.32, p=0.5727). Therefore, vertical cooperation and generalized cooperation both invest more in R&D relative to horizontal cooperation and R&D competition when knowledge spillovers are higher, and H6a is moderately supported. In Models 4 and 6, I find that the coefficients of Generalized cooperation*Knowledge spillover both have positive effects on R&D outputs and profits (t=2.69, p<0.01; t=5.91, p<0.01). These findings indicate that when knowledge spillovers are higher, generalized cooperation can lead to higher R&D outputs and profits relative to other R&D cooperation types. Thus H6b and H6c are supported.
Table 18: The impact of different R&D cooperation types on R&D investments, R&D outputs, and financial performance
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Dependent variable:
R&D investment
Dependent variable:
R&D investment
Dependent variable:
R&D output34
Dependent variable:
R&D output
Dependent variable:
Profit
Dependent variable:
Profit
Independent variables Coef. t-value Coef. t-value Coef. t-value Coef. t-value Coef. t-value Coef. t-value Vertical cooperation 1.222772 3.14*** .4131625 0.74 .0100367 0.58 -.0040158 -0.16 -.0013308 -0.23 -.0056685 -0.71
Horizontal cooperation -.6806009 -0.96 -1.125567 -0.98 -.0381425 -1.22 -.0311551 -0.61 -.0074627 -0.72 -.0132695 -0.81
Generalized cooperation 1.154383 2.25** .0457696 0.06 .0980286 4.26*** .0261674 0.74 .018125 2.38** -.033706 -2.96***
Knowledge spillover -.0108314 -3.01*** -.0001082 -0.67 -.0000204 -0.40
Vertical cooperation *
Knowledge spillover .0090275 2.08** .0001871 0.95 .0000704 1.12
Horizontal cooperation *
Knowledge spillover .0055933 0.63 -.0000209 -0.05 .0000813 0.64
Generalized cooperation
* Knowledge spillover .0125045 2.15** .0006911 2.69*** .0004981 5.91***
Sales growth -.0880232 -0.39 -.0507862 -0.23 .0035592 0.35 .003193 0.32 .0148265 4.42*** .0144372 4.42***
Capital structure -.0506867 -5.44*** -.0503852 -5.41*** .0001217 0.29 .0002 0.48 -.0006266 -4.51*** -.0005764 -4.25***
Firm size .8214437 6.28*** .7759058 5.90*** .0399771 6.43*** .0366618 5.80*** .0145043 7.03*** .0128751 6.36***
Upstream .7875584 2.01** .4800901 1.11 .0128476 0.73 .0170792 0.89 .0004926 0.09 .0061544 1.00
34
Midstream 1.897476 5.38*** 1.880061 5.23*** .0167572 1.06 .0220489 1.37 -.0053422 -1.01 -.0008259 -0.16
Optoelectronics .0466146 0.08 .0016619 0.00 .000599 0.02 -.0006121 -0.02 .0182017 2.09** .0150794 1.77*
Telecommunications -.0845574 -0.13 .0402893 0.06 -.042567 -1.47 -.0452526 -1.55 .0227114 2.36** .0178307 1.88*
Computer component -1.536795 -3.11*** -1.664756 -3.36*** -.0529776 -2.33** -.0587904 -2.59*** .0269065 3.61*** .0218431 2.99***
Computer peripheral .0707791 0.13 .8287599 1.23 .0044577 0.18 -.0051804 -0.17 .0371416 4.68*** .0234297 2.37**
System and equipment -.5895006 -1.06 -.559017 -1.00 -.0460872 -1.84* -.0497643 -1.98* .0196246 2.38** .0153534 1.90*
_cons -1.276931 -0.58 .1498993 0.07 -.7673383 -7.53*** -.7112588 -6.82*** 16.39586 484.21*** 16.42248 489.11***
N 571 571 555 555 558 556
Adj. R2 0.2526 0.2620 0.1939 0.1991 0.1699 0.1902
F-value 15.82*** 12.91*** 11.25*** 9.10*** 9.77*** 8.67***
Note: * significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level.
3. The direct and indirect impact of R&D cooperation intensity on R&D investments, R&D outputs, and financial performance
To understand the direct and indirect impact of R&D cooperation intensity on R&D investments, R&D outputs, and financial performance, I further use path analysis to examine their relationship. The models are as follows:
Model 1: Regressing Profit on R&D cooperation intensity:
ε
Model 2: Regressing R&D investment on R&D cooperation intensity:
ε
Model 3: Regressing R&D output on R&D cooperation intensity and R&D investment:
Model 4: Regressing Profit on R&D investment:
ε
Model 5: Regressing Profit on R&D output:
ε
Model 6: Regressing Profit on R&D cooperation intensity and R&D investment:
Model 7: Regressing Profit on R&D cooperation intensity and R&D output:
ε
Model 8: Regressing Profit on R&D cooperation intensity, R&D investments, and R&D output:
0 & int & investment &
Pr
Table 19 shows the path analysis between R&D cooperation intensity, R&D investments, R&D outputs, and financial performance in 2001. I find that R&D cooperation intensity has a direct effect on Profit (t=6.92, p<0.01), R&D investment (t=2.60, p<0.01), and R&D output (t=17.30, p<0.01) in Model 1, 2 and 3, which are consistent with the results of Heckman two-step model and treatment effects model.
R&D investment has no direct effect on Profit (t=0.37, p=0.713) in Model 4.
Therefore, H7a is not supported. In addition, the coefficient of R&D cooperation intensity becomes slightly larger (The coefficient of R&D cooperation intensity is changed from 0.311 in Model 1 to 0.313 in Model 6) after controlling for the effect of R&D investments on financial performance (t=6.92, p<0.01) in Model 6. Hence, H8a is not supported which indicate that the impact of R&D cooperation intensity on financial performance is not mediated by R&D investments. In contrast, R&D output has a positive impact on profit (t=10.55, p<0.01) in Model 5, which supports H7b.
Once I control for the effect of R&D outputs on financial performance, the direct effect of R&D cooperation intensity on profit no longer exists (t=1.34, p=0.181) (Model 7). Together, these findings support H8b’s prediction that the effect of R&D cooperation on financial performance is mediated by R&D outputs. See Figure 6 for the path between R&D cooperation, R&D investments, R&D outputs, and financial
Table 19: The path analysis between R&D cooperation intensity, R&D investments, R&D outputs, and financial performance
Model 1 Model 2 Model 3
Dependent variable:
Profit
Dependent variable:
R&D investment
Dependent variable:
R&D output
Independent variables: Coef. t-value Coef. t-value Coef. t-value
R&D cooperation intensity .310924 6.92*** .1143941 2.60*** .6550353 17.30***
R&D investment -.0522751 -1.45
R&D output
Sales growth .1135411 2.89*** -.0214036 -0.56 .0196114 0.60
Capital structure -.1986721 -5.00*** -.2022589 -5.19*** -.0304236 -0.89
Firm size .1587066 3.39*** .247595 5.40*** .0032276 0.08
Upstream -.0058575 -0.13 .1022043 2.23** -.0343485 -0.88 Midstream -.0534298 -1.15 .2284298 5.01*** .0544386 1.37
Optoelectronics .0243313 0.47 .030012 0.59 -.1084808 -2.49**
Telecommunications .0947205 1.91* .0233555 0.48 -.1197046 -2.88***
Computer component .1935287 3.19*** -.1702806 -2.86*** -.0912834 -1.78*
Computer peripheral .1988898 3.49*** .0163349 0.29 -.1201103 -2.52**
System and equipment .0823636 1.58 -.0435842 -0.85 -.1025022 -2.34**
N 577 577 577
Adj. R2 0.1985 0.2282 0.4378
Note: * significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level.
Table 19: The path analysis between R&D cooperation intensity, R&D investments, R&D outputs, and financial performance (cont.)
Model 4 Model 5 Model 6 Model 7 Model 8
Dependent variable:
Profit
Dependent variable:
Profit
Dependent variable:
Profit
Dependent variable:
Profit
Dependent variable:
Profit
Independent variables: Coef. t-value Coef. t-value Coef. t-value Coef. t-value Coef. t-value R&D cooperation intensity .3127386 6.92*** .0706144 1.34 .0700509 1.32
R&D investment .0163585 0.37 -.0158633 -0.37 .0035044 0.09
R&D output .4077024 10.55*** .3702451 7.76*** .3704956 7.75***
Sales growth .1154766 2.82*** .1053268 2.81*** .1132016 2.88*** .1058658 2.83*** .1059356 2.83***
Capital structure -.209904 -4.96*** -.1927574 -5.09*** -.2018805 -4.96*** -.1913225 -5.06*** -.1906087 -4.92***
Firm size .3189568 7.24*** .1872832 4.64*** .1626343 3.39*** .1623037 3.65*** .1614384 3.54***
Upstream .0153995 0.32 .0137599 0.31 -.0042362 -0.09 .008838 0.20 .0084897 0.19 Midstream -.0390159 -0.79 -.0680368 -1.54 -.0498062 -1.05 -.0691643 -1.56 -.0699755 -1.54 Optoelectronics -.0170982 -0.32 .0630654 1.27 .0248074 0.48 .0650767 1.31 .064999 1.31 Telecommunications .0788313 1.53 .1416988 2.99*** .095091 1.92** .1394926 2.94*** .139441 2.94***
Computer component .1293679 2.05** .2170861 3.76*** .1908275 3.12*** .2240302 3.87*** .2246476 3.85***
Computer peripheral .1641517 2.78*** .2430428 4.46*** .199149 3.49*** .2436762 4.47*** .2436493 4.47***
System and equipment .06866 1.26 .1210652 2.43** .0816722 1.56 .119471 2.40** .1196488 2.39**
N 577 577 577 577 577
Adj. R2 0.1307 0.2735 0.1987 0.2758 0.2758
Note: * significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level.
Figure 6: The path between R&D cooperation intensity, R&D investments, R&D outputs, and financial performance