5 Empirical estimation
5.1 Estimation model
Return on sales is collected in MIP survey as scale from 0 to
8 with known cut off points. Ordered Probit Model is the most suited for this
kind of dependant variables. According to Greene (2009), the general model
would be as follows:
The dependent variable in Ordered Probit Model is observed this
way:
Y=1 if Y*?å1
Y=2 if å1<Y*?å2
Y=3 if å2<Y*?å3
...
Y=7 if å6<Y*
In this case Y represents return on sales (Umren) and the cut off
points would be:
0 = 1, 0% to < 2% = 2, 2% to < 4% = 3, 4% to < 7% = 4,
7% to < 10% = 5, 10% to < 15% = 6, 15% and more = 7, estimation not
possible = 8
Y=1 if Y*=0%
Y=2 if 0%<Y*=2%
Y=3 if 2%<Y*=4%
Y=4 if 4%<Y*=7%
Y=5 if 7%<Y*=10%
Y=6 if 10%<Y*=15%
Y=7 if 15 %< Y*
The six independent variables corresponding to each hypothesis
are:
The control variables with a relevant theoretical impact on
return on sales are:
Three different models, M, M1 and M2, will be tested. M
representing the baseline without distinction between eco-innovative firms and
non-innovative firms. M1 being the model for eco-innovative firms and M3 for
the non-innovative firms in order to answer the central problematic of the
thesis.
5.2 Empirical results
Probit Estimation
|
Independent variables
|
Model M
|
Model M1
|
Model M2
|
|
Environmental Innovation
|
0.0822593**
(0.0403479)
|
|
|
|
Market share
|
0.0775999***
|
0.0862508**
|
0.0743373**
|
|
(0.022197)
|
(0.029501)
|
(0.0349319)
|
|
Product differentiation
|
0.0098879
|
0.0501308
|
-0.0946248
|
|
(0.0507557)
|
(0.0605704)
|
(0.0958762)
|
|
Patent stock
|
-0.0457845
|
-0.0516562
|
-0.0376007
|
|
(0.0494725)
|
(0.061838)
|
(0.0841409)
|
|
Martials and energy
|
-0.0692138**
|
-.0523971*
|
-0.1110295
|
|
efficiency
|
(0.0236537)
|
(.0284859)
|
(0.0447996)
|
|
Cost of capital
|
-0.0735411**
|
-0.0505324*
|
-0.2062648 ***
|
|
(0.0255448)
|
(0.0280471)
|
(0.0620886)
|
|
Labour productivity
|
0.8199367***
|
0.9515969***
|
.7446644***
|
|
(0.0427542)
|
(0.0678447)
|
(0.0558337)
|
|
Observations
|
3500
|
1966
|
1534
|
|
Likelihood Ratio
|
646.98***
|
348.64***
|
338.73***
|
|
Pseudo R2
|
0.480
|
0.0463
|
0.0572
|
Table3: Author's own calculation.
5.3 Model discussion
The empirical estimation provides a useful analytical tool to
judge the validity or not of the initial hypotheses postulated in the
introduction of the current thesis concerning the impact on eco-innovation on
firms' competitiveness with an insight of the key differences between
eco-innovative firms and non-innovative firms. The baseline model shows that
environmental innovation does have a significant positive effect
on return on sales which confirms the strong porter hypothesis.
The models M1 and M2 give more visibility of the key
differences between innovative and non-innovative firms in case of green
innovation, providing the reader with additional information concerning with
aspect of the eco-innovation does really explain a higher return on sales for
eco-innovative firms.
In fact, the higher return on sales of eco-innovative firms
compared to non-innovative firms is statistically explained by a greater
coefficient for market share, and labour productivity which confirms the two
hypotheses stipulated previously and lower additional cost (negative impact on
return on sales) in case of cost of capital and materiel and energy efficiency
(Porter justified it by that fact that the impact is actually dynamic rather
than static).
The coefficients of product differentiation, even if they are
not significant, they follow the same logic below with a positive effect for
eco-innovative firms and a negative one for other firms.
And lastly, the coefficient of patent stock has led to the
opposite conclusion but it is not statistically significant.
To put it differently, from the six initial hypotheses only
two were rejected and four confirmed. One may conclude that such results
support, even if only partly, the strong Porter hypothesis.
| 
