Analysis of the Relationship Between Agriculture, Economic Growth, and the Environment Through Keynesian Models - The Agricultural Economics of the 21st Century - page 111

4 The Model

The model was built based on the Verdoorn law, where the productivity growth rate
is endogenous and dependent upon the output growth rate, plus the investment
(Costantini and Monni 2008 ; Yanqing and Mingsheng 2012 ) and with some
environmental variables in growth rate. The environmental variables considered
are relative to the water quality (Kleemann and Abdulai 2013 ), expenditure on
environment (Costantini and Monni 2008 ), proportion of municipal waste collected
selectively (Costantini and Monni 2008 ; Cherniwchan 2012 ; Henriques and
Antunes 2012 ; Yanqing and Mingsheng 2012 ;A¸ıcı 2013 ), municipal waste col-
lected (Yanqing and Mingsheng 2012 ), and sales of liquid and gaseous fuels as a
proxy for gaseous emissions (Bartz and Kelly 2008 ; Costantini and Monni 2008 ;
Cherniwchan 2012 ;A¸ıcı 2013 ; Baek and Kim 2013 ; Kleemann and Abdulai 2013 ).
The model can be presented as following:

p it ¼ a 0 þ a 1 q it þ a 2 i it þ a 3 wt it þ a 4 ee it þ a 5 wss it þ a 6 wsr it þ a 7 sgl it

where all the variables are in growth rates, p is the productivity, q the output, i the
investment, wt the water quality, ee the expenditure on environment, wss the waste
collected selectively, wsr the waste collected, and the sgl is the sales of liquid and
gaseous fuels. The index i and t are, respectively, relative to the regions and time.

5 The Results

All the results obtained with the model presented before and with econometric
estimations, considering panel data techniques and the Stata ( 2011 ) software, are
presented in Table 8.2 .
Some industries such as the tobacco industry and the manufacture of coke,
refined petroleum products, and fuel pellets do not have sufficient available data
to make statistically significant estimations, due to this their results are not shown in
Table 8.2 .
In all estimations the statistical tests reject the hypotheses of fixed or random
effects and confirm that the OLS method is the most adjusted. This is more or less in
line with the fact that the constant coefficient is not statistically significant for
almost all the estimations, except in the clothing industry, printing and reproduction
of recorded media, and manufacture of other nonmetallic mineral products. On the
other hand, the majority of the R 2 are superior to 50 %.
All the Verdoorn coefficients are statistically significant, except for the manu-
facture of furniture and mattresses, and the majority are greater than 0.5. The
greatest values for this coefficient are presented by the manufacture of machinery
and equipment, i.e., (0.874), other manufacturing (0.886), and food industries
(0.861), respectively. The lowest values are shown by the repair, maintenance,