Cost–Benefit Analysis of Soil Degradation and Restoration - Principles of Sustainable Soil Management in Agroecosystems - page 237

where Y g = growth rate of agricultural or food production; A g = rate of change in labor
input; L g = rate of change in land cropped; F g = rate of change in fertilizer consump-
tion; M g = rate of change in machinery power utilization; Q = quality of arable land
or soil; and G = government policies, for example, price and land use.
Those variables with subscript “g” are flow variables, and they should be mea-
sured in growth rate. Q and G can be seen as stock variables, and the average values
are to be used in estimation.
The approach used in this study involves estimating a cross-country agricultural
growth function based on a sample of 23 developing countries. Variations in agricul-
tural growth rate are accounted for by differences in the growth rate of agricultural
inputs and related factors. All the data used in this study are from the period of 1971
to 1980. The data for flow variables are the average of 1971 to 1980, and the data for
stock variables are the 1975 actual figure. More detailed development of the study
methodology is presented in Zhao et al. (1991).
10.4.1.1 Analysis and General Results
Zhao et al. (1991) present a detailed discussion of the main results of the estimation
of the aggregate agricultural production and the food production growth function.
Considering the aggregate nature of the secondary data, the levels of statistical sig-
nificance of several of the estimated coefficients are quite good. The six independent
variables in the model can explain as high as 82% of the variance in total agricul-
tural production and about 78% of the variance of food production when growth is
measured by the average index. However, the models based on percentage growth
measure are less satisfactory, and the R 2 's for the TAP and TFP models are 0.66
and 0.68, respectively. The F tests show that one can be 95% to 99.99% confident
of rejecting the hypothesis that all the estimated coefficients are zero for the four
models. These models, as a whole, are quite well defined, and the multicollinearity
problem in the statistical analysis does not appear to be serious based on the SAS
collinearity diagnostics procedure.
Even though the data are categorical (three classifications) and thus gross mea-
sures, the statistical significance is relatively strong. Based on the percentage mea-
sure, the estimate of the moderate level of land degradation is significant at the 5%
level in the total agricultural production growth model, and the estimate for food
production growth is significant at the 2% level. If the growth is measured by the
average index, the estimate for moderate land degradation is not significant in the
total agricultural production growth model. It is, however, still significant at the 15%
level in the good production growth model.
When comparing the absolute values of the estimates and their corresponding
significance levels, the soil and land degradation variable has higher coefficient val-
ues, and the estimates are more significant in the TFP than in the TAP models. This
difference indicates that soil and land degradation tends to affect food production
more significantly than nonfood agricultural production. The result seems to confirm
the belief that soil and land degradation does threaten food production growth. It also
impedes income increases in rural areas because of a direct relationship between
farmers' income and food production growth. Notice that the estimated coefficients
of the high degradation level have wrong signs and are insignificant at the 15% level.