Explaining Growth in Dutch Agriculture: Prices, Public R&D, and Technological Change

Table 2. Augmented Dickey-Fuller Test for the variables in the model (1949-1997)

^a The letter(s) C or CT indicate whether that test contained a constant or a constant plus a linear
time trend. The letter L indicates that lagged value was included to exclude serial correlation

Estimation procedure

The model was estimated using a two-step estimation procedure. Firstly, the short-run system of
output supply and variable input demand equations (3) was estimated. Secondly, the remaining
undetermined parameters of the profit function were estimated, by regressing a ‘corrected profit
variable’ on the remaining variables, or (cf. Bouchet et al, 1989).

∏ = βo+∑jμ_iz_{i +} ∑ δkW_i ⁺ 2ΣjΣjγ_fZZ.

⁺2 Σ k Σ kXk WkW_t-+∑ j ∑ _lθ_llzjw_k

The estimation of (6) provides estimates for the γ_jj. and θ_jk parameters, that are necessary to
determine the long run price responses (see equation (4)).

As stated before, non-stationary series leads to spurious correlation characterized by over
excitingly high R-squares, low Durbin Watson-statistics and non-stationary residuals. This was
confirmed by the results found when estimating the model with the variables in levels following the
procedure described above. Even with cross equation restrictions for symmetry imposed, the R² ’s are
all above 0.96. Some own price reactions were in conflict with economic theory. When we calculated
the elasticities the order of magnitude was rather similar to Bouchet et al. (1989). The price sensitivity
of agricultural output and input was found to be low. The Augmented Dickey Fuller (ADF) test
statistic was calculated to check for stationarity of the residuals³. For arable and meat the null
hypothesis is rejected at a 5% of significance level, implying no presence of cointegration. For the
other variables a unit root could not be rejected. Instead of estimating the model in levels, therefore the
model is re-specified in differences. Since all variables appeared to be integrated of order 1, they
follow a difference stationary process and first differencing of the variables will generate stationary
series.

Firstly, the system of supply and demand equations (3) was redefined in first-difference form. As
can be seen from (3) the original intercept α_i will cancel out

<<   <   5   6   7   8   9   10   11   >   >>