Industrial Cores and Peripheries in Brazil

Table 5: Variables

^a Type A: firms which innovate, differentiate products and export.

^b Type B: firms that specialize in standardized products and that export.

^c Type C: firms which do not differentiate products, have lower productivity and do not export.

The spatial econometric models allow distinguish two types of spatial correlation, which result
in multiplier effects both locally and globally. Global effects are recorded using SAR (spatial
autoregressive) models and local effects using SMA (spatial moving average) models.

The two SAR models most commonly used in spatial econometrics are the spatial
autoregressive error and the spatial lag models. Global spatial dependence in error terms is taken into
account using spatial autoregressive error terms, as follows:

Y = Xβ + ε                                                               (1)

ε = λWε + u                                                             (2)

Y = Xβ + (I-λW)^-1 u                                                            (3)

Where ε is the autocorrelated error term and u é is an i.i.d. error term. The spatial error model is
suitable when the variables that are not included in the model but are present in the error terms are
spatially autocorrelated. The spatial lag model is specified as follows:

^{Y = ρWy + Xβ +ε}                                                      (4)

15

<<   <   12   13   14   15   16   17   18   >   >>