Page 153 - Special Topic Session (STS) - Volume 2

STS474 Gigih F. et al.

Fig 5. The F-test (left) and Wald Test (right) for Spatially-Dependent model.

As the next finding, we observed that the positive spatial correlation after
crisis is higher than before crisis (see in Fig 4). Connecticut, Florida, Michigan,
Maine, and South Carolina are states that have high spatial correlation
∗
(̂ + ̂ ) after crisis which indicates the recoveries from neighboring states

have effect toward those states’ regional economic recoveries. Wyoming
∗
become the only state that has (̂ + ̂ ) < 0.

Fig 5 shows the evidence that supports the structural break in US. The
Wald test results shows more appropriate test compared to the F-test. The
endogeneity (explanatory variable correlated with error term) in spatially-
dependent model might imply the F-test become non-robust for
autocorrelated disturbances (Krӓmer, 2003). On the other hand, the Wald test
produce more appropriate results hence the test focus to evaluate the distance
between the estimated parameters in constrained and unconstrained form
(Greene, 2012).

3.2 Breakpoint Analysis of Japan Regional Data
3.2.1 By Spatially-Independent Model
By using 2008 as the breakpoint, Kagawa prefecture shows positive trend
before crisis occurred (see Fig 6 and Fig 7). However, this exceptional case
was occurred due to the impact of the massive increase in manufacturing
production for Kagawa in 2007 based on Bank of Japan (BOJ) Report (2008).
The other prefecture showed a negative trend before crisis, such as Tokyo
prefecture. However, we observed that Tokyo shows a strong recovery for
regional economic growth after crisis.

142 | I S I W S C 2 0 1 9