2016 年 46 巻 3 号 p. 267-280
This paper aims to investigate interregional commodity flows linked to the current transportation network of the San-En-Nanshin Region in Japan. Since spatial units may differ in important characteristics such as size, homoscedasticity is a very strong assumption and may not hold in empirical cases. For this reason a heteroskedastic version of the conventional spatial interaction models was considered to test the presence of network dependence among commodity origin-destination flows. The empirical results show that network dependence affects modeling interregional flows and can be successfully captured by employing a Spatial autoregressive model with autoregressive disturbances (SARAR model) using generalized spatial two stage least squares (GS2SLS) estimation procedures. We also show that the indirect impacts of explanatory variables are significantly dominant with respect to the direct impacts in term of their magnitudes reflecting the presence of strong spatial autocorrelation. In addition, the highly significant total impact of travel distance in time on commodity flow emphasizes the paramount role of transportation connectivity in interregional freight shipments.
JEL Classifications: R12, R15, R41