Appropriate assumption on cross-hauling national input–output table regionalization

ABSTRACT

Input–output tables are useful for regional economic analyses. Although scholars often regionalize national input–output tables, cost-related issues make surveying regional trade flow difficult; hence, non-survey approaches are implemented instead. While location quotient (LQ) approaches have been used widely, they ignore cross-hauling in interregional trade. Therefore, alternative non-survey approaches with different assumptions on cross-hauling are used, such as cross-hauling depends on regional size and cross-hauling is proportional to its potential determined by output or demand. This study concludes that the most appropriate assumption, as per the relative performance of non-survey approaches, is that cross-hauling is in proportion to trade volume.

KEYWORDS:

• input–output table
• regionalization
• interregional trade
• cross-hauling

JEL:

• C67
• R15

ACKNOWLEDGEMENTS

The author thanks Editage (www.editage.jp) for English language editing.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the author.

Notes

1 Fujimoto (2015) is translated into English as Fujimoto (2017).

2 Techniques used to construct the import matrix data vary between countries; however, every country in the Organisation for Economic Co-operation and Development (OECD) database more or less made use of the import proportionality assumption in the construction of its import matrices (OECD, 2000, p. 12).

3 For the English versions of the interregional IO tables, see http://www.meti.go.jp/english/statistics/tyo/tiikiio/index.html.

4 $\begin{array}{r}{\sum }_j{\text{DSLQ}}_i^R\left(1-m_i^N\right)a_{ij}{X}_j^R+{\sum }_f{\text{DSLQ}}_i^R\left(1-m_i^N\right)c_{if}{Y}_f^R+E_i^R={\text{DSLQ}}_i^R\left(1-m_i^N\right)\left({\sum }_j{a}_{ij}{X}_j^R+{\sum }_f{c}_{if}{Y}_f^R\right)+E_i^R\end{array}$ $={\text{DSLQ}}_i^R\left(1-m_i^N\right)\left({\sum }_j{Z}_{ij}^R+{\sum }_f{Y}_{if}^R\right)+E_i^R={\text{DSLQ}}_i^R\left(1-m_i^N\right)D_i^R+E_i^R=X_i^R,$

therefore:

$E_i^R=X_i^R-{\text{DSLQ}}_i^R\left(1-m_i^N\right)D_i^R$.

Using equation (4),

$E_i^R=X_i^R-{\text{DSLQ}}_i^R\left(1-m_i^N\right)D_i^R$

can be deformed as:

$M_i^R=D_i^R-{\text{DSLQ}}_i^R\left(1-m_i^N\right)D_i^R$.

5 Flegg and Webber (1997) used employment as the measure, while this study uses value added due to limited data availability.

6 Along the columns of scrap sectors, zero inputs from the row sectors are recorded in all cells. Along the rows of scrap sectors, negative inputs to the column sectors that generate scraps are recorded and positive inputs to the column sectors that use scraps are recorded, which cancel out each other.

7 If the confidence interval of ${\mu }_1^R$ does not contain 0, the null hypothesis,${\mu }_1^R=0$, can be rejected. If the confidence interval of ${\mu }_2^R$ does not contain 1, the null hypothesis, ${\mu }_2^R$ = 1, can be rejected.

8 The paired t-test is a procedure to test whether the mean difference between two sets of observations is 0.

9 The generated regional production is underestimated in the sectors that output lower import propensity products because the import propensities are overestimated. However, they are overestimated in the sectors that output higher import propensity products because the import propensities are underestimated. However, the estimated output multipliers may not be biased because the over- and underestimation cancel each other. This can be considered a mixed success.