Research Papers

Developing an Input-Output Table Generation Algorithm from a Large Scale Company Database in Japan: How to Deal with Ambiguous Export and Import Information

Proceedings of The Japanese Society for Artificial Intelligence 2017(AI-Biz 2017)


Input-Output tables describe the interdependencies between different branches of a national economy or different regional economies and serves as a critical data source for the compilation of GDP statistics. It is also a useful tool in analyzing the impact and transmission channel of a particular shock in the economy. With the preparations for the Tokyo Olympics gaining momentum, there is an increased interest in using input-output tables to evaluate the economic impact of such a large event.

There are a number of shortcomings of the national input-output tables including the lack of timeliness. To address the lack of timeliness, there has been an increased effort to compile input-output tables using firm-level data. Teikoku Databank has been a key data source in obtaining inter-firm network information. However, Teikoku Databank does not distinguish between sales revenue from exports and purchases of imported goods/services. This paper attempts to construct input-output tables from firm-level data whilst accounting for export revenue in sales data and imports costs in purchases data using information collected during firm surveys.

The calculation of export and import values using available qualitative information contained in the firm level database produced results that were close to the official trade statistics published by the Ministry of Finance, with some differences in the estimated import values. The estimation of firm-level exports and imports and sales and purchases (exclusive of exports and imports) should pave the way for analysis looking at the magnitude and ripple effect arising from an overseas shock such as changes in exchange rates.

>For more details, please refer to the link here.



Tokyo Institute of Technology, Takaya Ohsato, Kaya Akagi, and Hiroshi Deguchi

Research and Collaborations TOP