DISASTER AND ECONOMIC STRUCTURAL CHANGE: CASE STUDY ON THE 1995 KOBE EARTHQUAKE

Abstract

In 1995, the Kobe Earthquake occurred in the second largest economic region of Japan, and its economic damages were accounted around 10 trillion yen. A catastrophic event of this magnitude would have surely created some long-run effects to the regional economy as well as to the surrounding regions. Additionally, the recovery and reconstruction activities would have affected the economic structure of the region and interdependence between regions in a potentially different way from the original growth trend before the event. While these long-run economic effects may have become sizable, few studies have been conducted to empirically measure or evaluate such effects, due to the significant noises in economic data muddled with macroeconomic influences from the outside. This paper presents an empirical investigation of long-run economic effects of the Kobe Earthquake, using structural decomposition methods. The results indicate significant changes in economic structure of the Kobe economy, and the changes are quite different across sectors and among factors. An additional investigation using shift-share analysis yielded the regional-specific changes; the corresponding decomposed factors of structural analysis with shift-share results appear complicated, and changes in regional final demand were found to be most influential to the changes in output for many sectors.

Keywords:

• Disaster
• Structural decomposition analysis
• Regional input–output table
• Long-run impact

Notes

¹Dietzenbacher and Los (1998) argued that structural decomposition does not provide a unique solution, having “a multitude of equivalent forms” (p. 307), and this creates a inconsistency in results among different forms. They recommended to calculate all the possible combinations (if the number of determinants is n, the number of equivalent decomposition forms becomes n!) and take average or report the range of the results. They also suggested that the average of two polar decompositions can be very close to the n! results, and Miller and Blair (2009) concluded that it is “often an acceptable approach” (p. 595). In addition, as presented in this sub-section, the structural decomposition in this paper contains a hierarchical form, while Chen and Wu (2008) claimed “calculating mathematical expectation of n! decomposition forms is not possible in hierarchical I-O SDA” (p. 887). Hence, in this paper, the average of two polar decompositions is used in the structural decomposition forms.

²The raw data were extracted from Cabinet Office of Japan's web site (http://www.esri.cao.go.jp/jp/sna/toukei.html#kenmin (in Japanese)). And, all the years are set in the Japanese fiscal year, starting from April of a year and ending at March of the next year.

³Since there are some negative decomposed changes, the sum of two can exceed 100% of the total output change.

⁴Oftentimes, the number of employments, rather than output level, is used for shift-share analysis, because it is easier to obtain in a regional context and for regional employment analysis. Shift-share analysis with output level can be found Mayor and López (2008) and Márquez et al. (2009) among others.