June 2014 Joji tokui (Shinshu University and rieti), Tsutomu miyagawa (Gakushuin University and rieti), Kazuyasu kawasaki

The Economic Impact and Recovery of the Supply Chain Disruptions in the Great East-Japan Earthquake

• 
  This paper is a revised version of the paper which was presented at the first workshop on the economic impacts on the 3.11 earthquake at Tokyo. We thank Professors Robert Dekle (University of Southern California), Jonathan Eaton (Pennsylvania State University), Theresa Greany (University of Hawaii) and Etsuro Shioji (Hitotsubashi University) for excellent comments. The preliminary version of this paper was a part of RIETI Policy Discussion Paper No. 12-P-004 “The Economic Impact of the Great East-Japan Earthquake: Comparison with Other Disasters, the Supply Chain Disruptions, and the Electric Power Supply Constraint” (in Japanese). This study is partly supported by a Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology (No.24530296 and No.22223004) of Japan
  

The Great East-Japan Earthquake of March 11th, 2011 had a serious negative economic impact on the Japanese economy. The earthquake substantially reduced production not only in regions directly hit by the earthquake but also in the other part of Japan by the propagated consequences of supply chain disruptions. We examine the economic impact of the supply chain disruptions immediately following the earthquake using regional IO tables, the JIP database and other statistics. Our estimate shows that the amount of production loss caused by the supply chain disruptions would count 1.3% of Japanese GDP at maximum level. We also analyzed the possible extent of the damage reducing effect by building multiple supply chains to cope with potential natural disasters in the future. On the other hand, economic linkage of the earthquake hit regions to the other regions of Japan may promote early recovery from damages relying on affluent cooperation by interested firms. We confirm this possibility using regional IO tables.

Key words: earthquake, economic damage, supply chain, regional IO tables,

JEL classification: L94, Q43, R11, R15

The Great East-Japan Earthquake of March 11, 2011 had a serious negative economic impact on the Japanese economy. The destructions in social infrastructures such as power plants, roads, railways, and ports gave negative effects on economic activities in Tohoku and North-Kanto areas. However, in the Great East-Japan Earthquake, as shown in Table 1 it is important to note that the short-term production activities in the private sector were greatly influenced not only in the earthquake hit area but also in the area that was not directly hit by the earthquake. Right after the earthquake, even in the countries outside of Japan, especially in the automobile and electronic equipment industries, the concern about the possible effect on their production activities caused by the supply shortage of essential parts of their product was widely publicized.

[Table1]

Although the major part of this concern was resolved a couple of months later by the devoted effort to restore the factories in the disaster-affected area and the substitution of the supply sources, this incident raised the awareness of the potential risk of the propagation of the disaster to the production activities outside of the disaster-affected area, transmitted by the supply chain disruptions. Since the incident caused by the disrupted supply chains is one of the phenomena caused by input-output interconnections of industries, we can analyze this incident in the framework of an input-output model. Therefore, in this paper we estimate the magnitude of such propagation effect as precise as possible using regional IO tables in the first place. Based on this result we assess the potential extent of the damage reducing effect by building multiple supply chains to cope with potential natural disasters in the future.

The previous studies on the economic impacts on the earthquake have focused on the demand-side aspect of these interconnections; that is, the propagation to upstream industries of the decreased input demand from the disaster-affected industries. On the other hand, our focus is on the propagation to downstream industries of declined intermediate-output supply from the disaster-affected industries. This supply-side propagation of input-output linkage is named ‘forward linkage’ by Miller and Blair (2009) while the demand-side linkage is named ‘backward linkage’. Although the concept and the methodology of the ‘forward linkage’ is well established, we need to slightly modify it to apply it to the case of the Great East-Japan Earthquake.

In the next section, we will explain how we estimate effects of supply chain disruption by using input output tables. To apply ‘forward linkage’ to the 3.11 earthquake, we have to start to estimate the damages in outputs by industry in the damaged areas. In the third section, we estimate them by using regional production data by Japan Industrial Productivity Database and estimated damage rates by Development Bank of Japan. In the fourth section, we move to estimate effects of supply chain disruptions based on the measured damages in the disaster affected areas. In the final section, we summarize our results.

As is well-known, the input-output table records how the outputs from the industries in the column were used as intermediate goods for the industries in the rows. When we analyze the demand-side linkage, we look at the rows of the table to capture the effect. On the other hand, when we analyze the supply-side linkage, we read the columns of the table. Let X be an output vector for each sector (X’ denotes its transpose), Z be an input-output matrix of the intermediate goods and V be a factor cost vector (V’ denotes its transpose), the relationship along the column of the input-output table can be expressed as

X’ = i’Z + V’.

Let B be the matrix whose row is equal to each row of the input-output matrix Z divided by the output of each sector. The entry of the matrix B={b_ji} in the j-th row and in the i-th column represents the ratio of the i-th sector’s usage of the j-th sector’s output to the entire output of the j-th sector.

(1) X’ = = X’B + V’

Thus the entry of the matrix B={b_ji} in the j-th row and in the i-th column shows us how the decrease in the output in the j-th entry of X on the right-hand side leads to the decrease in the output of the i-th entry of X’ on the left-hand side. In this sense, each entry of the matrix B shows the magnitude of the first-stage forward linkage effect. If this propagation of the forward linkage persists, the cumulative sum of the effects can be obtained by using an inverted matrix and solving for X’ in (1).

We denote the inverted matrix G. That is,

Using this new notation to re-write the equation above yields,

This is the basic idea of the forward linkage explained in Miller and Blair (2009). For our purpose, we modify this idea as follows. In our analysis, we replace the supply constraint with the decrease in the outputs of the particular sector (the estimated damage in terms of the value of output). This is easier than tracing back the damage to each fundamental factor of production and converting the loss into factor income units. To express this idea, we take advantage of the following relationships.

Combining these two equations, we obtain

This leads to the following.

(3)

Using this relationship, the cumulative impact on the i-th sector’s production by the forward linkage, or the disrupted supply chain, from the decrease in the j-th sector’s production due to the earthquake, can be computed by .

As shown in Appendix 1, in the analysis of the forward linkage above explained, the relationship between the intermediate goods input and the final output is assumed to be represented by the Cobb-Douglas production function. In other words, we assume the production technology with which the supply decrease in some intermediate goods can be substituted by other input goods. For retail industry, for example, this assumption is realistic, as an empty shelf due to the lack of the good from Tohoku region can be filled by the products from the other area and the business can keep running. However, in the manufacturing sector, where the final output consists of various parts, substitution will be difficult, at least in the short run. Right after the Great East-Japan Earthquake, the inability to find the substitute parts of the customized parts was revealed as the supply chain disruptions gathered attention.

Therefore, we compute the first-stage linkage effect by assuming that the decrease in the total output of a particular manufacturer is driven by the bottleneck in production, or the maximum decrease among all the intermediate goods from other manufacturers. The detail of the methodology is presented in Appendix 2. On the other hand, in computing the cumulative impacts after the first-stage linkage, we do not consider this extreme form of bottleneck effect for the second-order effects and later.

When we consider the bottleneck effect in the first-stage linkage, we need to decide whether the damage rates in Kanto and Tohoku regions should be treated jointly or separately. When we treat Tohoku and Kanto regions as one area and use the worst damage rate as the bottleneck, we assume no substitutability between inputs from Tohoku and Kanto industries. That is, even if each region produces the intermediate goods classified as the same sector products we regard they are essentially different goods. As an alternative assumption, by treating Tohoku and Kanto separately and use the maximum damage rate in each region as a measure of the bottleneck effect, we implicitly assume some substitutability in the same sector across different regions. Whether which assumption grasp reality may depend on how much detailed classification we use in input-output tables. To decide this selection problem, we compare two different projection of regional production decrease induced by the first-stage forward linkage under two different assumptions on substitutability to the actual production decline of manufacturing in each region within a few months after the earthquake. As we see in the following Section 4, we find that at least in the short period no substitutability assumption between inputs from Tohoku and Kanto industries captures the reality.

Even though short-term substitutability assumption between inputs from Tohoku and Kanto regions confronts the reality, it provides an interesting simulation result on how much extent supply chain diversification can mitigate the indirect damage caused by the forward linkage effect of supply chain disruptions. We report the result of such simulation in Section 6.