Estimated Damage by Industry in the Disaster-affected Area

In order to estimate the impact of the disrupted supply chain by applying the forward linkage methodology, as described above, we need to estimate the damage by industry in the disaster-affected area caused by the Great East-Japan Earthquake. Then, how do we estimate the damage by industry in the area? We do this by first estimating the output by industry in the each disaster-affected city or town, and then by multiplying these figures by the estimated damage rates for each city and town.

We can obtain the number of employees by industry in each city and town from the Economic Census 2009. We also have the country-level output per employee ratio and real net capital stock per employee ratio by industry, from Japan Industrial Productivity Database (JIP2010).¹ If we assume that these two ratios for each industry are same all over Japan, multiplying these ratios and the number of employees by industry in each city and town together gives us the estimated output and real net capital stock for each industry and for each city and town.

We obtain the damage rate for each city and town by applying the same methodology devised by Tomoyoshi Terasaki of Development Bank of Japan. He estimates the loss of capital stock in the earthquake hit four prefectures (Iwate, Miyagi, Fukushima and Ibaraki) dividing each prefecture into coastal area and the inland area. In calculating this estimation, he uses both the human damage rate (the ratio obtained by dividing the death toll and the number of missing and evacuees by the registered population in the area) and the corporate damage rate (the ratio obtained by dividing the affected number of firms reported in the newspaper by the number of corporate offices with the number of employees greater than or equal to 100) for each of coastal and inland area in four prefectures. Next, he multiplies the adjustment coefficient obtained by dividing the surveyed (that is, very close to actual) loss of capital stock in the Great Hanshin-Awaji Earthquake in 1995 by the estimated loss of capital stock applying the above described methodology to the Hanshin-Awaji case. We use the number of human damage rate for each city and town, and using the same numbers for the corporate damage rate and the adjustment coefficient as Terasaki we get the damage rate for each city and town hit by the Great East-Japan Earthquake.

Multiplying the estimated output and real net capital stock by industry at the city and town-level by the above estimated damage rate for each city and town we obtain the value of the damage (both for the output and the real net capital) by industry at the city and town-level. Then we aggregate these values at the city and town-level for three prefectures in Tohoku region (Iwate, Miyagi and Fukushima) to obtain the estimated damage for the Tohoku region. We do the same for Ibaraki prefecture to obtain the estimated damage for the Kanto region.

[Figure 1, Figure 2]

Figure 1 shows the bar plot of the estimated damage on the real net capital stock for each industry. Each bar is the sum of the Tohoku and the Kanto regions and these two regions are shown in different colors. Figure 2 shows the estimates on the output-level damage (at annual level). In both figures, we can confirm that the damage for the Tohoku region exceeds that for the Kanto region and that the Great East-Japan Earthquake hit the Tohoku region particularly hard. Also, Figure1 shows that the damage on the real net capital stock concentrates on non-manufacturing sectors, especially in the electricity industry. This reflects the fact that Fukushima Nuclear Power Plants were seriously damaged and many non-manufacturing firms are located in the coastal area where tsunami hit severely. On the other hand, Figure 2 shows that not only the non-manufacturing sector (such as commerce) saw the loss in the outputs but also the manufacturing sector, in particular the food industry, suffered a lot in terms of their output. The annualized direct damage (where annualized means the value assuming the damage after the earthquake persists at the same level for one year) is estimated to be 6.5 trillion Yen.

4. The Estimation on the Effects of Supply Chain Disruption

4-1 The Regional Propagation Pattern of Supply Chain Disruptions

As mentioned in the above Section 2, we choose short-term non-substitutability assumption of intermediate inputs from the Tohoku and the Kanto regions by comparing estimated regional propagation pattern of supply chain disruptions under two different assumptions with the actually observed regional production decline pattern of manufacturing sector right after the earthquake (as shown in Table 1). Figure 3 shows the comparison between estimated regional propagation pattern of first-stage forward linkage under non-substitutability assumption and the actual production decline right after the earthquake (that is, within 20days after March 11th).

[Figure 3]

Figure 3 shows that our estimates of production decrease is rather underestimation even under the non-substitutability assumption. Referring to only two examples, in the Tohoku region our estimate is 23 percent decline while actual manufacturing decline in the region is 53 percent, and in the Kanto region our estimate of 15 percent decline falls short of actual 30 percent decline. But the similarity of regional propagation pattern of supply chain disruptions to actual pattern (the sole exception is the Chubu region) and the scale of the impact lead us to choose the non-substitutability assumption.

We can suggest a few possible reasons of the underestimation of our estimation method. First, our estimation of the damage does not include damages on the public infrastructure such as road and harbor, which may cause additional influence on production activities in the earthquake hit region. Second, while we compare first-stage forward linkage with actual production decline right after the earthquake, second-stage and further-stage linkage to downstream industries may already take place. Third, actual input-output interconnections of industries may be more complicated; that is, not only forward linkage but also backward linkage may come about at the same time. For example, suppose Company A supplying parts to Company B is hit by the earthquake, which stops production activities of both Company A and Company B. Suppose the Company B also buys the other parts from Company C. Even if the Company C is free from the earthquake damage and is upstream industry, the stop of operation of the Company B results in the production decline of the Company C. In the Chubu region where automobile-related industries are agglomerated this kind of interaction between forward linkage and backward linkage may be biting, which explains exceptional underestimation in this region.

4-2. The Magnitude of the Supply Chain Disruptions

Now let us look at the magnitude of the supply chain disruptions and its influence to each industry. Both first-stage forward linkage and the total forward linkage are calculated. As explained in the above sections first-stage forward linkage is calculated assuming bottleneck effect in the manufacturing industries and no substitutability between Tohoku and Kanto regions, while in calculating further-stage forward linkages we assume no such special conditions.

[Figure 4]

Figure 4shows our calculated result of the first-stage forward linkage effect by industry. A notable feature of the first-stage effect of the supply chain disruptions is that the effect is particularly concentrated in the manufacturing industries while the direct damages by the earthquake itself are highly concentrated in the non-manufacturing industries (as we see in Figure 1 and Figure 2). In the manufacturing sector, the chemical material, steel, general machinery, electric parts and automobile-related industries suffer a large loss, in addition to the food industry which suffer a large direct damage. In particular, the automobile-related industries suffer relatively less in terms of the direct damage by the earthquake while they are affected significantly by the supply chain disruptions. This is due to the fact that a car consists of more than several ten thousand parts and thus car manufacturing depends on a complex division of labors, involving a web of subcontractors. When we aggregate theses first-stage forward linkage effects of the supply chain disruptions the total amount count 27.3 trillion yen per year base. The estimate is about four times as large as that of the direct damage, showing that the supply chain disruption is crucial factor dragging Japanese economy significantly after the disaster.

[Figure 5]

Figure 5 shows the total forward linkage effect of supply chain disruptions; that is, the cumulative effect assuming forward linkage effect continues infinitely. In this calculation, in addition to the industries that suffered a lot from the first-stage effect, chemicals, machinery and metal engineering show the large cumulative effects. It is notable that the automobile-related sector has a significantly large cumulative effect. It is surprising that the estimated damage per year base is as high as 142 trillion yen, a large number even in terms of the gross output.

We should be careful for the interpretation of our estimates. It is the simple aggregation of numbers on gross output base, and based on the unrealistic assumption that the most severe damage to the production right after the earthquake would continue the whole one year without any recovery. To get more realistic numbers we should translate the numbers on gross output base to those on value-added base, and reflect the actual recovery of production in the earthquake hit regions. By multiplying the estimated numbers on gross output base by the ratio of value-added to output for each industry, we can get numbers on value-added base. Our estimated direct production damage by the earthquake is equivalent to 0.7 percent of GDP, first-stage forward linkage effect is 1.7 percent of GDP, and total forward linkage effect is 8.9 percent of GDP.

When we normalize the output loss in the damaged area right after the earthquake in March to be 100, we can calculate the output loss in the following months from the manufacturing productions index. The loss is 62 in April, 33 in May, and 20 in June, showing a sign of recovery from the earthquake. Therefore, adding two thirds of the damage in March (the quake occurred on March 11th) to the loss over the period between April and June gives us 1.82/12 =(1×2/3＋0.62＋0.33＋0.20) /12 of the annualized loss. Converting the annualized estimates to the 4 months estimates reflecting recovery (from March through June) using this fraction yields the direct output loss by the earthquake to be 0.11 percent of GDP, first-stage forward linkage effect to be 0.26 percent of GDP, and total forward linkage effect to be 1.35 percent of GDP. We can still confirm the large impact of forward linkage effect in comparison with the direct production damage by the earthquake.

Since we observe the significant damage of forward linkage effect caused by the supply chain disruptions, it is worthwhile to consider the damage reduction through building multiple supply chains. What is needed first of all for this consideration is the estimation of the benefit of multiple supply chains in the situation of natural disasters. How would be the damage from forward linkage effect when we can count on the multiple supply chains from both Tohoku and Kanto regions. Figure 6 and Figure 7 show respectively the first-stage effect and the total forward linkage effect under such hypothetical situation.

[Figure 6]

Comparing Figure 6 with Figure4, we can see that the substitution in parts supply between Kanto and Tohoku regions reduces the first-stage forward linkage effect significantly. In this case, the first-stage forward linkage effect is estimated to be 6.1 trillion yen (at annual base), which is slightly more than one fifth of the case without such substitution. For the industry-level breakdown, we do not see any difference between relatively more affected industries among the non-manufacturing sector, such as commerce and construction, and the manufacturing industries. The most severely affected sectors among manufacturing include the electric parts, food industries followed by the auto parts industries. Thus, we do not see any prominent first-stage forward linkage effects on the overall automobile-related industries.

[Figure 7]

Figure 7 shows the total forward linkage effect with the substitution possibilities in parts supply between Kanto and Tohoku regions. The impact on the automobile-related industries becomes larger due to its complicated interdependence. However, the magnitude is still similar to the food industry which faced the large direct damage and to some of the non-manufacturing industries which suffered relatively large losses (such as commerce and construction). The total forward linkage effect is 30.5 trillion yen (at annual base), which is slightly more than one fifth of the corresponding value in Figure 5 with no substitution among Kanto and Tohoku regions.

The bottom line is that only by means of diversifying the parts supply sources to two different regions, such as Tohoku and Kanto, we can mitigate the forward linkage effect from supply chain disruptions to the level of one fifth in case of such huge natural disasters as the Great East-Japan Earthquake. Applying this result to the estimated effects from March through June after the 3.11 earthquake, the diversification in parts supply can mitigate the first-stage forward linkage effect to the level of 0.05 percent of GDP, and total forward linkage effect to the level of 0.3 percent of GDP, which are rather small compared with normal economic fluctuations. The benefit from such diversification is important in the case that the recovery from earthquake damage takes time, for the industries with complex web of supply chains such as automobile-related industries.

5. Concluding Remarks

Though Japan’s land has been hit by many tremendous earthquakes in the past, the Great East-Japan Earthquake is unique in the sense that it hit large areas of Tohoku and Kanto where complex web of supply chains exists. This situation raised concern about the propagated consequences of supply chain disruptions right after the earthquake. Our calculation confirms that such concern has a reason because estimated production decline in Japan caused by forward linkage effect from supply chain disruptions is much larger than that caused by the direct earthquake damage.

The experience of the Great East-Japan Earthquake reminds us the importance of damage mitigation of natural disasters, the most vital of which is of course human life. But mitigating the damage to economic activities should be noted. Our calculation suggests that the benefit of supply chain diversification is quite significant.

Some firms may have already started to consider the supply chain diversification as a part of their business continuity plans (BCP), based on the lessons from the Great East-Japan Earthquake. But there also lie difficulties in the realization of such diversification plan. The very reason that some intermediate goods are hard to substitute is that their production requires intangible assets owned by a particular supplier. Therefore, it is not easy to diversify the source of supply of such inputs. Building the system of supply chains that is robust to disasters requires the innovation of the usage of intellectual property rights to circumvent this difficulty at the same time.

It may not make sense for individual firms to spend resources to construct robust supply chains to reduce the damage to one fifth in the rare disaster which occurs only once in a century. Thus, it may be more realistic to make plans about how to recover from the damage smoothly after the disaster. If the prompt recovery in the disaster-affected area is possible, supporting the recovery effort is one of the most effective measures.

However, if the disaster is so serious that it is not easy to achieve a prompt recovery, then one might have to find alternative factories to resume its operation. Without a prescribed plan, it might take more than several months to resume business in the alternative factory. It should be effective to exchange information between the industrial clusters or the regions located by large factories which shares similar production technologies and to make agreements about renting excess spaces in the factories to each other in case of a serious disaster. To this end, it is quite essential that more than one area with vibrant manufacturing industries remain in Japan.

We will explain the assumption about the production function on which the forward linkage model described in Section 2 is based. We start from the equation that yields the forward linkage.

X’ = X’B + V’

Taking difference for the both sides of the equation yields,

Pre-multiplying , we have,

Here, the term corresponds to the standard input coefficient matrix A, as shown in Appendix 3. Thus we can rewrite the equation above such that,

In other words, the following will hold.

Let a_ji be the entry in the j-th row and the i-th column of the input coefficient matrix A. Then the effect of the change in the first term of the right-hand side on the i-th sector in the left-hand side can be computed as

(A-1)

Let us rewrite X_j on the right-hand side as Z_j to clarify that it is an input, then the equation becomes,

In other words,

Therefore, we can see that this model is based on the Cobb-Douglas production function with coefficients a_ji.