Externalities Spillover Costs & Spillover Benefits Chapter 10 (pages 203-208)
The Regional Spillover E ects of the Tohoku Earthquake...The Regional Spillover E ects of the Tohoku...
Transcript of The Regional Spillover E ects of the Tohoku Earthquake...The Regional Spillover E ects of the Tohoku...
The Regional Spillover Effects of the Tohoku
Earthquake∗
Robert Dekle, Eunpyo Hong, and Wei XieDepartment of Economics, Kaprielian Hall, USC
March 3, 2014
Abstract
In this paper, we trace out how a decline in industrial productionin one region can be propagated throughout Japan. We model Kantoas the dominant region, which includes the Tokyo area, and then use themodel to analyze how a shock to industrial production in Tohoku–owing tothe earthquake–can be propagated throughout Japan. In our econometricmodel, regions and industries within regions are linked by an input-outputstructure and this input-output structure disciplines how the shocks arespatially propagated.
1 Introduction
On March 11, 2011, a devastating earthquake and tsunami hit the Tohoku andNorthern Kanto regions of Japan. The damage was mostly concentrated in theIwate, Miyagi, and Fukushima prefectures. In particular, all three prefectureswere swept by the tsunami, with much of the immediate damage caused by thetsunami. In the areas impacted by the tsunami, industrial production declinedby over 95 percent between March and July of 2011.
Nearly 23000 people were killed (or missing) in these prefectures; and inthe days after the earthquake, about 125,000 people (or 2 percent of the threeprefectures’ populations) evacuated. Destruction to the capital stock was esti-mated to be about $180 billion, or 10 percent of the total capital stock in thethree prefectures.
The overall weight of Iwate, Miyagi, and Fukushima in Japan is small, com-prising about 4 percent of both the Japanese population and GDP in 2010.Still the immediate impact of the earthquake and tsunami on Japanese aggre-gate production was huge, with the negative effect on aggregate GDP lingering
∗We thank the helpful comments from Professors Etsuro Shioji and other participants fromthe March 2013 Gakushuin Workshop on the Economic Recovery from the Great East JapanEarthquake. We deeply thank the financial support from the Japan Foundation Center forGlobal Partnership grant, without which this project would not have been possible.
1
on for a year or more. This is because these three prefectures were major pro-ducers of electronics and other intermediate parts used for production in otherJapanese regions (and even the world), and the stoppage in production of theseintermediate parts meant that production of the final goods in the electronics,automotive, and other industries were stalled all over Japan. For example, To-hoku accounted for 42 percent of the micro-semiconductors and 40 percent ofthe flat screen filters used in the Japanese production of automobiles and cellphones.
The importance of this collapse in Tohoku intermediate input production canbe seen in how Japan’s aggregate GDP declined in the immediate aftermath ofthe earthquake. Compared to the previous quarter (before the earthquake),Japanese aggregate GDP declined by 1.9 percent in the first quarter of 2011.1
The declines in aggregate consumption and inventories contributed 0.9 and 0.6percent to the overall GDP decline, respectively.2 Inventories dropped sharply,as firms nationwide dug into their inventories to supply the intermediate parts–disrupted by the earthquake–necessary for production.
In subsequent quarters, while consumption recovered, inventories continuedtheir depletion. Between the last quarter of 2010 and the third quarter of 2012,aggregate GDP grew by 0.5 percent. Aggregate consumption contributed 1.2percent to this growth, but the depletion of inventories and the decline in netexports contributed to dragging down GDP by 0.6 percent and 1.8 percentbetween the last quarter of 2010 and the third quarter of 2012.3
In this paper, we trace out how a decline in industrial production in oneregion can be propagated throughout Japan. We take Kanto (Tokyo) as thedominant region and explain how a shock to industrial production in Tohoku–owing to the earthquake–can be propagated throughout Japan. Kanto (Tokyo)was chosen as the ”dominant” region, since we estimate our model using datafrom 1998, and during most of this period, more and large shocks have emanatedfrom Kanto than from other regions.4Our maintained a priori assumption is that
1It is important, however, to keep the magnitude of the impact of the Tohoku earthquakein perspective. In fact, the negative impact of the global financial crisis in late 2008 onoverall Japanese GDP was far larger than the negative impact of the Tohoku earthquake.Moreover, how the 2008 global financial crisis caused the Japanese recession at that time isvastly different from how the Tohoku earthquake caused the latest Japanese recession.
While the recession after the 2008 financial crisis was caused by a decline in Japanese invest-ment and an exogenous fall in exports, owing to a collapse in foreign demand, the recessionpost-earthquake was related to the inability of Japan to produce inputs to production, suchas intermediate products and energy, which led to a drawdown in inventories, a decline in theability to supply exports, and the increased imports of raw materials.
2Let GDP=C+I. Then in an accounting sense, the contribution of variable C to the growthin GDP is approximately (C/GDP)*∆C/C.
3During this longer period, net exports declined because of the fall in total exports andthe increase in total imports. The decline in total exports contributed to dragging down GDPgrowth by 0.6 percent and the rise in total imports contributed to dragging down GDP growthby 1.2 percent. Much of the increase in imports was driven by the increase in natural gasand other fossil fuel imports. Energy imports increased, since Japan was faced with an energyshortage. The energy shortage was caused by a shutdown of almost all of the country’s nuclearpower plants, which normally provides 30 percent of Japan’s total energy.
4The assumption of choosing Tokyo as a dominant region seems more justifiable when we
2
during the period including the Tohoku earthquake, Kanto can be treated as adominant region. This allows us to consistently estimate separate conditionalerror correction models for different regions of Japan, which we then combinewith a model for Tohoku to solve for a full set of spatio-temporal impulse re-sponse functions. Conditional impulse response analysis traces out the effects ofshocks over time. However, with a spatial dimension, dependence is both spa-tial and temporal. In our impulse responses using our econometrically estimatedmodel, we trace out the effects from a shock to Tohoku.
In our econometric model, regions and industries within regions are linked byan input-output structure and this input-output structure disciplines how theshocks are spatially propagated. Our emphasis on the input-output structure inthe propagation of shocks after the Tohoku earthquake is motivated by the factthat much of the immediate impact of the Tohoku earthquake on other regionswas driven by the decline in intermediate inputs produced in Tohoku. Theshocks to Tohoku are propagated spatially to other regions. The other regionsin turn impact still other regions with a delay. We also allow these lagged effectsto echo back to Tohoku.
This is not the first paper to trace out the effects of the earthquakes andother natural disasters on Japanese output and industrial production. Davisand Weinstein (2002), Okazaki, Ito, and Imaizumi (2002), and Tokui, Kawasaki,and Miyagawa (2013) examine how the distribution of economic activity withinJapan are impacted by natural disasters. Uchida, Miyakawa, Hosono, Ono, andUesugi (2013) examine how shocks arising from earthquakes, when interactedwith fianncing constraints, can lower firm-level industrial production.
2 The Impact of the 2011 Tohoku Earthquakeon Aggregate and Regional Industrial Produc-tion.
GDP includes a sizable component of non-manufacturing production, in-cluding the production of services. To better isolate the impact of the disruptionof the production of parts in Tohoku on Japanese manufacturing production, forthe remainder of the paper, we focus on the measure of industrial production,which mainly captures manufacturing production. Figure 1 depicts the patternin industrial production from the third quarter of 2008 to the third quarter of2012. We can observe that disruptions owing from the Lehman crisis sharplylowered Japanese aggregate industrial production in the first quarter of 2009.Compared to the decline in production from the Lehman crisis, the decline inproduction from the earthquake was far milder.
This aggregate pattern, however, masks the wide regional disparities inthe impact of the earthquake. Not surprisingly, the decline in production inTohoku was far larger during the earthquake than during the financial crisis.
are estimating our model using data from 1998. In an earlier version of our paper, we choseTohoku as the ”dominant” region in our estimates, but our impulse responses were unaffected.
3
The impact of the earthquake was much more regionally concentrated than theimpact of the financial crisis.
In Figure 2, we show a map when the 47 prefectures are aggregated into8 regions. We aggregate the prefectures up to this level, since the input-outputtables that we use extensively below are only available at this regional break-down. With this aggregation, Tohoku now includes Aomori, Akita, and Yam-agata, in addition to the three heavily impacted prefectures of Iwate, Miyagi,and Fukushima. The Kanto region includes Japan’s largest cities of Tokyo andYokohama (Kanagawa); and the Chubu region includes the important heavymanufacturing prefectures of Aichi and Shizuoka. In this aggregation, sinceChubu also includes the Hokuriku region, Chubu also turns out to be adjacentto Tohoku.
Figures 3(a) and 3(b) plot the monthly regional industrial production indices(seasonally adjusted) from 1998 to 2012 for the eight regions. Compared toFebruary 2011, industrial production in Tohoku fell by 35 percent in March2011. This decline in industrial production was much steeper than the post-financial crisis decline of 28.6 percent (between December 2008 to February2009) in Tohoku.
While the decline was not as steep as during the financial crisis, productiondeclined sharply post-earthquake in Kanto and Chubu. In March 2011, indus-trial production fell by 20 percent in Kanto and 25 percent in Chubu. TheKanto prefectures of Chiba, Saitama, Ibaragi, Tochigi, and Tokyo were directlyimpacted by the earthquake, but not the tsunami, so the direct damage to theircapital stock was minimal. However, the Kanto region has many factories usinginputs produced in the Tohoku region, so production was halted in many of thefactories. Likewise, the Chubu region is Japan’s industrial heartland, and manyof the factories located there such as the automobile factories used inputs madein Tohoku.
Despite its geographic proximity to Tohoku, Hokkaido was spared of muchof the impact of the earthquake. Kyushu, Shikoku, Kinki, and Chugoku are alllocated far from Tohoku. While Chugoku and Shikoku’s industrial productiondeclined after the earthquake, Kyushu’s industrial production, while decliningslightly after the earthquake has bounced back strongly. It is said that Kyushuproduces many products that are substitutes to Tohoku’s, so that Kyushu was infact a beneficiary of the damage to Tohoku’s production facilities. Surprisingly,Kinki, while including the industrial cities of Osaka and Kobe, was spared ofthe direct effects from the supply disruption of the intermediate parts producedin Tohoku.
3 Indices of Interactions Among Japanese Re-gions.
As discussed above, the earthquake to Tohoku affected different regions in dif-ferent ways. Some regions like Kanto and Chubu experienced a sharp fall in
4
industrial production, while industrial production in Kinki, Chugoku, and otherSouthern regions barely budged. We have argued that the different propagationmechanisms in industrial production may be related to how different regionsused the inputs produced in Tohoku or were substitutes to the inputs producedin Tohoku.
In this Section, using input-output matrices that include 17 industries in our8 regions, we show how the different regions in Japan are ”interrelated.” Weconsider three measures of ”interrelatedness.” The 17 industries and 8 regionsare depicted in Table 1. The measures of ”interrelatedness” are: 1) how tworegions are ”similar” (Conley and Dupor (2003)) 2) how much two regions buyfrom each other; and 3) the geographical adjacency of two regions.
3.1 ”Interrelatedness” Measures
We use the Japanese regional input-output matrices for 2005 compiled by RI-ETI, in which there are N = 8 regions. The raw input-output matrices includesrows (suppliers of commodities) and columns (purchasers of commodities) thatdo not correspond to any industries. On the column side, besides intermedi-ate users of commodities such as manufacturing, mining, and construction, theinput-output table contains columns for other components of gross domesticproduct: consumption, investment, change in business inventories, and govern-ment purchases. On the row side, the input-output table contains rows forcompensation to nonindustries such as wages and taxes. We address these com-ponents of the regional input-output table by: (a) removing all the final-usecolumns of the input-output table; and (b) dropping all additional rows of thetable. Finally, the original matrix has 29 industries, but we drop ”public ad-ministration”, ”medical services”, ”business services”, ”personal services”, and”others”, to arrive at M = 17 industries, which are primarily in manufacturing.
3.1.1 Notation
Γ is the input-output matrix of dimension N ×M by N ×M . A typical (s, b)-th
element of Γ is Γ(s, b), which is the total value of transactions between s’s sup-ply and b’s purchase. In other words, the s-th row of Γ corresponds to the valueof sales of s, and the b-th column of Γ corresponds to the value of purchases of b.
For i, j = 1, · · · , N and m,n = 1, · · · ,M , denote Γ(i(m), j(n)
)as the total
value of sales from region-i’s industry-m to region-j’s industry-n.
3.1.2 ”Similarity” Regional Matrix
This economic distance measure (modified from Conley and Dupor (2003)) holdsthat two regions are close if they buy goods from similar industries. We use the
5
argument that regions with similar input requirements are likely to have similartechnology; so that the same shock to a given region is likely to affect the outputof another ”similar” region.
Steps to compute the ”similarity” matrix.
• calculate Bm
Bm(i, j) =Γ(i(m), j(m)
)∑k Γ(k(m), j(m)
)• calculate B
B(i, j) =∑m
Bm(i, j)
• calculate Db
Db(i, j) =
{∑k
[B(k, i)−B(k, j)]2
}1/2
for i, j = 1, ....N . This matrix is depicted in Table 2(a). According tothis matrix, prefectures most related to Tohoku (in order) are: Kanto,Shikoku, Hokkaido, Kinki, Chubu, Chugoku, and Kyushu.
3.1.3 Mutual Buying Regional Matrix
Our second measure of ”interrelatedness” measures how much two regions arebuying from each other, relative to their purchases from other regions. The morethe two regions are buying from each other, the more dependent or ”interrelated”are the two regions.X with (i, j)-the element
X (i, j) =
∑m,n Γ
(i(m), j(n)
)∑k,m,n Γ
(k(m), j(n)
) +
∑m,n Γ
(i(m), j(n)
)∑l,m,n Γ
(i(m), l(n)
)The first term is the weight of sales from region i to region j among all the
regions’ sales to region j. The second term is the weight of purchases by regionj from region i among all the regions’ purchases from region i. This matrix isdepicted in Table 2(b). According to this matrix, prefectures most related toTohoku (in order) are: Kanto, Chubu, Kinki, Hokkaido, Chugoku, Kyushu, andShikoku.
3.1.4 Contiguity Matrix
The last matrix of ”interrelatedness” simply assigns a value of one if theregion shares a border with another region, deeming that if they share a border,they are ”similar.” This matrix is depicted in Table 2(c). According to thismatrix, prefectures most related to Tohoku (in order) are: Hokkaido, Kanto,Chubu, Kinki, Chugoku, Shikoku, and Kyushu.
6
4 Regional Spillover Effects
4.1 Model of Regional Spillover Effects
Holly, Pesaran and Yamagata (2011) designed a method for analyzing the spatialand temporal diffusion of shocks to a dominant region, which was applied toevaluate the effects on UK housing prices due to shocks on the housing priceto London. The method treats the house price of London as a common factorand then models the contemporaneous as well as lagged dependencies amongregions conditional on London house prices.
We employ the diffusion model of Holly, Pesaran and Yamagata (2011) toassess the shock of Tohoku earthquake on the other regions in Japan. We makeuse of monthly data of industrial production for the 8 Japan regions defined inthe previous section. The data ranges from January 1998 to October 2012, sothat T = 178.
Denote pit as the industrial production data of region i at time t, for i =1, · · · , N and t = 1, · · · , T . The diffusion model treats the dominant region (i =1) and the rest of the regions (i = 2, · · · , N) differently by allowing for the shockon the dominant region to affect the other regions not only contemporaneouslybut also through lagged impacts, while allowing for no contemporaneous effectsfrom the rest of the regions on the dominant region.
For regions i = 2, · · · , N ,
∆pit = φis(pi,t−1 − p̄si,t−1
)+ φi1 (pi,t−1 − p1,t−1) + ai
+
kia∑l=1
ail∆pi,t−l +
kib∑l=1
bil∆p̄si,t−l +
kic∑l=1
cil∆p1,t−l + ci0∆p1t + εit (1)
For region i = 1, φ11 and c10 are set to be 0 in the above equation (1), where
p̄sit =
N∑j=1
Sijpjt, with
N∑j=1
Sij = 1
Sij ≥ 0 is the (i, j)-th element of weighted spatial matrix S, which measuresthe spatial connection between region i and region j. Notice that S is rowstandardized in that each row of S sums up to 1. In practice, S is estimatedby row standardizing the ”interelatedness” measures defined in the previoussection, namely, the ”Similarity Matrix”, the ”Mutual Buying Matrix”, and the”Contiguity Matrix”.
As pointed out by Holly, Pesaran and Yamagata (2011), the error correctingspecification of equation (1) is a parsimonious representation of pair-wise coin-tegration of the data across regions. In addition, weak exogeneity of ∆p1t inequation (1) can be tested by the procedure of (1973).
In the estimation of the model above, we take Kanto (Tokyo) as the dominantregion. Tokyo’s industrial production is assumed to be only affected by its ownlagged industrial production and the lagged effects of its neighbor’s industrial
7
production. The industrial production of other regions is assumed to be affectedby not only the lagged effects of Tokyo and the remaining regions, but also thecontemporary effects of the shocks to Tokyo.
4.2 Spatio-temporal Impulse Response Functions
We can use the estimates from the model above to examine impulse responsesboth over time and space.The persistence profile of shocks to the system overtime and across regions can be evaluated using generalized impulse responsefunction (GIRF), initially advanced by Pesaran and Shin (1998).
For horizons h = 0, 1, · · ·, the impulse response of a unit (i.e. a standarddeviation) shock on the dominant region is computed as
g1(h) = E(pt+h|ε1t =√σ11,Ft−1)− E(pt+h|Ft−1)
=√σ11ΨhRe1 (2)
where pt = (p1t, · · · , pNt)′ is the vector of industrial production data at
time t, Ft is the filtration of information up to time t, σ11 = var(ε1t), ande1 = (1, 0, · · · , 0)′. By stacking the N regressions in (1), Holly, Pesaran andYamagata (2011) derived that5
∆pt = a + Hpt−1 +
k∑l=1
(Al + Gl)∆pt−l +
k∑l=0
Cl∆pt−l + εt
where a, H, Al, Gl, and Cl are matrices of model parameters. It can besolved from the above expression to get
∆pt = µ+ Πpt−1 +
k∑l=1
γl∆pt−l + Rεt
where k = maxi{kia, kib, kic}, µ = Ra with R = (IN − C0)−1, Π = RH,γl = R(Al + Gl + Cl).
In a VAR form, this implies that
pt = µ+
k+1∑l=1
Φlpt−l + Rεt
where Φ1 = IN + Π + γ1, Φl = γl − γl−1 for l = 2, · · · , k and Φk+1 = −γk.Then for h = 0, 1, · · ·, Ψh in equation (2) is defined as
Ψh =
k+1∑l=1
ΦlΨh−l
5See Holly, Pesaran and Yamagata (2011) for detailed derivations of the generalized impulseresponse function in the spatial temporal model.
8
5 Empirical Results
Kanto (Tokyo) is set as the dominant region in model (1) to account for both ofits contemporaneous and intertemporal impacts. We follow Holly, Pesaran andYamagata (2011) to estimate model (1) equation by equation using OLS, andthe estimation results are reported in Table 3.
We construct the weighted spatial matrix basing on three measures of re-gional ”interrelatedness”.
5.0.1 Total Industrial Production
Table 4, Figures 6-8 , and Figure 9, contain the results when evaluating theTohoku diffusion effects using total industry production data.
Table 4(a) reports the results based on the row standardized ”Similarity”matrix, Table 4(b) reports the results based on the row standardized ”MutualBuying” matrix, and Table 4(c) reports the results based on the row standard-ized ”Contiguity” matrix. We can see that results from Table 4(a), (b), and (c)are similar in the following ways.
”Own lag” is the estimated∑kia
l=1 ail. A positive ”own lag” effect impliesthat the series continues to drift in the same direction as last period, exhibitingeither an upward trend or a downward trend. A negative ”own lag” effectimplies that the series adjusts to last period’s increase by a decrease in thecurrent period, exhibiting a property like mean reverting. Estimation basedon the ”Similarity” matrix identifies the own lag effects of Tohoku, Hokkaido,Chubu, Kinki, Chugokku, and Shikoku to be significant. Estimations based onthe ”Mutual Buying” matrix and the ”Contiguity” matrix identify the same setof significant own lag effects, ie. own lag effects are only found to be insignificantfor Kanto and Kyushu.
”Neighbour lag” estimates the dynamic spillover effects∑kib
l=1 bil. A positive”neighbour lag” effect implies that the series moves in the same direction as theweighted average of its neighbour in the last period. A negative ”neighbour lag”effect implies the series moves in the opposite direction. Both the estimationbased on the ”Similarity” matrix and the estimation based on the ”MutualBuying” matrix identify the same set of significant neighbour lag effects inHokkaido, Kinki, Chugoku, Shikoku, and Kyushu. Estimation based on the”Contiguity” matrix identifies significant neighbour lag effects in Chubu, Kinki,Chugoku, Shikoku, and Kyushu. Finally, based on all three ”interrelatedness”measures, the estimated neighbour lag effects on all the regions are positive,except for the neighbour lag effect on Tohoku and the neighbour lag effect ofKanto when the Contiguity matrix is used as the ”interrelatedness” measures.
”Kanto lag” is the estimated lagged effect of Kanto. A positive ”Kanto lag”effect implies that the series moved in the same direction as Kanto did in the lastperiod. Based on all the connectedness measures, the estimated ”Kanto lag”effects are found to be significantly positive for Tohoku. Significantly positive
9
”Kanto lag” effects are also observed for Chubu when using the ”Similarity ma-trix” and ”Mutual buying matrix” and for Hokkaido when using the ”Contiguitymatrix”.
”Kanto current” is the estimated contemporaneous effect of Kanto, ci0. Apositive ”Kanto current” effect implies that the series simultaneously moves inthe same direction as Kanto. Based on all the connectedness measures, theestimated ”Kanto current” effects are similar, and all of them are significantlypositive.
EC1 is estimated φi1, which is referred to as the error correction term of(pi,t−1 − p1,t−1), the deviations of region i from Kanto. The estimated EC1are similar based on the three connectedness measures, which give a signifi-cantly negative EC1 for Chugoku; the Similarity matrix additionally identifiesthat Tohoku also has a significantly negative EC1. EC2 is the estimated φis,the error correction term of (pi,t−1 − p̄si,t−1), the deviation of region i from itsneighbours. The estimated EC2 based on the three ”interrelatedness measures”identify Chugoku and Shikoku to have significantly negative EC2; the ”MutualBuying matrix” and the ”Contiguity Matrix” both identify Tohoku to have asignificantly negative EC2.
WH-stat is the Wu-Hausman test statistics (Wu, 1973) testing the null hy-pothesis that production changes in the dominant region Kanto are exogenousto production changes in the other regions. The results show that most of theregressions passed the Wu-Hausman test, except for the regression of Hokkaidobased on the ”Contiguity” matrix.
kia, kib, and kic are all selected by the Schwarz Bayesian criterion (SBC).Based on all three ”interrelatedness” measures, SBC selected kia to be equal to 1and kib to be equal to 1 or 2. SBC selected the lag orders kic = 0, producing theestimated ”Kanto lag” effects,
∑kic
l=1 cil, to be 0 for Kinki, Chugoku, Shikoku,and Kyushu.
Figure 6, Figure 7, and Figure 8 plot the estimated generalized impulseresponse functions (GIRF) caused by a 1 unit (i.e. 1 standard deviation) positiveshock to Tohoku’s industrial production. Figure 6, Figure 7, and Figure 8 allowus to observe a similar pattern across regions.
The persistence profile of Tohoku shows that it takes about 2 years forTohoku to fully absorb a positive unit of shock to its monthly IP level. Thepersistence profiles of all the regions other than Tohoku show that a positiveunit of shock to Tohoku can be quickly and well adjusted within about a year’stime. Figure 6, Figure 7, and Figure 8 also reveal similar persistence profiles interms of the shape and the degree of the persistence basing on the three different”interrelatedness” measures.
For selected time periods h = 0, 3, 5, 10, 20, 50, Figure 9 depicts the GIRFacross regions and over time. The results are estimated for the ”Similarity,””Mutual Buying,” and ”Contiguity” matrices. The regions are ordered on thehorizontal axis from left to right according to their ”closeness” (according toeach of the three ”interrelatedness” matrices) to Tohoku. For example, in Figure9(a), according to the ”similarity” matrix, the ordered horizontal axis shows thatthe ”closest” region to Tohoku is Kanto, followed by Hokkaido, Kinki, Shikoku,
10
Chubu, Chugoku, and Kyushu.If ”proximity”–according to the various definitions– results in higher spillovers,
then we should see a declining pattern in the graphs. As the regions becomefurther from Tohoku, the impact of the Tohoku shock should dissipate. In gen-eral we see no such pattern in the graphs, except for Kanto. A positive Tohokushock always tends to raise Kanto and Shikoku industrial production. WhileKanto is ”close” to Tohoku by all three measures, Shikoku is rather ”far,” butis strongly affected by Tohoku shocks. Chubu is relatively ”close” to Tohoku,but its total industrial production is scarcely affected by Tohoku’s.
5.0.2 Key Industries
There certainly have been other shocks hitting Japan in the few months follow-ing the 2011 earthquake. There was a nationwide adverse demand shock in theSpring and Summer of 2011, owing to negative consumer sentiment. The regu-lation and forced conservation of electrical power usage throughout Japan afterthe earthquake lowered production in other sectors. In principle, the impulseresponse methodology above controls for these other factors, since the impulseresponse shock is supposed to identify only the partial effect on other regionsfrom the adverse earthquake shock to Tohoku.
To focus more on the role of Tohoku’s intermediate products productionon the output in the rest of Japan, below we trace out the shocks emanatingfrom the decline in the production of electrical and automotive parts in Tohoku.Tohoku was an important producer of the two industries, and other regions usedthe output of these industries from Tohoku as intermediate goods.
Electric Machinery Industry Tables 5(a) and (b), Figure 4, and Figures10-12 contain the results when evaluating the effects of the Tohoku earthquakeusing production data of only the Electric Machinery Industry. Tables 5(a) and5(b) report the results based on the ”Mutual Buying” and ”Contiguity” matrices(the ”Similarity” matrix is undefined when we have only one industry.) All ofthe estimates in the Tables appear reasonable.
Figure 10 depicts the effects of a one standard error positive shock to electri-cal industry production in Tohoku. The impacts are relative large and persistentin Hokkaido, Chubu, Shikoku, and Kyushu. The effects are relatively small inKinki and Hokkaido. In Figure 12, we order the regions according to theirproximity to Tohoku in the electrical machinery industry. As expected from theimpulse responses, the effects to Kanto and Kinki are small, while they are largeand persistent for Hokkaido, Chubu, Shikoku and Kyushu. It appears that theshocks to the electrical machinery industry had relatively large and importanteffects on the electric machinery industries throughout Japan, except for thevery large (in terms of GDP and size) regions of Kanto and Kinki, where theeffects were minimal.
11
Transportation Equipment Industry Tables 6(a) and (b), Figure 5, andFigures 13-15 contain the results when evaluating the effects of the Tohokuearthquake using production data of only the Transportation Industry. Tables6(a) and 6(b) show that the effects are relatively large for Chubu, Kinki, andKyushu, although the effects dissipate quickly for Kinki. In Figure 15, weorder the regions according to their proximity to Tohoku in the transportationequipment industry. With regards to the effect on Chubu, a major automotiveproducer, the immediate effects are large, but they dissipate after one period,then the effects grow again. The effects are relatively large for Kyushu andHokkaido through all periods.
6 Conclusion
In this paper, we traced out how a decline in industrial production in one re-gion can be propagated throughout Japan. We take Kanto as the dominantregion and explain how a shock to industrial production in Tohoku–owing to theearthquake–can be propagated throughout Japan. In our econometric model,regions and industries within regions are linked by an input-output structureand this input-output structure disciplines how the shocks are spatially propa-gated.
Our emphasis on the input-output structure in the propagation of shocksafter the Tohoku earthquake is motivated by the fact that much of the im-pact of the Tohoku earthquake on other regions was driven by the decline inintermediate inputs produced in Tohoku.
In general, while we definately find effects on industrial production from theTohoku earthquake, the regional effects do not seem to depend on our defini-tions of proximity, although we observe significant heterogeneity in how differentprefectures were affected by the spillovers from the Tohoku earthquake.
References
[1] Conley, T. and Dupor, B. (2003), “A spatial analysis of sectoral comple-mentarity,” Journal of Political Economy, 111(2):311-352.
[2] Davis, D. and D. Weinstein (2002), “Bones, Bombs and Break Points:The Geography of Economic Activity,” American Economic Review, 92:1269-1289.
[3] Holly, S., M. Pesaran, and T. Yamagata (2011), “The spatial and temporaldiffusion of house prices in the uk,” Journal of Urban Economics, 69(1):2-23.
[4] Okazaki, T., K. Ito, and A. Imaizumi (2009), “The Impact of NaturalDisasters on Industrial Agglomeration: The Case of the 1923 Great KantoEarthquake,” working paper, University of Tokyo.
12
[5] Pesaran, M.H. and Y. Shin (1998), “Generalized impulse response analysisin linear multivariate models,” Economics Letters, 58(1):17-29.
[6] Tokui, J., K. Kawasaki, and T. Miyagawa (2014), “Reallocation and Re-gional Economies in Japan: Towards an Application to the Great-EastJapan Earthquake,” working paper, Gakushuiin University.
[7] Uchida, H., I. Miyakawa, K. Hosono, A. Ono, H. Uchino, and I. Uesugi(2013), “Natural Disasters and Natural Selection,” working paper, Uni-versity of Tokyo.
[8] Wu D. (1973), “Alternate tests of independence between stochastic re-gressors and disturbances”, Econometrica, 41(4):733-50.
13
Results of May 6
Data 1998 to 2011 for IP, Electric Industry, and Transport Industry
Table 1: Regions and Industries(a) Regions
01 Hokkaido02 Tohoku03 Kanto04 Chubu05 Kinki06 Chugoku07 Shikoku08 Kyushu + Okinawa
(b) Industries
020 Mining030 Beverages and Foods040 Textile products050 Timber, wooden products and furniture060 Pulp, paper, paperboard, building paper070 Chemical products080 Petroleum and coal products090 Plastic products100 Ceramic, stone and clay products110 Iron or steel products120 Non-ferrous metal products130 Metal products140 General machinery150 Electrical machinery160 Transportation equipment170 Precision instruments180 Miscellaneous manufacturing products
1
Table 2: Distance Measures(a) Similarity Matrix
Kanto Tohoku Hokkaido Chubu Kinki Chugoku Shikoku Kyushu
Kanto 0 11.366 12.227 13.034 12.703 14.064 12.999 13.905Tohoku 11.366 0 11.843 12.612 11.996 13.105 11.998 13.295
Hokkaido 12.227 11.843 0 12.352 12.127 13.169 12.046 13.476Chubu 13.034 12.612 12.352 0 11.497 12.993 12.050 13.356Kinki 12.703 11.996 12.127 11.497 0 11.648 9.848 12.287
Chugoku 14.064 13.105 13.169 12.993 11.648 0 10.812 12.414Shikoku 12.999 11.998 12.046 12.050 9.848 10.812 0 11.796Kyushu 13.905 13.295 13.476 13.356 12.287 12.414 11.796 0
(b) Mutual Buying Matrix
Kanto Tohoku Hokkaido Chubu Kinki Chugoku Shikoku Kyushu
Kanto 0 0.269 0.160 0.261 0.182 0.106 0.109 0.166Tohoku 0.361 0 0.038 0.080 0.062 0.032 0.016 0.037
Hokkaido 0.239 0.045 0 0.107 0.059 0.019 0.011 0.021Chubu 0.256 0.117 0.063 0 0.206 0.097 0.076 0.140Kinki 0.224 0.083 0.043 0.205 0 0.130 0.147 0.106
Chugoku 0.147 0.039 0.025 0.120 0.165 0 0.109 0.119Shikoku 0.187 0.025 0.014 0.096 0.137 0.070 0 0.076Kyushu 0.147 0.029 0.011 0.083 0.107 0.096 0.056 0
(c) Contiguity Matrix
Kanto Tohoku Hokkaido Chubu Kinki Chugoku Shikoku Kyushu
Kanto 0 1 0 1 0 0 0 0Tohoku 1 0 1 1 0 0 0 0
Hokkaido 0 1 0 0 0 0 0 0Chubu 1 1 0 0 1 0 0 0Kinki 0 0 0 1 0 1 1 0
Chugoku 0 0 0 0 1 0 1 1Shikoku 0 0 0 0 1 1 0 1Kyushu 0 0 0 0 0 1 1 0
2
Table 3: Distance Measures(a) Electric Machinery Mutual Buying Matrix
Kanto Tohoku Hokkaido Chubu Kinki Chugoku Shikoku Kyushu
Kanto 0 0.270 0.426 0.254 0.262 0.238 0.230 0.240Tohoku 0.364 0 0.122 0.090 0.053 0.046 0.083 0.075
Hokkaido 0.538 0.031 0 0.056 0.096 0.037 0.056 0.062Chubu 0.308 0.090 0.022 0 0.255 0.123 0.091 0.173Kinki 0.381 0.117 0.071 0.170 0 0.107 0.202 0.118
Chugoku 0.356 0.021 0.003 0.152 0.184 0 0.010 0.038Shikoku 0.476 0.066 0.129 0.038 0.076 0.019 0 0.175Kyushu 0.398 0.030 0.007 0.096 0.118 0.085 0.067 0
(b) Transportation Mutual Buying Matrix
Kanto Tohoku Hokkaido Chubu Kinki Chugoku Shikoku Kyushu
Kanto 0 0.215 0.127 0.345 0.166 0.093 0.061 0.329Tohoku 0.622 0 0.001 0.096 0.040 0.013 0.000 0.084
Hokkaido 0.133 0.007 0 0.734 0.026 0.013 0.000 0.038Chubu 0.291 0.447 0.496 0 0.412 0.222 0.063 0.353Kinki 0.181 0.019 0.050 0.320 0 0.269 0.266 0.126
Chugoku 0.097 0.058 0.176 0.071 0.149 0 0.522 0.215Shikoku 0.053 0.027 0.001 0.075 0.267 0.167 0 0.120Kyushu 0.188 0.020 0.004 0.204 0.019 0.092 0.044 0
3
Tab
le4:
Est
imat
ion
resu
lts
ofre
gion
spec
ific
diffu
sion
equa
tion
for
Tot
alIn
dust
rial
Pro
duct
ion
(a)
Sim
ilari
tyM
atr
ix
Ow
nLag
Nei
ghbL
Kanto
Lag
Kanto
Curr
ent
EC
1E
C2
Wu-H
aus
kia
kib
kic
Kanto
-0.3
(-2
.624
)0.8
8(
6.1
49
)-
--
--
12
-Tohoku
-0.2
62
(-3
.253
)-0
.026
(-0
.168
)0.4
42
(3.2
2)
0.8
15
(11
)-
-0.2
2(
-3.6
53
)-0
.815
11
1H
okka
id-0
.478
(-6
.518
)0.3
94
(4.0
42
)-
0.3
05
(3.6
66
)-
-1.1
34
11
0C
hubu
-0.2
17
(-3
.031
)0.4
71
(3.8
18
)-
1.0
12
(14.4
97
)-
-0.2
63
11
0K
inki
-0.5
41
(-4
.785
)0.5
06
(5.4
37
)-
0.5
13
(7.7
52
)-
-0.1
08
(-2
.343
)0.0
93
21
0C
hugoku
-0.1
19
(-1
.394
)0.3
07
(2.8
6)
-0.6
29
(7.4
24
)-
-0.3
7(
-4.5
61
)-0
.632
11
0Shik
oku
-0.5
9(
-4.7
79
)0.5
19
(3.3
19
)-
0.5
26
(4.1
92
)-
-2.0
99
21
0K
yush
u-0
.059
(-0
.658
)0.1
(0.6
22
)0.3
66
(2.8
96
)0.7
85
(10.8
89
)-
-0.0
7(
-2.1
38
)1.0
39
11
1
(b)
Mutu
alB
uyin
gM
atr
ix
Ow
nLag
Nei
ghbL
Kanto
Lag
Kanto
Curr
ent
EC
1E
C2
Wu-H
aus
kia
kib
kic
Kanto
-0.3
09
(-2
.566
)0.8
68
(6.0
22
)-
--
--
12
-Tohoku
-0.2
54
(-3
.024
)0.4
45
(4.1
32
)-
0.8
1(
11.3
91
)-
-0.2
54
(-3
.609
)0.3
87
11
0H
okka
id-0
.482
(-6
.703
)0.4
01
(4.5
28
)-
0.3
28
(4.1
)-
-0.5
27
11
0C
hubu
-0.2
62
(-3
.432
)0.5
44
(4.1
53
)-
1.0
01
(14.3
83
)-
-0.1
42
11
0K
inki
-0.5
9(
-5.1
11
)0.5
29
(6.1
13
)-
0.5
22
(7.7
6)
--
-0.9
06
21
0C
hugoku
-0.1
74
(-2
.048
)0.3
3(
3.1
02
)-
0.6
25
(7.1
5)
--0
.251
(-3
.569
)-0
.912
11
0Shik
oku
-0.6
17
(-4
.993
)0.5
63
(3.6
98
)-
0.5
19
(4.1
87
)-
-1.6
42
10
Kyush
u-0
.252
(-2
.564
)0.6
07
(5.2
83
)-
0.7
8(
10.9
96
)-
-0.8
72
21
0
(c)
Conti
guity
Matr
ix
Ow
nLag
Nei
ghbL
Kanto
Lag
Kanto
Curr
ent
EC
1E
C2
Wu-H
aus
kia
kib
kic
Kanto
-0.1
59
(-1
.306
)0.5
76
(4.8
2)
--
--
-1
2-
Tohoku
-0.2
16
(-2
.628
)-0
.152
(-1
.038
)0.5
(3.4
44
)0.8
31
(11.8
98
)-
-0.2
86
(-4
.098
)0.0
91
11
Hokka
id-0
.454
(-6
.345
)-0
.089
(-1
.064
)0.4
44
(2.5
37
)0.3
73
(4.6
96
)-
--0
.699
11
1C
hubu
-0.1
93
(-2
.603
)0.3
69
(3.2
)-
1.0
42
(15.0
52
)-
--0
.458
11
0K
inki
-0.5
7(
-5.0
56
)0.1
14
(1.2
45
)0.3
63
(3.6
63
)0.5
74
(8.5
78
)-
-0.1
04
(-2
.605
)-1
.812
21
1C
hugoku
-0.1
85
(-2
.303
)0.2
76
(2.9
16
)-
0.6
31
(6.8
88
)-0
.148
(-2
.882
)-
-1.6
49
11
0Shik
oku
-0.2
92
(-3
.81
)-0
.371
(-1
.554
)0.5
34
(2.8
55
)0.6
77
(4.9
25
)-0
.097
(-2
.359
)-
-0.5
43
12
1K
yush
u-0
.252
(-2
.681
)0.1
46
(2.1
42
)0.4
31
(2.8
38
)0.8
31
(11.5
97
)-
--1
.027
21
1
Not
e:t-
stat
isti
csin
pare
nthe
ses.
Kan
to’s
lagg
edeff
ect
are
esti
mat
edto
be0
and
thus
omit
ted
from
the
repo
rt.
Lag
orde
rsar
ese
lect
edse
para
tely
bySc
hwar
zB
ayes
ian
crit
erio
nfr
oma
max
imum
lag
orde
rof
4.
4
Tab
le5:
Est
imat
ion
resu
lts
ofre
gion
spec
ific
diffu
sion
equa
tion
for
Ele
ctri
cM
achi
nery
(a)
Ele
ctri
cM
ach
iner
yM
utu
alB
uyin
gM
atr
ix
Ow
nLag
Nei
ghbL
Kanto
Lag
Kanto
Curr
ent
EC
1E
C2
Wu-H
aus
kia
kib
kic
Kanto
0.1
9(
1.5
46
)0.1
72
(2.0
09
)-
21
-Tohoku
-0.2
78
(-3
.267
)0.3
46
(2.6
63
)0.5
95
(5.4
85
)-0
.224
(-3
.338
)0.1
89
11
0H
okka
id-0
.166
(-2
.033
)0.5
12
(2.2
01
)0.6
04
(3.0
8)
-0.0
91
(-2
.121
)0.4
54
11
0C
hubu
-1.0
11
(-9
.304
)1.2
74
(5.6
88
)0.2
03
(1.3
92
)-0
.153
22
0K
inki
-0.3
44
(-4
.38
)0.1
25
(0.9
32
)0.3
14
(1.0
45
)0.4
74
(5.1
75
)0.0
71
11
1C
hugoku
-0.1
49
(-1
.731
)0.5
02
(3.0
13
)0.5
41
(4.9
16
)-0
.093
(-2
.278
)-0
.525
12
0Shik
oku
-0.1
91
(-2
.329
)0.3
37
(1.9
93
)0.6
08
(4.2
76
)0.0
66
11
0K
yush
u-0
.186
(-2
.325
)0.7
28
(3.1
4)
0.4
(2.8
22
)-0
.103
(-2
.471
)1.0
21
12
0
(b)
Conti
guity
Matr
ix
Ow
nLag
Nei
ghbL
Kanto
Lag
Kanto
Curr
ent
EC
1E
C2
Wu-H
aus
kia
kib
kic
Kanto
0.2
64
(2.3
32
)0.0
74
(1.3
59
)-
21
-Tohoku
-0.3
14
(-3
.923
)0.0
24
(0.2
23
)0.3
43
(2.6
61
)0.5
92
(5.4
41
)-0
.13
(-2
.945
)-0
.293
11
1H
okka
id-0
.223
(-2
.849
)0.4
85
(2.2
78
)0.4
93
(2.4
45
)1.2
95
12
0C
hubu
-1.0
26
(-9
.262
)1.4
96
(5.4
59
)0.1
05
(0.6
86
)0.4
49
23
0K
inki
-0.3
37
(-4
.298
)0.0
41
(0.5
38
)0.3
75
(1.3
)0.4
8(
5.2
47
)0.8
98
11
1C
hugoku
-0.1
32
(-1
.587
)0.3
67
(2.9
28
)0.5
85
(5.3
39
)-0
.089
(-2
.179
)-0
.175
12
0Shik
oku
-0.1
41
(-1
.764
)0.1
03
(0.6
93
)0.6
12
(4.1
19
)-0
.114
11
0K
yush
u-0
.268
(-3
.489
)0.7
21
(4.3
49
)0.3
64
(2.6
56
)-0
.094
(-2
.497
)0.2
32
12
0
Not
e:t-
stat
isti
csin
pare
nthe
ses.
Kan
to’s
lagg
edeff
ect
are
esti
mat
edto
be0
and
thus
omit
ted
from
the
repo
rt.
Lag
orde
rsar
ese
lect
edse
para
tely
bySc
hwar
zB
ayes
ian
crit
erio
nfr
oma
max
imum
lag
orde
rof
4.
5
Tab
le6:
Est
imat
ion
resu
lts
ofre
gion
spec
ific
diffu
sion
equa
tion
for
Tra
nspo
rtat
ion
(a)
Tra
nsp
ort
ati
on
Mutu
alB
uyin
gM
atr
ix
Ow
nLag
Nei
ghbL
Kanto
Lag
Kanto
Curr
ent
EC
1E
C2
Wu-H
aus
kia
kib
kic
Kanto
0.2
61
(2.1
71
)-0
.07
(-0
.883
)-
21
-Tohoku
-0.1
42
(-1
.742
)0.3
13
(2.4
57
)1.0
49
(12.7
54
)-0
.053
(-1
.815
)-0
.489
11
0H
okka
id-0
.052
(-0
.767
)-0
.091
(-1
.796
)0.9
35
(7.8
39
)-0
.056
(-2
.404
)-0
.786
11
0C
hubu
-0.0
17
(-0
.213
)-1
.67
(-4
.249
)-0
.012
(-0
.048
)-0
.326
(-5
.434
)0.2
99
13
0K
inki
-0.6
8(
-5.6
58
)-0
.044
(-0
.623
)0.2
59
(2.6
55
)0.5
52
(8.4
81
)-0
.118
(-3
.218
)-1
.493
21
1C
hugoku
-0.0
73
(-1
.027
)0.1
54
(1.4
62
)0.7
75
(9.4
41
)-0
.114
(-3
.116
)-0
.445
11
0Shik
oku
-0.2
59
(-3
.353
)0.0
89
(0.7
52
)0.2
57
(2.4
62
)0.9
36
11
0K
yush
u-0
.205
(-2
.561
)-0
.294
(-2
.361
)0.5
19
(3.3
03
)0.9
01
(9.9
44
)-0
.028
(-1
.401
)0.9
17
12
2
(b)
Conti
guity
Matr
ix
Ow
nLag
Nei
ghbL
Kanto
Lag
Kanto
Curr
ent
EC
1E
C2
Wu-H
aus
kia
kib
kic
Kanto
0.2
61
(2.1
71
)-0
.07
(-0
.883
)-
21
-Tohoku
-0.1
42
(-1
.742
)0.3
13
(2.4
57
)1.0
49
(12.7
54
)-0
.053
(-1
.815
)-0
.489
11
0H
okka
id-0
.052
(-0
.767
)-0
.091
(-1
.796
)0.9
35
(7.8
39
)-0
.056
(-2
.404
)-0
.786
11
0C
hubu
-0.0
17
(-0
.213
)-1
.67
(-4
.249
)-0
.012
(-0
.048
)-0
.326
(-5
.434
)0.2
99
13
0K
inki
-0.6
8(
-5.6
58
)-0
.044
(-0
.623
)0.2
59
(2.6
55
)0.5
52
(8.4
81
)-0
.118
(-3
.218
)-1
.493
21
1C
hugoku
-0.0
73
(-1
.027
)0.1
54
(1.4
62
)0.7
75
(9.4
41
)-0
.114
(-3
.116
)-0
.445
11
0Shik
oku
-0.2
59
(-3
.353
)0.0
89
(0.7
52
)0.2
57
(2.4
62
)0.9
36
11
0K
yush
u-0
.205
(-2
.561
)-0
.294
(-2
.361
)0.5
19
(3.3
03
)0.9
01
(9.9
44
)-0
.028
(-1
.401
)0.9
17
12
2
Not
e:t-
stat
isti
csin
pare
nthe
ses.
Kan
to’s
lagg
edeff
ect
are
esti
mat
edto
be0
and
thus
omit
ted
from
the
repo
rt.
Lag
orde
rsar
ese
lect
edse
para
tely
bySc
hwar
zB
ayes
ian
crit
erio
nfr
oma
max
imum
lag
orde
rof
4.
6
Fig
ure
1:Ja
pane
seIn
dust
rial
Pro
duct
ion
Gro
wth
(Qua
rter
ly,sa
)
−0
.25
−0
.2
−0
.15
−0
.1
−0
.050
0.0
5
0.1
2008Q3
2008Q4
2009Q1
2009Q2
2009Q3
2009Q4
2010Q1
2010Q2
2010Q3
2010Q4
2011Q1
2011Q2
2011Q3
2011Q4
2012Q1
2012Q2
2012Q3
Ind
ustr
ial P
rod
uctio
n G
row
th (
Qu
art
erly)
7
Figure 2: Japanese Regional Map
8
Fig
ure
3:T
ime
Seri
esP
lot
ofTot
alIn
dust
rial
Pro
duct
ion
Dat
a
Tim
e
2000
2005
2010
7090110
Toho
kuH
okka
ido
Kan
toC
hubu
Tim
e
2000
2005
2010
7090110
Kin
kiC
hugo
kuS
hiko
kuKy
ushu
9
Fig
ure
4:T
ime
Seri
esP
lot
ofE
lect
ric
Mac
hine
ryIn
dust
ryD
ata
Tim
e
2000
2005
2010
40100
Toho
kuH
okka
ido
Kant
oC
hubu
Tim
e
2000
2005
2010
60120180
Kink
iC
hugo
kuSh
ikok
uKy
ushu
10
Fig
ure
5:T
ime
Seri
esP
lot
ofTra
nspo
rtat
ion
Indu
stry
Dat
a
Tim
e
2000
2005
2010
40100160
Toho
kuH
okka
ido
Kant
oC
hubu
Tim
e
2000
2005
2010
40100160
Kink
iC
hugo
kuSh
ikok
uKy
ushu
11
Figure 6: Shock on IP based on Similarity matrix
0 10 20 30 400
0.5
1
Kanto90% Bootstrap Bound
0 10 20 30 400
1
2
Tohoku90% Bootstrap Bound
0 10 20 30 400
0.5
1
Hokkaido90% Bootstrap Bound
0 10 20 30 400
0.5
1
1.5
Chubu90% Bootstrap Bound
0 10 20 30 400
0.5
1
Kinki90% Bootstrap Bound
0 10 20 30 400
0.5
1
Chugoku90% Bootstrap Bound
0 10 20 30 40−0.5
0
0.5
1
Shikoku90% Bootstrap Bound
0 10 20 30 400
0.5
1
1.5
Kyushu90% Bootstrap Bound
12
Figure 7: Shock on IP based on Mutual Buying Matrix
0 10 20 30 400
0.5
1
Kanto90% Bootstrap Bound
0 10 20 30 400
1
2
Tohoku90% Bootstrap Bound
0 10 20 30 400
0.5
1
Hokkaido90% Bootstrap Bound
0 10 20 30 400
0.5
1
1.5
Chubu90% Bootstrap Bound
0 10 20 30 400
0.5
1
Kinki90% Bootstrap Bound
0 10 20 30 400
0.5
1
1.5
Chugoku90% Bootstrap Bound
0 10 20 30 40−0.5
0
0.5
1
Shikoku90% Bootstrap Bound
0 10 20 30 400
0.5
1
1.5
Kyushu90% Bootstrap Bound
13
Figure 8: Shock on IP based on Contiguity Matrix
0 10 20 30 400
0.5
1
Kanto90% Bootstrap Bound
0 10 20 30 400
1
2
Tohoku90% Bootstrap Bound
0 10 20 30 40−0.5
0
0.5
1
Hokkaido90% Bootstrap Bound
0 10 20 30 400
0.5
1
1.5
Chubu90% Bootstrap Bound
0 10 20 30 400
0.5
1
Kinki90% Bootstrap Bound
0 10 20 30 400
0.5
1
Chugoku90% Bootstrap Bound
0 10 20 30 40−1
−0.5
0
0.5
Shikoku90% Bootstrap Bound
0 10 20 30 400
0.5
1
1.5
Kyushu90% Bootstrap Bound
14
Figure 9: GIRF of IP by 1 unit shock on Tohoku(a). Similarity Matrix
Tohoku Kanto Hokkaido Kinki Shikoku Chubu Chugoku Kyushu
0
0.5
1
1.5
GIRF
h=0h=3h=5h=10h=20h=50
(b). Mutual Buying Matrix
Tohoku Kanto Chubu Kinki Hokkaido Chugoku Kyushu Shikoku
0
0.5
1
1.5
GIRF
h=0h=3h=5h=10h=20h=50
(c). Contiguity Matrix
Tohoku Kanto Hokkaido Chubu Kinki Chugoku Shikoku Kyushu
−0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
GIRF
h=0h=3h=5h=10h=20h=50
15
Figure 10: Shock on Electric Machinery based on Electric Machinery MutualBuying Matrix
0 10 20 30 400
0.5
1
Kanto90% Bootstrap Bound
0 10 20 30 400
2
4
6
Tohoku90% Bootstrap Bound
0 10 20 30 400
1
2
Hokkaido90% Bootstrap Bound
0 10 20 30 400
1
2
Chubu90% Bootstrap Bound
0 10 20 30 400
0.5
1
Kinki90% Bootstrap Bound
0 10 20 30 400
1
2
Chugoku90% Bootstrap Bound
0 10 20 30 400
1
2
3
Shikoku90% Bootstrap Bound
0 10 20 30 40−1
0
1
2
Kyushu90% Bootstrap Bound
16
Figure 11: Shock on Electric Machinery based on Contiguity Matrix
0 10 20 30 40−0.5
0
0.5
Kanto90% Bootstrap Bound
0 10 20 30 400
2
4
6
Tohoku90% Bootstrap Bound
0 10 20 30 400
2
4
Hokkaido90% Bootstrap Bound
0 10 20 30 400
1
2
3
Chubu90% Bootstrap Bound
0 10 20 30 400
0.5
1
Kinki90% Bootstrap Bound
0 10 20 30 40−1
0
1
2
Chugoku90% Bootstrap Bound
0 10 20 30 400
1
2
3
Shikoku90% Bootstrap Bound
0 10 20 30 40−1
0
1
2
Kyushu90% Bootstrap Bound
17
Figure 12: GIRF of Electric Machinery by 1 unit shock on Tohoku(a). Electric Machinery Mutual Buying Matrix
Tohoku Kanto Kinki Chubu Shikoku Hokkaido Kyushu Chugoku 0
0.5
1
1.5
2
2.5
3
3.5
4
GIRF
h=0h=3h=5h=10h=20h=50
(b). Contiguity Matrix
Tohoku Kanto Hokkaido Chubu Kinki Chugoku Shikoku Kyushu 0
0.5
1
1.5
2
2.5
3
3.5
4
GIRF
h=0h=3h=5h=10h=20h=50
18
Figure 13: Shock on Transportation based on Transportation Mutual BuyingMatrix
0 10 20 30 40−0.5
0
0.5
Kanto90% Bootstrap Bound
0 10 20 30 400
2
4
6
Tohoku90% Bootstrap Bound
0 10 20 30 400
1
2
3
Hokkaido90% Bootstrap Bound
0 10 20 30 40−2
0
2
4
Chubu90% Bootstrap Bound
0 10 20 30 40−1
0
1
2
Kinki90% Bootstrap Bound
0 10 20 30 40−0.5
0
0.5
1
Chugoku90% Bootstrap Bound
0 10 20 30 40−1
−0.5
0
0.5
Shikoku90% Bootstrap Bound
0 10 20 30 400
1
2
Kyushu90% Bootstrap Bound
19
Figure 14: Shock on Transportation based on Contiguity Matrix
0 10 20 30 40−0.5
0
0.5
Kanto90% Bootstrap Bound
0 10 20 30 400
2
4
6
Tohoku90% Bootstrap Bound
0 10 20 30 400
1
2
3
Hokkaido90% Bootstrap Bound
0 10 20 30 40−2
0
2
4
Chubu90% Bootstrap Bound
0 10 20 30 40−1
0
1
2
Kinki90% Bootstrap Bound
0 10 20 30 40−0.5
0
0.5
1
Chugoku90% Bootstrap Bound
0 10 20 30 40−1
−0.5
0
0.5
Shikoku90% Bootstrap Bound
0 10 20 30 40−1
0
1
2
Kyushu90% Bootstrap Bound
20
Figure 15: GIRF of Transportation by 1 unit shock on Tohoku(a). Transportation Mutual Buying Matrix
Tohoku Chubu Kanto Chugoku Shikoku Kyushu Kinki Hokkaido
0
0.5
1
1.5
2
2.5
3
3.5
GIRF
h=0h=3h=5h=10h=20h=50
(b). Contiguity Matrix
Tohoku Kanto Hokkaido Chubu Kinki Chugoku Shikoku Kyushu
0
0.5
1
1.5
2
2.5
3
3.5
GIRF
h=0h=3h=5h=10h=20h=50
21