Measuring changes in international production from a

disruption: Case study of the Japanese earthquake and

tsunami

Cameron A. MacKenziea,1, Joost R. Santosb,2, Kash Barkera,3

aSchool of Industrial and Systems Engineering, University of OklahomabDepartment of Engineering Management and Systems Engineering, The George

Washington University

Abstract

The earthquake and tsunami that struck Japan on March 11, 2011 causeda tremendous loss of life and property. The disaster also disrupted globalsupply chains, which was blamed for anemic growth in the global economy. Amultiregional input-output model can quantify the international impacts onproduction due to changes in demand from companies reducing their ordersbecause of a disruption. By using the input-output model to conceptualize asupply chain, we present a unique method for calculating indirect productionlosses caused by disabled production facilities. Methods for calculating thepossible transfer of demand to industries in other countries are also discussed.We apply the multiregional input-output model to the Japanese earthquakeand tsunami. Comparing results generated by Japanese consumer sales withthose generated by Japanese production data reveals that Japanese demandwas satisfied by other countries and that inventory in the production pipelinelikely allowed consumer sales to remain strong.

Keywords: disruption, input-output, Japan earthquake, inventory

1202 West Boyd, Room 124, Norman, OK 73019, Phone 405.325.3721 Fax:405.325.7555, Email address: cmackenzie@ou.edu

2joost@gwu.edu3kashbarker@ou.edu

Preprint submitted to International Journal of Production Economics March 6, 2012

1. Introduction

On March 11, 2011, a 9.0 magnitude earthquake and tsunami struckJapan. More than 15,000 people were confirmed dead with almost 5,000missing people, and 120,000 homes and buildings were destroyed. In additionto the humanitarian and reconstruction costs, the disruption was blamed inpart for anemic growth in the U.S. and global economy. On July 20, U.S.Treasury Secretary Timothy Geithner (2011) stated:

The economy absolutely slowed in the first half of the year. . . . Itslowed because . . . gas prices went up a lot because we had a hugesupply disruption in the Mideast. You saw some really terribleweather across the country which slowed construction spending.. . . You saw Japan suffer catastrophic damage.

As Secretary Geithner suggested, several other global events occurredduring the first half of 2011, and separating the economic impacts due tothe earthquake and tsunami from impacts caused by other events poses achallenge for modelers. Accurately quantifying the international impacts onproduction caused by major natural disasters can help national and inter-national policymakers better understand how a disruption that occurs inone country may exacerbate an economic slowdown in other countries. Thisunderstanding can encourage better risk management at an international,national, and firm level.

Although the events in Japan adversely impacted companies by increasingtheir costs and delaying some production (Nanto et al., 2011), the companies’ability to manage disruptions within their supply chains may have limitedthe macroeconomic impacts of supply shortages in other countries (Lohr,2011; Salter, 2011). The slowdown in manufacturing in European and NorthAmerican countries that occurred in spring 2011 was partly due to supplydisruptions caused by the earthquake and tsunami, but this slowdown didnot seem to impact consumers’ ability to purchase commodities.

Unlike supply shortages, demand fluctuations caused by drops in con-sumer and intermediate demand may represent more permanent productionchanges. The earthquake and tsunami rendered many Japanese produc-tion facilities inoperable for several weeks and months. Production in manyJapanese industries dropped significantly, final sales to Japanese consumersfell for some industries, and Japanese imports and exports fluctuated in themonths after the earthquake (Bank of Japan, 2011; Wassina, 2011). Because

2

production fell in Japan, companies in Japan and those that deliver suppliesto Japanese industries may have experienced a drop in the demand for theirgoods and services. Changes in Japanese exports and imports reflect changesin producer and consumer behavior, as Japanese consumers and companiessubstituted good and services produced by foreign companies because of in-operability in domestic production.

Multiregional input-output (I-O) models (Isard et al., 1998; Crowtherand Haimes, 2010) can help policymakers understand the impacts due todemand fluctuations and changes in consumer and producer behavior thatare caused by a disruption. I-O models produce results that demonstratethe interdependent effects on industries that are not directly impacted bythe disruption. Industries in foreign countries may also increase their pro-duction if the country in which the disruption occurs increases its imports ordecreases its exports because of domestic production difficulties. I-O analysisprovides a richer understanding of both the positive and negative impacts ofa disruption on an industry-by-industry and a country-by-country basis.

Although we use ideas developed in several I-O models, which are de-scribed in Section 2, this paper advances the field in several ways. Section3 presents several different manipulations of the multiregional I-O model toreflect different types of impacts, including the effects of disabled produc-tion facilities, possible mitigation impacts of inventory, changes in demandresulting from a disruption, and consumers substituting goods and services.Section 4 applies this model to the Japanese earthquake and tsunami anduses production and consumer data collected by the Japanese government asinputs into the multiregional model. Finally, the multiregional model is usedto analyze the structural features of the Japanese automotive industry andexplore how this industry’s business structure influenced the internationaleffects of the disaster.

2. Methodological background

A variety of methods and models have been proposed for measuring thedirect and indirect effects on production caused by disruptions (Rose, 2004;Okuyama, 2008). In addition to econometric models (Ellson et al., 1984), I-Omodels have formed the core of modeling the economic impacts of disasters.

3

2.1. Literature reviewI-O models describe the amount of production needed to satisfy a given

level of demand where each industry’s production is used in the productionof other goods and services or is consumed as final demand (Leontief, 1936,1951). Because I-O models rely on an exogenously determined level of de-mand to calculate each industry’s production in an economy, estimating theinterdependent effects due to disabled production facilities presents a chal-lenge to I-O modelers. If changes in industry production can be convertedto changes in final demand, the final demand can be incorporated within anI-O model to estimate total economic losses due to a disruption (Boisvert,1992; Rose et al., 1997). Supply-side I-O models (Ghosh, 1958), in whichproduction output is expressed as a function of primary inputs like labor,can measure the impacts of constrained supply due to a disruption (Davisand Salkin, 1984; Park, 2008). However, supply-side I-O models have beencriticized for the underlying assumption that supply generates demand (seeOosterhaven, 1998, 1989), and these models may more accurately measureprice rather than production deviations (Dietzenbacher, 1997).

Derived from the Leontief I-O model, the Inoperability Input-OutputModel (IIM) (Santos and Haimes, 2004) measures production and demandshortfalls in percentage or fraction terms. The dynamic version of the IIM(Lian and Haimes, 2006) has calculated the interdependent effects of con-strained production due to employees falling ill from a pandemic (Orsi andSantos, 2010a,b) and due to supply shortages (MacKenzie et al., 2012). How-ever, the dynamic model’s results depend on a resilience parameter that isdifficult to estimate.

In order to more accurately incorporate industry actions before and dur-ing a disruption, I-O models have been adapted to incorporate uncertainty(Santos, 2008; Barker and Haimes, 2009), industry mitigation activities suchas inventory (Barker and Santos, 2010a,b), the possibility of alternate routesduring a transportation disruption (Gordon et al., 2005; MacKenzie et al.,2012), the substitution of different inputs (MacKenzie and Barker, 2011),and timing effects that differ among industries (Okuyama et al., 2004). Theadaptive regional I-O model (Hallegatte, 2008, 2011) measures the impact ofsupply constraints and new sources of inputs on production in the wake ofHurricane Katrina.

Computable general equilibrium (CGE) models (Shoven and Whalley,1992) build on the basic I-O framework but create flexibility in the model byallowing consumers and producers to optimize simultaneously and substitute

4

other inputs for constrained supply (Rose and Ghua, 2004). Because theyfocus on a new long-term equilibrium after a disruption, CGE models mayunderestimate the impacts on production (Rose and Liao, 2005; Okuyama,2008). Andreoni et al. (2011) uses a hybrid I-O, CGE model (Kratenaand Streicher, 2009) to reconcile the disequilibrium between demand andsupply caused by the Japanese earthquake and tsunami, but it may be overlyoptimistic about Japanese demand transferring to other countries.

2.2. Multiregional I-O model

Eq. (1) describes the output xs of goods and services in dollars as a func-tion of intermediate production Asxs that is used by other industries andfinal consumer demand cs. For an economy with n industries, xs and cs arevectors of length n that represent economic production and final demand,respectively, for each industry in country s. The n × n technical coefficientmatrix As describes the economic interdependency among industries: indus-try j requires asij dollars of production input from industry i for every dollarof production output by industry j.

xs = Asxs + cs ⇒ xs = (I−As)−1 cs (1)

A multiregional I-O model (Isard et al., 1998) connects all countries basedon international trade. Let T rs

i be the proportion of goods and servicesin industry i consumed by country s that are produced in country r, asdetermined by Eq. (2), where mrs

i is the value of industry i’s goods andservices that are imported by country s from country r, and ms

i and esi arecountry s’s total imports from and exports to all other countries included inthe model for industry i.

T rsi =

mrs

i

xsi +msi − esi

if s 6= r

xsi − esixsi +ms

i − esiif s = r

(2)

The formula for T ssi represents goods and services that are produced and

consumed in the same country. Including both exports and imports in thecalculation ensures that T rs

i captures all the production in industry i andthat

∑∀r T

rsi = 1.

The interregional matrix T is presented in Eq. (3), where each Trs is ofsize n and the ith element on the diagonal is T rs

i . The variable p represents

5

the total number of countries in the model.

T =

T11 T12 · · · T1p

T21 T22 · · · T2p

......

. . ....

Tp1 Tp2 · · · Tpp

(3)

The multiregional I-O model as shown in Eq. (4) incorporates the inter-regional matrix into the I-O model from Eq. (1). (For notational simplicity,we use the notation yq:p =

[(yq)ᵀ , (yq+1)

ᵀ, . . . , (yp)ᵀ

]ᵀto represent a vector

of length n(p − q + 1) for countries q, q + 1, . . . , p, where ys is a vector ofinterest such as production, final demand, or changes in production). Thenp square matrix A = diag (A1,A2, . . . ,Ap).

x1:p = TAx1:p + Tc1:p ⇒ x1:p = (I−TA)−1Tc1:p (4)

The right-hand side of Eq. (4) demonstrates how a given level of demand inone or more countries determines production levels in p countries.

By necessity, multiregional models include fewer countries than the totalnumber of countries in the world. Trade between countries in the model andall the countries not included in the model, which is called the rest of theworld (ROW), can help us estimate the impacts on countries not includedin the model. Let ms

ROW be a vector of length n that represents the valueof imports that country s imports from the ROW. Because the final demandvector cs subtracts imports from final household consumption in the originalI-O model, the ROW imports are added back into the original final demandvector to preserve the equilibrium between supply and demand (Isard et al.,1998). The new multiregional I-O model that includes the ROW is given inEq. (5).

x1:p + m1:pROW = TA

(x1:p + m1:p

ROW

)+ T

(c1:p + m1:p

ROW

)(5)

Including the ROW ensures that the model represents a complete economicsystem.

3. Changes in production due to a disruption

For the purposes of this paper, a major disruption can directly impactproduction in two different ways. First, production facilities may be de-

6

stroyed or severely damaged such that only a portion of the normal produc-tion can occur in that facility. Second, the disruption may cause an increasein the demand for certain industries, such as machinery manufacturing andconstruction that are necessary to rebuild the nation’s infrastructure.

We define indirect impacts as the changes in production among indus-tries who supply goods and services to companies directly impacted by thedisruption. Because industries represent the aggregation of companies thatproduce similar goods and services, a single industry may include compa-nies that are directly impacted, companies that are indirectly impacted, andcompanies that are not impacted at all.

We assume that data exist that describe production in some of a country’sindustries for a period of time following a disruption. The available data mayonly record the direct impacts for those industries, or it may record both thedirect and indirect impacts for those industries. Each of these cases and theirimplications for production losses are presented in separate subsections.

3.1. Data representing direct impacts

Building on the frameworks offered by Cronin (1984) and Oosterhaven(1998), we develop a method to calculate the indirect impacts due to dis-abled production facilities that explicitly uses the technical coefficient matrix(the A matrix) as a means of estimating how a company whose productionfacilities are disabled will be forced to reduce its demand to its suppliers.Conceptualizing the I-O model as a supply chain composed of producers andsuppliers echoes other studies (see Lin and Polenske, 1988; Albino et al.,2002) in which a company’s supply chain is modeled as an I-O process.

Without loss of generality, we assume that industries 1 . . . l in country 1are directly impacted by the disruption where l ≤ n and n is the number ofindustries in each county. The direct impacts in country 1 due to a disrup-tion are given by the vector δx1 (0) where δx1i (0) is non-zero for 1 ≤ i ≤ land is zero for l + 1 ≤ i ≤ n. Because of the direct impacts, these l indus-tries adjust their demand to their immediate suppliers to reflect their newproduction levels. The resulting change in production for the first echelonof suppliers is calculated by premultiplying δx1 (0) by TA. This first ech-elon of suppliers consequently adjusts its demand to its suppliers, and thechange in production for the second echelon is calculated by premultiplyingthe production change in the first echelon by TA. This pattern continues adinfinitum.

7

A model based on this logic might overestimate the indirect impacts incountry 1, however, because some of the directly impacted companies mayalso be suppliers to other directly impacted companies. Eq. (6) calculatesδxs, the total production changes (direct plus indirect impacts) in countrys where the ROW impacts are momentarily ignored. The parameter ρ is avector of length n representing the proportion of companies in each industrythat are not directly and adversely impacted by the disruption.

(δx1

δx2:p

)=

∞∑k=0

(PTA)k(δx1 (0)

0

)= (I−PTA)−1

(δx1 (0)

0

)(6)

where

P =

(diag (ρ) 0

0 I

)If we assume that suppliers are uniformly distributed throughout the country,ρi can be estimated for industry i via Eq. (7).

ρi =

1 +δx1i (0)

x1iif δx1i (0) < 0

1 otherwise(7)

If δx1i (0) > 0, which represents a positive direct impact on industry i, weassume no suppliers in this industry are directly impacted in an adversemanner.

We assume the indirect impacts of the disruption are spread proportion-ally between the country’s domestic production and the ROW imports. Eq.(8) calculates the indirect impact in industry i in country s as representedby δxsi (1), where δxsi − δxsi (0) represents the indirect impacts in industry iin both country s and the ROW, and xsi

/(xsi +ms

ROW,i

)is the proportion of

that industry’s output that is produced in country s.

δxsi (1) =[δxsi − δxsi (0)]xsixsi +ms

ROW,i

(8)

After separating out the ROW imports, industry i’s total production changein country s can be calculated as the summation of direct plus indirect im-pacts: δxsi (0) + δxsi (1).

8

3.2. Data representing both direct and indirect impacts

We have been assuming that the direct impacts are known, but it maybe easier to observe the total economic impact on industries. Aggregatingoutput levels for an industry from company surveys will likely include bothcompanies whose production changes due to indirect impacts as well as com-panies whose production is reduced due to disabled facilities. Under thisformulation, the total impacts δx1i (0 : 1) = δx1i (0) + δx1i (1) are known forthe first l industries in country 1 but are unknown for industries l + 1 . . . n.The direct impacts δx1i (0) are unknown for the first l industries. We assumethere are no direct impacts for industries l + 1 . . . n.

We seek to express δx1i , the change in industry i’s production, as a func-tion of the direct and indirect impacts on domestic production. By substi-tuting the formula for δx1i (1) as given in Eq. (8) into the calculation ofδx1i (0 : 1) as shown in the previous paragraph, we derive Eq. (9) to calculateδx1i .

δx1i =

(1−

x1i +m1ROW,i

x1i

)δx1i (0) +

x1i +m1ROW,i

x1iδx1i (0 : 1) (9)

We assign αi =(x1i +m1

ROW,i

)/x1i as the inverse of the proportion of

total production changes that impact domestic production. The data, whichwe assume measure direct and indirect impacts for industries 1 ≤ i ≤ l, arerepresented by δx1i (0 : 1). Eq. (10) is identical to Eq. (6) except that Eq.(9) is substituted for δx1i . (We also let y1

1:l be a vector representing industries1 . . . l in country 1 for a quantity of interest y1i ).

diag(α1:l ) δx1

1:l (0 : 1)+ [I− diag ( α1:l ) ] δx1

1:l (0)δx1

l+1:n

δx2:p

= (I−PTA)−1

δx11:l (0)00

(10)

The total impacts in the first l industries δx11:l (0 : 1) are known and ob-

served, but the direct impacts δx11:l (0) in those industries as well as the total

impacts in other industries δx1l+1:n and other countries δx2:p are unknown.

Eq. (10) can be solved for the direct impacts on the first l industries andthe total impacts on all other industries and countries. As shown in Eq. (7),P assumes the proportion of direct impacts to total production is known a

9

priori. We estimate P using the results derived from assuming the data onlyrepresent direct impacts.

3.3. Impact on demand

Direct production losses in country 1 may be replaced by production fromother countries or by inventory. If a company whose production is disableduses finished goods inventory to satisfy demand, we still expect to observeproduction losses. As long as its facility is inoperable, that company is notproducing, and the model assumes that a non-producing company reduces or-ders to suppliers. Using finished goods inventory should allow the companyto recover those losses in the future once its facility is restored. Finishedgoods inventory might not prevent production losses in the immediate after-math of a disruption, but because inventory can satisfy current demand, weexpect to observe later an increased amount of production in industries whorely on inventory to weather the disruption.

If inventory is not available, other competing companies may be ableto increase their production to replace the lost production. We develop amethodology to estimate the increased production that would be needed inother countries to satisfy demand in country 1. We assume that the fractionof total imports that country 1 imports from each country remains constantwhile country 1’s total imports increase to replace its domestic production.As we will discuss in Section 4, Japan increased its imports following theJapanese earthquake and tsunami. The multiregional I-O model translatesthese additional Japanese imports into increased production quantities inother countries.

Customers in country 1 may not purchase goods and services from othercountries. Industries in other countries may not be able increase their produc-tion, customers may not want to buy from foreign industries, or disruptionsin shipping may limit the ability of country 1 to import products. If demandin country 1 is not satisfied by increased production in other countries orwith inventory, purchases could be delayed and customers may increase theirdemand in later months.

3.4. Import substitution

A country that is not directly impacted by the disaster may replace lostimports from the disrupted country with its own domestic production. Weassume that the fraction of country 1’s constrained production that manifestsitself as lost exports to country s equals the fraction of country 1’s normal

10

production that is exported to country s. Eq. (11) calculates the reducedlevel of imports m1s that industry i in country s receives from country 1,where m1s

i is the pre-disaster value of imports that industry i in country simports from country 1 and x1i is the constrained production of industry i incountry 1.

m1si =

m1s

i x1i

x1iif x1i < x1i

m1si otherwise

(11)

A new interregional matrix is constructed to reflect that countries are notimporting as much from the country that has suffered a disruption. The newnp× np interregional matrix T is calculated via Eq. (2) with m1s

i replacingm1s

i for each country s. The total imports by country s, msi , is also reduced

by m1si − m1s

i .When the disruption occurs, industries in countries not directly disrupted

may experience conflicting impacts. They may suffer from lost demand dueto industries in the disrupted country reducing their demand for production,but they may also see an increase in demand if the country suffering fromthe disruption increases its imports. These industries’ domestic customersmay also substitute domestically produced goods and services in the placeof imports from the disrupted country. This last impact is calculated byincorporating the new interregional matrix into Eq. (5) and solving for a newproduction vector x1:p while keeping the final demand vector c1:p constant.

This section has explored several different methods that extend the multi-regional I-O model to estimate the international impacts of a major disruptiveevent. Any one of them may be appropriate to analyze a situation, but itis difficult to know a priori which method or methods is most appropriate.We apply these methods to the 2011 Japanese earthquake and tsunami anddiscuss which method seems to describe most accurately the impacts of thisdisaster.

4. Empirical application: 2011 Japanese earthquake and tsunami

We next deploy these different methods to estimate the internationalimpacts on production following the 2011 Japanese earthquake and tsunami.

11

4.1. Data sources

Before showing and analyzing the different results, we review the datasources used for this application. The Organization of Economic Coopera-tion and Development (OECD, 2011) collects and publishes I-O data in U.S.dollars for all of the OECD countries and for 11 non-OECD countries in Asiaand South America. The most recent data come from the mid-2000s, andeach national economy is divided into 37 industries. We use this data to cre-ate the technical coefficient matrices and production and final consumptionvectors. Using data from the mid-2000s to study the economic consequencesof an disruption that occurred in 2011 assumes that the structural coefficientshave not drastically changed in the years since this data was collected. Thisis customary in I-O modeling in the absence of more current data.

We select 18 countries in addition to Japan to include in the model:Australia, Belgium, Brazil, Canada, Chile, China, France, Germany, India,Indonesia, Italy, South Korea, Mexico, the Netherlands, Taiwan, Thailand,the United Kingdom, and the United States. These countries represent 66%of all of Japan’s imports. The rest of Japan’s imports that are not included inthe model mostly derive from the Middle East and other Asian countries forwhich the OECD does not publish I-O tables. The I-O analysis includes thedifference between a country’s total imports and imports from the other 18countries to estimate the impacts on the ROW, which includes these MiddleEastern and Asian countries.

The OECD also provides bilateral trade matrices for 21 of the 37 indus-tries. The imports from 2009 (the most recent year available) are used togenerate the T matrix for OECD countries. In order to estimate trade datafor the other 16 industries, we assume that the fraction that country s im-ports from country r for those industries is equal to the overall proportionthat country s imports from that country r. The total value of trade betweenthe five non-OECD countries as published by the United Nations (2009) isused to estimate T for these five countries.

Data to estimate the impacts of the Japanese earthquake and tsunamicome from the Japanese Ministry of Economy, Trade and Industry (METI),which publishes monthly production data for 14 manufacturing industriesand the mining, food and tobacco, and construction industries (Japan, 2011b).For each industry, METI reports the production index (using 2005 as the baseyear) and the percent change in the index from the same month of the pre-vious year. It also publishes indices and changes from the previous year for

12

Table 1: Percent change in industry production in 2011 from same months in 2010

IndustryMarch April May

Pro Ship Inv Pro Ship Inv Pro Ship Inv

Mining and quarrying 1.4 -3.2 -4.5 -5.9 0.5 11.6 -1.8 1.1 -9.2Food and tobacco -7.5 -2.2 -47.1 -2.4 -4.2 -36.2 1.9 1.1 -29.3Textiles 2.3 -2.1 -5.2 -0.1 -2.1 -2.9 2.2 1.8 -3.7Wood Products 1.8 3.8 -11.0 3.2 6.1 -11.0 5.6 8.8 -11.1Paper products -6.2 -6.1 -5.9 -7.2 -5.4 -7.2 -6.5 -3.9 -8.1Coke and refinedpetroleum products -10.0 -9.0 -4.6 -13.0 -13.7 5.0 -9.7 -7.0 -7.0Chemicals -7.5 -6.2 0.7 -9.0 -8.5 1.2 0.3 -5.3 4.6Rubber andplastic products -9.7 -11.7 1.3 -5.9 -9.0 2.4 0.6 -2.2 3.3Other non-metallicmineral products -2.9 -7.7 2.3 -2.2 -5.0 6.0 -2.4 -3.7 6.6Basic metals -9.3 -7.2 7.5 -9.3 -8.4 6.7 -9.2 -12.2 10.6Fabricated metal -6.9 -6.0 -1.9 -7.0 -6.2 -0.3 -0.9 -1.5 1.9Machinery and equipment 6.9 4.7 0.7 8.5 3.8 3.9 17.3 16.1 10.2Electrical machinery -4.1 -2.7 30.8 -8.8 -7.3 25.6 -6.0 -5.2 32.4Medical andprecision instruments -10.8 -8.5 -5.4 0.1 -11.9 3.3 3.0 3.5 5.7Motor vehicles -47.7 -38.9 -51.8 -49.0 -52.6 -43.3 -26.5 -31.0 -23.2Other manufacturing -7.3 -9.0 -0.9 -3.2 -3.3 -0.1 -0.2 1.9 -1.1Construction -1.3 -0.8 0.2 2.7 1.4 4.5 3.7 1.8 8.2

Production abbreviated as “Pro,” Shipments as “Ship,” and Inventory as “Inv.”

industry shipments and inventory (Table 1). The percent changes in pro-duction for March, April, and May 2011 become the production changes inthese 17 industries that are directly impacted by the disruption. We assumethat 2010 represents as-planned production so that the measured productiondeviations in 2011 are changes from typical production.

Comparing industry production with its shipments provides insight intohow industries relied on inventory. If the negative percent change in an indus-try’s production is less than the percent change in its shipments, we assumethe industry uses inventory to fill the gap between production and shipments.As Table 1 shows, in many of the months where industry shipments exceededproduction, inventory also decreased during those months.

Japan’s monthly consumer sales provide another important data sourcefor this analysis. Monthly consumer sales published by METI (Japan, 2011d)

13

Table 2: Percent change in consumer sales in 2011 from same months in 2010

Industry March April May

Agriculture, forestry, and fishing -11.1 -14.1 -7.3Food and tobacco -0.2 -0.6 3.0Textiles -7.9 -6.0 2.0Wood products -1.5 1.5 2.4Coke and refined petroleum products 5.3 -1.0 1.3Chemicals -6.3 -2.5 -3.7Other non-metallic mineral products 11.3 5.6 5.0Fabricated metal 11.3 5.6 5.0Machinery and equipment -2.9 -1.0 8.3Electrical machinery 10.5 -2.7 3.9Motor vehicles -9.3 -13.9 -10.8Other transport equipment -3.5 -2.4 3.3Other manufacturing 0.1 3.9 7.0Construction -1.5 1.5 2.4Wholesale trade 3.4 2.3 -1.5

allow us to estimate changes in consumer sales for the agriculture, food andtobacco, textile, wholesale trade, construction, and 10 manufacturing indus-tries (Table 2).

The consumer sales data represent changes in the final consumption vec-tor c1 for Japan and can be incorporated into Eq. (4) to calculate the impacton production. Using consumer sales as a proxy for changes in final consump-tion sheds more light onto the impacts of the earthquake and tsunami on theJapanese consumer and how the direct impacts of production are manifestedin final sales.

4.2. Data representing direct impacts

We first examine the economic impact if the observed industry productiondata for the 17 industries recorded by METI represent only the direct impactsof the earthquake and tsunami. Fig. 1 shows the direct impacts without in-ventory and indirect impacts with and without inventory for March, April,and May 2011. We separate the results for the most impacted countries:Japan, China, Germany, South Korea, and the United States. Other Asiarepresents the Asian countries included in the model: Australia, India, In-donesia, Taiwan, and Thailand. Other Europe represents the countries of

14

−50

−40

−30

−20

−10

0

Japan China Germany Korea UnitedStates

OtherAsia

OtherEurope

Other WestHemisphere

ROW

Bill

ions

of U

.S. d

olla

rs

Direct impacts with no inventoryIndirect impacts with inventoryIndirect impacts with no inventory

March April May

Figure 1: Monthly changes in production when the data represent direct impacts. OtherAsia represents Australia, India, Indonesia, Taiwan, and Thailand. Other Europe repre-sents Belgium, France, Italy, and the United Kingdom. Other West Hemisphere representsBrazil, Canada, Chile, and Mexico.

Belgium, France, Italy, and the United Kingdom; and Other West Hemi-sphere represents Brazil, Canada, Chile, and Mexico.

Japan’s total production losses in March and April exceeded $51.9 billion,which corresponded to 7.3% of Japan’s monthly output. Production lossesonly totaled $20.7 billion in May. Principal drivers behind the increasedproduction in May include the motor vehicles industry whose productionrose from 51.0% to 73.5% of normal in May and general machinery thatproduced 17.3% more in May 2011 than it had in 2010.

These results also show that the macroeconomic impact on other countrieswas minimal relative to Japan. If inventory is ignored, the production lossoutside of Japan was $17.2 billion over the span of those three months. Chinaexperienced the most severe impact with losses about $1.5 billion in Marchand April and $584 million in May, but even the losses in March and Aprilaccounted for less than 0.3% of China’s monthly production.

The model also demonstrates the importance of inventory. Japanese in-dustries that used inventory to maintain their shipments should produce morein order to replenish their inventory. With inventory, Japan should recover$20.3 billion or 15.8% of its production that was lost from March to May.

15

−30

−20

−10

0

Japan China Germany Korea UnitedStates

OtherAsia

OtherEurope

Other WestHemisphere

ROW

Bill

ions

of U

.S. d

olla

rs

Direct impacts with no inventoryIndirect impacts with inventoryIndirect impacts with no inventory

March April May

Figure 2: Monthly changes in production when the data represent both direct and indirectimpacts

4.3. Data representing both direct and indirect impacts

The METI production data may incorporate both direct and indirect im-pacts on production. As demonstrated in Fig. 2, assuming that the datainclude both direct and indirect effects significantly changes the analysis ontotal production losses. Japanese production losses in March and April ex-ceeded $32.4 billion (or about 4.6% of monthly production), and productionlosses in May were $11.6 billion. Production losses in the other countriestotaled $10.8 billion over the three months. The estimate of the earthquakeand tsunami’s impact on production decreases by approximately 40% usinga model that assumes the observed data include both types of impacts.

Comparing the analyses from Figs. 1 and 2 raises the question of whichassumption is more accurate. As information on industry production is usu-ally gathered via company surveys, answering that question depends in parton the length of time it takes for these indirect impacts to ripple throughsupply chains. Some industries, like agriculture and mining, produce monthsor even a year in advance to meet anticipated demand. We would not expectto observe indirect impacts in these industries for several months or more.Manufacturing industries can generally react to demand changes in weeks,and still other industries, like service industries, can react to demand changesalmost immediately (Okuyama et al., 2004). Because manufacturing indus-tries comprise most of the observed data, indirect impacts should appear in

16

the data within a matter of weeks as opposed to months.The data reveal that many industries had much smaller production losses

in May. If the lag time to observe the indirect impacts were larger than a fewweeks, the month of May would have had greater production losses becausethe indirect impacts resulting from the direct effects that occurred in Marchand April would not have been observable until May or even later. Jour-nalistic accounts (Lohr, 2011; Rowley, 2011) of the earthquake and tsunamialso suggest that the economic reverberations were felt within the first coupleof months. It appears more reasonable to assume that the published datainclude both direct and indirect impacts as opposed to only direct effects.

I-O analysis provides important insight into the impacts on individualindustries. We aggregate the 37 industries into 10 industries and explorethe direct and indirect impacts in each industry in Japan with and with-out inventory (Fig. 3). These results are based on the assumption thatthe METI production data include both direct and indirect impacts. Thetransportation and office equipment industry, which includes the automotivesector, suffered the greatest production losses, both in direct and in indirectimpacts. The minerals and metals industry, which includes a lot of basicmanufacturing, had $10.3 billion in indirect production losses even thoughthe industry’s direct losses only totaled $1.1 billion. Japanese service indus-tries (wholesale and retail trade, transportation and telecommunication, andbusiness services) had no direct impacts but suffered $11.6 billion in produc-tion losses during the three months following the earthquake and tsunami.Conversely, the agriculture, mining, food, and textiles industries only lost$2.6 billion, of which $1.6 billion were direct production losses. As Barkerand Santos (2010b) discuss, the breakdown of industry production losses canhelp policymakers determine the key sectors that are impacted by a disrup-tion. The disruption in the Japanese automotive sector led to tens of billionsof dollars of production losses in the service and manufacturing sectors.

4.4. Impact on Japanese demand

The other data set, the commercial sales indices, demonstrates the impactof the earthquake and tsunami on the Japanese consumer. Fig. 4 showsthe impact of demand fluctuations on production over March, April, andMay. The total production loss due to demand changes was an order ofmagnitude less than the impact due to disabled production facilities. Thelargest month, April, only had production losses of $3.0 billion, which is lessthan a tenth of production losses as calculated from production losses. Final

17

−60

−50

−40

−30

−20

−10

0

10

Agri−culture

Mining, food, and

textiles

Mineralsand

metals

Transp and

office equip

Util andcon−

struction

Whole−sale and

retail trade

Transp and

telecom services

Businessservices

Educationand health

Otherservices

Bill

ions

of U

.S. d

olla

rs

Direct impactsIndirect impacts

With inventory

Without inventory

Figure 3: Changes in production in Japanese industries

−3

−2

−1

0

1

2

Japan China Germany Korea UnitedStates

OtherAsia

OtherEurope

Other WestHemisphere

ROW

Bill

ions

of U

.S. d

olla

rs

March April May

Figure 4: Monthly changes in production due to demand changes

consumption increased in May beyond 2010 levels, leading to a gain of $2.1billion in production for Japan and minor increases in production for theother countries. These results suggest that much of the demand that wasnot satisfied in March and April returned in May.

18

How should we interpret the results based on production data in lightof these results based on commercial sales? First, Japanese demand seemsto be resilient. Even though final consumption dropped during the first twomonths after the tsunami and the earthquake, it started to revive by thethird month. Second, commercial sales data only include a fraction of theJapanese industries. Including changes in consumer demand for more of theindustries might result in greater production losses for March and April, butit still would not reflect the results based on METI’s production data.

Third, Japanese final consumption never fell as far as Japanese produc-tion. The results derived from the production data suggest that Japaneseindustries were not producing enough to meet the current demand, and yet,much of that demand was actually met. For example, motor vehicle produc-tion in March and April fell 45% from 2010 levels, but final demand for thatindustry only fell 10% to 15%.

One explanation for this large difference is that Japan increased its im-ports to make up for lost production. According to the Japanese Statis-tics Bureau (Japan, 2011a), Japanese imports increased by 10.7% in March,April, and May 2011 compared to those same months in 2010. If Japan hadreplaced all of its lost production with imports, the multiregional I-O modelpredicts that Japan’s imports should have increased by 14.7% during thosethree months. Table 3 depicts the increased production in other countriesthat we would expect to have occurred given that Japanese imports increasedby 10.7%. Our approach assumes that the fraction of Japan’s total importsfrom each country remains constant. Germany and China benefited the mostfrom increased imports as their industrial production should have increasedby $10.5 billion and $6.8 billion, respectively.

Inventory in the pipeline likely accounts for the rest of production short-fall that was not met by increased foreign production and imports (Calunson,2011; Shameen, 2011). The commercial sales data derive mostly from retailsales, and the stores’ inventories were probably not reflected in the produc-tion, inventory, and shipment data released by METI. Because inventoryenabled some Japanese demand to be met, Japanese production should riseabove normal levels once facilities are restored as the inventory in the pipelineand stores’ inventories would need to be replenished.

4.5. Import substitution

From March through May 2011, the total value of Japan’s exports fell by5% compared with 2010. Overall, 5% is not a large difference, but the lost

19

Table 3: Changes in production from different models (millions of U.S. dollars)

CountryDriver of production changes

Disabled Increased exports Importfacilities to Japan substitution

Japan -78,068China -2,230 6,774 3,350Germany -824 10,495 1,244South Korea -595 3,383 1,629United States -826 3,790 10,771Other Asia -1,991 3,771 3,401Other Europe -629 3,790 2,629Other West Hemisphere -411 1,722 2,377ROW -3,300 7,178 1,020

exports were heavily concentrated in a few industries. Exports from Japan’smotor vehicles industry fell by approximately 21% and computer exportsdropped by 17% (Japan, 2011a).

It seems likely that countries’ domestic industries produced more to re-place some of the lost imports from Japan. For example, U.S. automobilemanufacturers General Motors, Ford, and Chrysler increased their share ofproduction in North America from 55% of total production in January toalmost 59% by June 2011, whereas Toyota’s share fell by 2% and Honda’sshare fell by 3% (Ward’s, 2011). The third column in Table 3 shows the pro-duction changes if countries’ domestic industries compensated for the loss inimports from Japan by increasing their production. According to this model,other countries benefited by increasing their domestic production: the UnitedStates increased its production by $10.8 billion, China by $3.4 billion, andSouth Korea by $1.6 billion.

As depicted in columns 2 and 3 in Table 3, increasing both exports toJapan and domestic production to replace lost imports from Japan greatlysurpassed the indirect production losses that countries experienced due todisabled Japanese facilities. The multiregional I-O model separates andquantifies the changes in production resulting from these different drivers.According to the model, which aligns closely with the actual trade data, theoverall impact from the Japanese earthquake and tsunami provided macroe-conomic benefits to other countries.

20

−100

−80

−60

−40

−20

0

Japan China Germany Korea UnitedStates

Bill

ions

of U

.S. d

olla

rs

Direct impacts in home countryIndirect impacts in home countryIndirect impacts in foreign countries

Data representing directimpacts

Data representing both directand indirect impacts

Figure 5: Changes in production due to impacts on the automotive industry

4.6. Japanese automotive industry structure

Japan’s motor vehicles or automotive industry suffered the most from theearthquake and tsunami. The Japanese auto industry, as exemplified by theToyota keiretsu, stresses an integrated and closely coordinated supply chain(Ellram and Cooper, 1992). If a disruption occurs, the keiretsu can bandtogether by directing certain companies to produce more to replace the lostproduction at other facilities (Sheffi, 2005). The pace at which companieslike Toyota and Honda resumed production after the disaster was quickerthan many observers had expected (Rechtin, 2011; Tabuchi, 2011).

The I-O model can compare the interdependent effects of a disruptionin this industry with similar disruptions in other countries. In addition toJapan, we select the automotive or motor vehicles industry in four othercountries: China, Germany, South Korea, and the United States.

We assume that each country’s automotive industry proportionally suffersthe same impact as that of Japan’s automotive industry from March throughMay: automotive production at 38.1% less than normal. Each situation isanalyzed separately, i.e., only one country’s automotive industry is directlyimpacted at one time. Fig. 5 shows the impact of each disruption wherethe first bar for each country assumes the 38.1% are direct impacts and thesecond bar assumes the 38.1% includes both direct and indirect impacts inthe automotive industry.

21

Understanding the reasons for the differences in Fig. 5 is important forforecasting the interdependent effects of large disruptions that might occur indifferent countries. Indirect losses in Japan are proportionally smaller thanthose in the other countries because Japan’s automotive industry is veryself-dependent. From the technical coefficient matrix, Japan’s automotiveindustry requires 46 cents of production input from its own industry forevery dollar of production output. None of the other countries’ automotiveindustries require more than 40 cents of input from their own industries,and the U.S. automotive industry only needs 29 cents. If a disaster disables38.1% of the automotive industry in Japan, many suppliers who are also partof the automotive industry will also be directly impacted. Because so manysuppliers are already directly impacted, the model predicts fewer indirectproduction losses.

The international impacts from a disruption in Japan are smaller than ifthe disruption occurs in another country. The Japanese automotive industryrelies more on industries within Japan than industries in foreign countries.Imports account for less than 4% of Japan’s automotive production, whereasimports account for 25% of Germany’s automotive production and 40% ofU.S. automotive production. Because Japanese automotive industries relyon industries within its own country, the international impacts of the earth-quake and tsunami were limited. If a similar disruption were to occur inthe United States or a European country, the international impacts could bemore significant.

5. Conclusions

This paper has presented several approaches to estimate the differentproduction impacts caused by a major disruption. The 2011 Japanese earth-quake and tsunami provide an appropriate case study for this multiregionalI-O modeling framework. The framework is parameterized using I-O tablesand trade data from the OECD, and production and consumer sales datapublished by METI provide a reliable source of input data into the model.

The model does not attempt to quantify every possible international im-pact resulting from this disaster, and supply shortages may have causedglobal production to slow down, especially in the automotive and electron-ics industries. Other possible impacts of the disruption include interna-tional price fluctuations and changes in Japanese wages or employment (U.S.Department of Commerce, 1997). Shortages increased the prices of some

22

consumer products, such as automobiles (Hirsch, 2011) and cameras (Lam,2011), but employment and wages in Japan remained fairly constant fromMarch through May (Japan, 2011c).

Without I-O tables and the Leontief economic model, understanding thedifferences between consumer sales and production data would be more chal-lenging because the former involves sales to final consumers and the latterincludes intermediate and final production. I-O methods translate both setsof data into a common measurement, total production losses in Japan. If thedata include both direct and indirect impacts for the 17 industries, produc-tion losses in Japan totaled $78.1 billion and Japan’s gross domestic productlost $41.7 billion from March to May. These losses represent 3.6% of Japan’stypical economic output. Disaggregating the production losses by industryreveals that minerals and metals manufacturing, transportation equipment,and several services industries suffered the greatest indirect impacts. Theconsumer sales data suggest a different picture, however, as changes in finalconsumption only led to $3.7 billion in production losses.

This $74.4 billion difference generated by these two methods can be ex-plained by increased Japanese imports from other countries and inventory inJapan’s retail stores. Japanese imports increased by 10.7% during the threemonths following the earthquake and tsunami. This increase satisfied about73% of the shortfall between demand and production.

The multiregional I-O model can distinguish among different impacts onproduction and quantify the changes in production due to losses in intermedi-ate demand, the use of inventory, increased imports, and import substitution.The I-O model suggests that contrary to popular perception, industries inother countries may have actually benefited from the Japanese earthquakeand tsunami. Japanese industries do not import enough from other countriesthat those countries would experience serious production losses due to a dropin intermediate demand. Japan increased its imports to replace some of itsdomestic production, which likely led to more production in countries likeGermany and China. Some industries in other countries, especially the U.S.automotive industry, also benefited as they produced more to meet demandin their home countries.

The importance of this work for planning purposes is that policymakersmay not need to be concerned about the adverse economic impact of large-scale disruptions that occur in foreign countries. Certainly, the humanitarianneeds of a disaster like an earthquake or tsunami require that nations andinternational organizations react quickly to assist in saving lives, caring for

23

displaced people, and ensuring that basic necessities are met. Although na-tional economies are linked together, the indirect impacts from a disruptionwould likely be dispersed among several different countries such that any in-dividual country that escaped direct impacts would probably not experiencelarge production losses. Industries in those other countries may also benefitfrom the disruption due to increasing exports to the disrupted country andreplacing imports from that country. The resilience of the global economy,the likelihood that demand will stay high or recover after a couple of months,and the ability of companies to rely on inventory can help ensure that anyinternational production losses will be temporary and limited in scope.

Acknowledgments

The authors would like to thank Jennifer Spencer, Associate Professorof International Business & International Affairs at The George WashingtonUniversity, for her thoughtful insights. We also thank the thoughtful sug-gestions of the reviewers. This work was supported in part by the NationalScience Foundation, Division of Civil, Mechanical, and Manufacturing Inno-vation under award 0927299 and by the Center for International BusinessEducation and Research at The George Washington University.

References

Albino, V., Izzo, C., Kuhtz, S., 2002. Input-output models for the analysis ofa local/global supply chain. International Journal of Production Economics78 (2), 119-131.

Andreoni, V., Arto, I., Rueda-Cantuche, J., Mongelli, I., 2011. Worldwideeconomic effects of changes in international trade due to natural disaster:the case of Japan earthquake and tsunami. Paper presented at the 19thInternational Input-Output Conference, Alexandria, VA, June 16.

Bank of Japan, 2011. Monthly report of recent economic and financial devel-opments, June 15 <http://www.boj.or.jp/en/mopo/gp 2011/> (accessedJuly 15, 2011).

Barker, K., Haimes, Y. Y., 2009. Assessing uncertainty in extreme events:applications to risk-based decision making in interdependent infrastructuresectors. Reliability Engineering and System Safety 94 (7), 819-829.

24

Barker, K., Santos, J. R., 2010a. Measuring the efficacy of inventory witha dynamic input-output model. International Journal of Production Eco-nomics 126 (1), 130-143.

Barker, K., Santos, J. R., 2010b. A risk-based approach for identifying keyeconomic and infrastructure sectors. Risk Analysis 30 (6), 962-974.

Boisvert, R. N., 1992. Indirect losses from a catastrophic earthquake and thelocal, regional, and national interest. In: Indirect Economic Consequencesof a Catastrophic Earthquake. Development Technologies, Inc., pp. 207-265.

Calunsod, R., 2011. Japan supply disruptions test Philippine industries’ re-silience. BBC Monitoring Asia Pacific, May 10.

Cronin, F. J., 1984. Analytical assumptions and causal ordering in interindus-try modeling. Southern Economic Journal 51 (2), 521-529.

Crowther, K. G., Haimes, Y. Y., 2010. Development of the multiregionalinoperability input-output model (MRIIM) for spatial explicitness in pre-paredness of interdependent regions. Systems Engineering 13 (1), 28-46.

Davis, H. C., Salkin, E. L., 1984. Alternative approaches to the estimationof economic impacts resulting from supply constraints. Annals of RegionalScience 18 (2), 25-34.

Dietzenbacher, E., 1997. In vindication of the Ghosh model: a reinterpreta-tion as a price model. Journal of Regional Science 37 (4), 629-651.

Ellram, L. M., Cooper, M. C., 1993. The relationship between supply chainmanagement and keiretsu. The International Journal of Logistics Manage-ment 4 (1), 1-12.

Ellson, R. W., Milliman, J. W., Roberts, R. B., 1984. Measuring the regionaleconomic effects of earthquakes and earthquake predictions. Journal ofRegional Science 24 (4), 559-579.

Geithner, T., 2011. Meet the Press, July 20. <http://www.msnbc.msn.com/id/43672884/ns/meet the press-transcripts/t/meet-press-transcript-july/> (accessed August 30, 2011).

25

Ghosh, A., 1958. Input-output approach in an allocation system. Economica25 (97), 58-64.

Gordon, P., Moore II, J. E., Richardson, H. W., Pan, Q., 2005. The economicimpact of a terrorist attack on the twin ports of Los Angeles-Long Beach.In: Richardson, H. W., Gordon, P., Moore II, J. E. (Eds.), The EconomicImpacts of Terrorist Attacks. Edward Elgar, Cheltenham, UK, pp. 262-286.

Hallegatte, S., 2008. An adaptive regional input-output model and its appli-cation to the assessment of the economic cost of Katrina. Risk Analysis 28(3), 779-799.

Hallegatte, S., 2011. Modeling the role of inventories in the assessmentof natural disaster economic costs: application on Hurricane Katrina inLouisiana, submitted for publication.

Hallegatte, S., Przyluski, V., 2010. The economics of natural disasters:concepts and methods. Policy Research Working Paper 5507. Officeof the Chief Economist of the Sustainable Development Network, TheWorld Bank <http://ssrn.com/abstract=1732386> (accessed September25, 2011).

Hirsch, J., 2011. Auto sales remain lackluster as carmakers skimp on incen-tives. Los Angeles Times, July 2.

Isard, W., Azis, I. J., Drennan, M. P., Miller, R. E., Saltzman, S., Thor-becke, E., 1998. Methods of Interregional and Regional Analysis. AshgatePublishing, Aldershot, UK.

Japan, 2011a. Foreign trade and balance of payments. Japan MonthlyStatistics, Statistics Bureau, Director-General for policy Planning(Statistics Standards) & Statistical Research and Training In-stitute <http://www.stat.go.jp/english/data/getujidb/index.htm> (ac-cessed August 20, 2011).

Japan, 2011b. Indices of industrial production (revised report). Researchand Statistics Department, Economic and Industrial Policy Bureau, Min-istry of Economy, Trade and Industry (METI), May 19, June 14, andJuly 13 <http://www.meti.go.jp/english/statistics/tyo/iip/index.html>(accessed September 25, 2011).

26

Japan, 2011c. Provisional report of monthly labour survey - Au-gust 2011. Employment Statistics Division, Ministry of Health,Labour and Welfare <http://www.mhlw.go.jp/english/database/db-l/23/2308pe/2308pe.html> (accessed October 4, 2011).

Japan, 2011d. Table 1, commercial sales value by type of business andthe percentage change from the same month/term of the previous year.Preliminary report on the current survey of commerce, Office of CurrentSurvey for Service Industry, Research and Statistics Department, METI<http://www.meti.go.jp/english/statistics/tyo/syoudou/index.html>(accessed July 23, 2011).

Kratena, K., Streicher, G., 2009. Macroeconomic input-outputmodelling – structures, functional forms and closure rules. In:E. Dietzenbacher, B. Thage (Eds.), Working Papers in Input-Output Economics. International Input-Output Association<http://www.iioa.org/working%20papers/WPs/WPIOX09-009.pdf>(accessed September 25, 2011).

Lam, L., 2011. Camera prices rise after earthquake. South China MorningPost, April 12.

Leontief, W. W., 1936. Quantitative input and output relations in the eco-nomic systems of the United States. Review of Economics and Statistics18 (3), 105-125.

Leontief, W. W., 1952. Input-output economics. Scientific American 185 (4),15-21. Reprint in: Leontief, W. W., 1986. Input-Output Economics, seconded. Oxford University Press, New York, pp. 3-18.

Lian, C., Haimes, Y. Y., 2006. Managing the risk of terrorism to interde-pendent infrastructure systems through the dynamic inoperability input-output model. Systems Engineering 9 (3), 241-258.

Lin, X., Polenske, K. R., 1998. Input-output modeling of production pro-cesses for business management. Structural Change and Economic Dy-namics 9 (2), 205-226.

Lohr, S., 2011. Japan appears dispensable as a supplier. The New YorkTimes, May 30.

27

MacKenzie, C. A., Barker, K., 2011. Conceptualizing the broader impactsof industry preparedness strategies with a risk-based input-output model.In: Lahr, M. L., Hubacek, K. (Eds.), Proceedings of the 19th Interna-tional Input-Output Conference <http://www.iioa.org/Conference/19th-downable%20paper.htm> (accessed September 24, 2011).

MacKenzie, C. A., Barker, K., Grant, F. H., 2012. Evaluating the con-sequences of an inland waterway port closure with a dynamic multire-gional interdependency model. IEEE Transactions on Systems, Man, andCybernetics-Part A: Systems and Humans 42 (2), 359-370.

Nanto, D. K., Cooper, W. H., Donnelly, J. M., Johnson, R.,. 2011. Japan’s2011 earthquake and tsunami: economic effects and implications for theUnited States. Congressional Research Service, The Library of Congress,Washington, DC.

Okuyama, Y., 2008. Critical review of methodologies on disas-ter impacts estimation. Background paper for the World Bank- UN Assessment on the Economics of Disaster Risk Reduction.<http://gfdrr.org/gfdrr/node/284> (accessed September 23, 2011).

Okuyama, Y., Hewings, G. J. D., Sonis, M., 2004. Measuring economic im-pacts of disasters: interregional input-output analysis using sequentialinterindustry model. In: Okuyama, Y., Chang, S. E. (Eds.), ModelingSpatial and Economic Impacts of Disasters. Springer-Verlag, Berlin, pp.77-101.

Oosterhaven, J., 1988. On the plausibility of the supply-driven input-outputmodel. Journal of Regional Science 28 (2), 203-217.

Oosterhaven, J., 1989. The supply-driven input-output model: a new inter-pretation but still implausible. Journal of Regional Science 29 (3), 459-465.

Organization for Economic Co-Operation and Development, 2011.OECD.StatsExtrats <http://stats.oecd.org/Index.aspx> (accessedSeptember 25, 2011).

Orsi, M. J., Santos, J. R., 2010. Incorporating time-varying perturbationsinto the dynamic inoperability input-output model. IEEE Transactions onSystems, Man, and CyberneticsPart A: Systems and Humans 40 (1), 100-106.

28

Orsi, M. J., Santos, J. R., 2010. Estimating workforce-related economic im-pact of a pandemic on the Commonwealth of Virginia. IEEE Transactionson Systems, Man, and CyberneticsPart A: Systems and Humans 40 (2),301-305.

Park, J. Y., 2008. The economic impacts of dirty bomb attacks on the Los An-geles and Long Beach ports: applying the supply-driven NIEMO (nationalinterstate economic model). Journal of Homeland Security and EmergencyManagement 5 (1), 1-20.

Rechtin, M., 2011. Toyota sales, inventories still weak; rebound predicted inJuly. Automotive News, July 1.

Rose, A., 2004. Economic principles, issues, and research priorities in hazardloss estimation. In: Okuyama, Y. Chang, S. E. (Eds.), Modeling Spatialand Economic Impacts of Disasters. Springer-Verlag, Berlin, pp. 13-36.

Rose, A., Guha, G.-S., 2004. Computable general equilibrium modeling ofelectric utility lifeline losses from earthquakes. In: Okuyama, Y., Chang, S.E. (Eds.), Modeling Spatial and Economic Impacts of Disasters. Springer-Verlag, Berlin, pp. 119-141.

Rose, A., Liao, S.-Y., 2005. Modeling regional economic resilience to disas-ters: a computable general equilibrium analysis of water service disrup-tions. Journal of Regional Science 45 (1), 75-112.

Rose, A., Benavides, J., Chang, S. E., Szczesniak, P., Lim, D., 1997. Theregional economic impact of an earthquake: direct and indirect effects ofelectricity lifeline disruptions. Journal of Regional Science 37 (3), 437-458.

Rowley, A., 2011. Japan reports highest monthly rise in factory output in 58yrs. The Business Times, June 30.

Salter, P., 2011. Just-in-time is on the rocks but not sunk yet. City A.M.,June 14.

Santos, J. R., 2006. Inoperability input-output modeling of disruptions tointerdependent economic systems. Systems Engineering 9 (1), 20-34.

Santos, J. R., 2008. Interdependency analysis with multiple probabilistic sec-tor inputs. Journal of Industrial and Management Optimization 4 (3), 489-510.

29

Santos, J. R., Haimes, Y. Y., 2004. Modeling the demand reduction input-output (I-O) inoperability due to terrorism of interconnected infrastruc-tures. Risk Analysis 24 (6), 1437-1451.

Shameen, A., 2011. Big money: dissecting the global supply chain. The Edge,April 4.

Sheffi, Y., 2005. The Resilient Enterprise: Overcoming Vulnerability forCompetitive Advantage. The MIT Press, Cambridge.

Shoven, J. B., Whalley, J., 1992. Applying General Equilibrium. CambridgeUniversity Press, New York.

Tabuchi, H., 2011. Toyota raises profit outlook 40%, defying strong yen andearthquake. The International Herald Tribune, August 3.

United Nations, Trade Statistics Branch, 2009. International Trade Statis-tics Yearbook Annex II: Country Pages. International merchandise tradestatistics <http://comtrade.un.org/pb/first.aspx> (accessed September25, 2011).

U.S. Department of Commerce, Bureau of Economic Analysis, 1997.Regional Multipliers: A User Handbook for the Regional Input-Output Modeling System (RIMS II), third ed. Washington, DC.<http://www.bea.gov/scb/pdf/regional/perinc/meth/rims2.pdf> (ac-cessed October 4, 2011).

Ward’s AutoInfoBank, 2011. WARD’S North America vehicle produc-tion by manufacturer. North American Production Summary - June2011. <http://wardsauto.com/keydata/NAProductionSummary/> (ac-cessed July 19, 2011).

Wassener, B., 2011. Exports drop more than expected in Japan. The Inter-national Herald Tribune, April 21.

30