INCLUDING COSTA RICA INTO AN INTERNATIONAL INPUT- OUTPUT TABLE, AN EXERCISE BASED ON THE WIOD

Transcription

1 INCLUDING COSTA RICA INTO AN INTERNATIONAL INPUT- OUTPUT TABLE, AN EXERCISE BASED ON THE WIOD David Buón Satoshi Inomata Bo Meng Tayutic Mena Nataia Sánchez Gabriea Saborio May 2014 This research was possibe thanks to the support of IDE JETRO, the Ministry of Foreign Trade of Costa Rica and the Centra Bank of Costa Rica. ABSTRACT This paper describes the methodoogy used to incude Costa Rica into a preexisting Internationa Input-Output Tabe (IIOT), in this case the Word Input Output Tabe (WIOD). We have embedded the first nationa IOT of Costa Rica into the WIOD by using a detaied biatera approach that ensures consistency with the nationa accounts and the WIOD framework. The paper highights some new methodoogica insights that coud be of use to other countries that seek to insert their domestic IOT into an internationa IOT. This is the first IOT of Costa Rica, and the first attempt to incude a county into an aready existing IIOT. This exercise reveaed key areas in which Costa Rica s data and partner s country data coud be strengthened to improve the capture of information, especiay in the area of trade in services. The views expressed herein are those of the authors and do not necessariy refect the views of the accompanying institutions by David Buón, Satoshi Inomata, Bo Meng, Tayutic Mena, Nataia Sánchez and Gabriea Saborio. A rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without expicit permission provided that fu credit, incuding notice, is given to the source. 1.

2 1. Why are Internationa Input-Output Tabes (IIOTs) usefu for the anaysis of trade in vaue-added? The vaue-added approach to internationa trade is neither new nor surprising. It has been widey discussed at venues ike the Word Trade Organization (WTO). In practice, however, the idea was hardy reaized due to the ack of an appropriate methodoogy and database. The conventiona approach to tracing cross-border vaue chains can be found in the studies that use firms micro-eve data. As seen in the famous exampe of the iphone s production networks, the approach generay aims to identify the structure of the production process and/or the saes networks of a particuar product, based on the information provided by manufacturers. These firm-eve approaches are usefu in drawing the actua structure of suppy chains since they utiize the data directy provided by individua firms rather than resorting to any forms of statistica inference. The weakness, however, is aso apparent in its fipside. Firsty, their appicabiity is imited in considering macroeconomic issues ike trade poicies, since the anaytica focus is cast ony to a particuar product and/or activity of a few firms. It is far from being sufficient to capture the entire vaue fows at the nationa context. Second, the majority of firm-eve data does not expicity present compensation of empoyees, an important component of vaue-added items in the nationa account framework, but merges it with other types of production costs. The vaue-added anaysis based on a firm s micro-data, therefore, is bound to be an approximation by the information on a firm s operating surpus (profit). Finay, since vaues are generated at every point of the production process, the vaue-added anaysis shoud be abe to trace a the production stages aong the entire suppy chain. The firm-eve approach, however, ony considers the vaue-added structure of direct input suppiers (the first tier), but eaves a the rest of the vaue-added stream untracked. 1 Given these imitations of the conventiona approach, increasing attention is directed to a new strand of studies that use IIOTs. An IIOT provides a comprehensive mapping of internationa transactions of goods and services. Since the tabes contain information on suppy-use reations between industries across countries, which is totay absent in foreign trade statistics, it is possibe to identify the vertica structure of internationa production sharing. Unike the firm-eve approach, the input output anaysis covers an entire set of industries that comprise an economic system, and thereby enabes to capture the crossborder vaue fows at the eve of a country or a region. Theoreticay, it has capacity to track the vaue-added generation process of every commodity in every country at every production stage. 2. The basic framework of IIOTs The IIOT is a major effort bringing together pieces taken from various nationa input-output 1 Monge Arino (2011) presents a rare research exampe that successfuy overcame these anaytica imitations. By conducting an extensive survey on the suppy-use reations of eading companies in Costa Rica, the study identified the vaue-added structure of the country s core economic system. Though insightfu, it is however difficut to envisage the approach to be appied to other countries, since its feasibiity is fundamentay attributabe to the idiosyncratic feature of Costa Rica that a few number of mutinationa corporations (such as Inte) are assumed to sufficienty represent the nationa economy.

3 tabes, and hence the resuts can be read exacty in the same manner as for nationa tabes. The major difference being that it expicity presents internationa transactions between industries in the form of import matrices and export matrices by trading partners, which enabes us to draw a comprehensive mapping of goba production networks. IIOTs can take various forms depending on anaytica purposes and data avaiabiity. Figure 1 presents a genera format of the tabe based on the recommendation in the United Nations Handbook on the compiation of input-output tabes. Figure 1 Schematic image of an internationa input-output tabe (three-country case) Internationa input-output tabe for Country A, B and C Industry/ FD/VA codes Intermediate demand Fina demand Country A Country B Country C Country A Country B Country C PCE GCE GFCF PCE GCE GFCF PCE GCE GFCF Export to ROW Tota outputs Country A Z AA Z AB Z AC Y AA Y AB Y AC E AW X A Country B Z BA Z BB Z BC Y BA Y BB Y BC E BW X B Country C Z CA Z CB Z CC Y CA Y CB Y CC E CW X C Net taxes on products of trading partners Internationa freight and insurance costs TZ A IZ A TZ B IZ B TZ C IZ C TY A IY A TY B IY B TY C IY C Rest of the Word Z WA Z WB Z WC Y WA Y WB Y WC Duties & import commodity taxes DZ A DZ B DZ C DY A DY B DY C Wages Vaue added Op.surpus V A V B V C Tota Inputs X A X B X C Source Drawn by the author. Each ce in the coumns of the tabe shows the input composition of industries of the respective country. The sub-matrix Z AA, for exampe, shows the input composition of Country A s industries in reation to domesticay produced goods and services, that is, the domestic transactions of Country A. Z BA and Z CA in contrast show the input composition of Country A s industries for the imported goods and services from Country B and those from Country C, respectivey. Z WA indicate the imports from other countries, as grouped under the name

4 of Rest of the Word. TZ A, IZ A and DZ A give domestic taxes (net), internationa freight and insurance, and duties and commodity taxes evied on import transactions of Country A s industries. The fourth coumn of sub-matrices from the eft shows the composition of goods and services that have gone to the fina demand sectors of Country A. Y AA and Y BA, for exampe, present respectivey the goods and services produced domesticay and those imported from Country B that fow into Country A s fina demand sectors. The rest of the coumn is read in the same manner as for the 1 st set of sub-matrices of the tabe. E* W are exports (vectors) of each country to the Rest of the Word, and V* and X* are vaueadded and tota input / tota output, as seen in the conventiona nationa input output tabe. Figure 2 shows compositiona correspondence of the IIOT to a nationa input-output tabe. It shoud be noted that the export vectors to the Rest of the Word contain various statistica discrepancies arising out of data conficts among different sources since the vectors are constructed as residuas of the entire IIOT matrix. One of the sources of discrepancies stems from the fact that during the inking process of nationa input-output tabes, the export data in each nationa tabe is repaced by the import matrices of trading partners, and hence the probem of symmetries arises. Theoreticay, Country A exports of a particuar product to Country B shoud be equa to Country B imports of the same product from country A, given that they are compared at the appropriate vauation scheme (c.i.f. or f.o.b.). In practice, this is often not the case due to many reasons, for exampe - Improper decaration of product cassification at the customs border, either entry or exit; - Confusion about the rea country of origin/destination for re-exported products; - Shipping time-ag across different accounting periods (quarters or years); - Presence of merchandising trade; -Smugging; and - Other unrecorded transactions. The discrepancies can be aso attributed to inappropriate estimation of transaction margins (internationa freight and insurance costs, trade and transport margins, various taxes) in converting the vauation scheme of import matrices. Furthermore, there may be mismatches between the record of internationa transactions in SUTs/Nationa Accounts and the custom statistics/baance of payments statistics. Some existing IIOTs, notaby the Inter-country Input-Output Tabes of the OECD and the Word Input-Output Tabes of the European Commission (in which the Costa Rican I-O tabe has been embedded, as described ater) have featured Rest of the Word as a singe endogenous region in the transaction matrices. This aows the inter-country Leontief inverse to be derived with respect to the corresponding segments, see Dietzenbacher, et.a (2013) for the derivation method of endogenous Rest of the Word matrices in the case of the Word Input-Output Tabes, see the next section.

5 Figure 2 Compositiona correspondence of the IIOT to a nationa input-output tabe Industry/ FD/VA codes Source Drawn by the author. Country A Country B Country C Net taxes on products of trading partners Internationa freight and insurance costs Rest of the Word Duties & import commodity taxes Vaue added Wages Op.surpus Tota Inputs Intermediate demand Country A Country B Country C Country A Internationa input-output tabe for Country A, B and C * Light-shaded segments are directy transpanted from the origina nationa I-O tabe (country A), whie dark-shaded segments require some processing before being embedded into the IIOT. PCE GCE GFCF Dom estic products Im ported products Vaue added Fina demand PCE Country B GCE GFCF Industry/ FD/VA codes Wages Op.surpus Tota Inputs PCE Country C GCE GFCF Interm ediate dem and Export to ROW and PCE stat. discrepancies GCE Tota outputs Fina demand GFCF Export Country A's nationa inputoutput tabe Tota outputs

6 3. Construction of the nationa IOT The IOT for 2011 was constructed in accordance with the best practices recommended by the System of Nationa Accounts 2008 (SNA 2008), adopted by the United Nations Statistica Commission (UNSC). The SNA 2008 addresses important issues brought about by changes in the economic environment, advances in methodoogica research and the needs of users. The process used for the domestic construction of the IOT draws upon data from the fuy baanced version of the SUT for 2011 in basic prices. The IOT was buit at a product eve rather than industry eve. A secondary production was assigned to the same categories as primary production and intermediate consumption of secondary consumption was aso aocated to the primary product category. In a other aspects, the SNA 2008 methodoogy was foowed Basic ayout of the SUT The SUT incudes 183 products and 136 economic activities or industries (see appendix for detaied ist). These were cassified using the Centra Product Cassification (Rev 2), the Centra American Tariff System (2013), the Broad Economic Categories (Rev 4) and the Internationa Standard Industria Cassification (Rev 4) Data sources Domestic Production Domestic production was estimated using saes data from the Centra Bank s Registry of Economic Variabes for Businesses and Estabishments (REVBE), which contains over 26,500 observations. Internationa production Import data from the Costa Rican Genera Customs Directorate and the Baance of Payments data from the Centra Banks were used to buid an import matrix in CIF vaues that contains intermediate demand, fina demand and fixed capita formation of imported products. The REVBE was used to assign imports to specific economic activities. The Ministry of Finance s Annua Decaration of Cients, Suppiers and Specific Expenses (ADCSSE) was used in conjunction with administrative data from the REVBE to determine the destination of saes in order to distribute imports to the industries that use that use them. The ADCSSE is used by a firms and organizations that pay income taxes to decare saes and purchases over 2.5 miion coones (approximatey USD 5,000) and rentas, professiona services and commission and interest over 50 miion coones (approximatey USD 100,000). These vaues were adjusted from CIF to FOB by removing the insurance and transport margin. Taxes and subsidies a) Vaue added tax was cacuated for fina consumption by househods, intermediate consumption and fixed capita formation.

7 Fina consumption by househods The appicabe tax rate for each product category was estimated using data from the Centra Bank s Nationa Survey of Income and Expenditures, appying a 13% tax to a taxabe goods and services consumed by each of the 2,500 househods in the sampe and then extrapoating for a househods. Intermediate consumption The Genera Saes Tax Law was used to determine the activities for which different firms are aowed to deduce the saes tax from their purchase of goods and services. This information was used to cacuate the share of economic activities that are egay exempt from taxes and the share for the rest of the activities. The government and a companies in the Free Trade Zone Regime are exempt from paying taxes. Fixed capita formation Since the majority of products in the category are imported, the 13% tax rate was appied after adjusting for a customs taxes. Goods with taxexempt status were not incuded in the cacuation, such as machinery and farm equipment. The 13% tax rate was appier directy to gross domestic capita formation. b) Import and export taxes The appicabe tax rate was estimated based on data from the Costa Rican Genera Customs Directorate. c) Seective consumption tax and other taxes on products The appicabe tax rate was estimated based on information provided by the Ministry of Finance. Distribution margins Administrative data from the Ministry of Finance for a sampe of 620 companies was used to cacuate trade and transport margins for each product by usage (intermediate consumption, fina consumption, gross capita formation and exports). Domestic usage The structure of domestic usage was estimated using the foowing data sources a) Agricuture A sampe of 100 case studies which were constructed based on severa site visits and interviews with the main producers of 38 products. b) Services and manufacturing A sampe of 450 companies was constructed pooing data from three data sources. The first was the Centra Bank s Economic Study of Firms, which incudes firms with more than five empoyees and above the 15 th percentie in revenues. The second was The Trade Promotion Agency s financia and administrative data from firms in specia economic regimes. The third was interviews with specific firms for carification purposes. b) Whoesae and retai trade Trade production was assumed equivaent to trade margins. c) Construction Both pubic and private construction projects were incuded. Data on state owned construction projects was obtained from the Ministry of Transportation and from oca governments. Data on private construction projects was obtained trough the Federated Association of Engineers and Architects, an institution required by aw to coect this information in the process of granting construction permits. d) Government and state owned enterprises Administrative data reported by a of these institutions, incuding financia and expenditure data for each economic activity.

8 e) Financia intermediation and insurance services Census of financia firms, except cooperatives and pawnshops, for which there was a sampe of 177 enterprises. Data from audited financia statements was obtained from the financia reguatory bodies. Internationa usage Export data from the Genera Customs Directorate and the Baance of Payments data from the Centra Bank. Fina consumption of househods This data was estimated based on the Nationa Institute for Statistics and Census Nationa Househod Survey (ENIG) for 2004, and extrapoated to the year 2011 using price and voume indicators. This survey contains information for 2,415 products and was compemented with saes records from the Ministry of Finance. Fina consumption of NGOs This data was estimated based on information reported by non-profit associations coected through the Centra Bank s Economic Enterprises Survey. Fina consumption of the government A dataset was constructed from basic data provided by government institutions. Fixed capita formation Since most goods in this category are imported, data was constructed based on foreign trade records from the Genera Customs Directorate and information required by aw for the Ministry of Finance. This data was used to determine which products are capita assets and the fina economic activities in which they are used. Additiona data from the Federated Association of Engineers and Architects was aso used. Changes in inventories Tax decarations to the Ministry of Finance were used Recommendations adopted from SNA 2008 The foowing new standards were adopted from the SNA 2008 recommendations Exports of manufacturing services Ony imports and exports that incude a change in ownership are registered. In the case of Costa Rica, this affected 14 arge firms. Centra Bank production A distinction is drawn between non-market services, estimated using costs, and market services, estimated using to revenues. Financia intermediation

9 services are treated as market services. Services that reate to monitoring, currency stabiity and the production of statistics are considered as non-market activities. Production of non-ife insurance Compensations and premiums were smoothed over time in order to remove the appearance of voatiity from the sector. Indirecty measured financia intermediation services A oans and deposits offered by or deposited in financia institutions are incuded and cacuated using a common risk-free interest rate. 4. Seection of the internationa IOT At a very high eve, there are two possibe approaches to constructing an internationa IOT. The first is to construct the entire tabe using nationa sources of a the reevant countries. This is the strategy that organizations such as IDE-JETRO, the OECD and WIOD have foowed to construct their IIOTs. The second approach is to add new countries to an existing Internationa IOT that was buit using the first approach. This research project foowed the second approach in order to faciitate the integration of Costa Rica s data into an IIOT. When seecting the motherboard tabe in which to integrate Costa Rica, the foowing pre-existing IIOT s were considered Database Tabe 1 Main characteristics of existing IIOT s Producing organisation Reference years AIIOT IDE-JETRO EORA University of Sydney yeary Number of countries Industry/product cassification (1975) 78 ( ) 76 ( ) Approx ~500 EXIOPOL Database (CREEA) GTAP-MRIO European Commission Purdue University ROW 129 industries 129 products ICIO OECD industries 56 + ROW 18 industries WIOD European Commission yeary 40 + ROW 35 industries 59 products Source IDE-JETRO

10 The foowing criteria were considered in the seection of the motherboard tabe (a) Country coverage An economic interdependency anaysis was performed to determine Costa Rica s main trade and investment partners. It is important to note that severa of the main trade partners are not in any of the pubicay avaiabe IIOTs and do not even have a domestic IOT. (b) Sector coverage Since the domestic IOT is quite detaied, we wanted to seect a tabe that woud take advantage of these sectors. (c) Assumptions made Typicay, the tabes with the greatest number of sectors and countries are based on broad underying assumptions, which eads to a significant margin of error in any of the concusions that can be drawn from the tabe. WIOD provided a country coverage that incude most of the main partners of Costa Rica, whie its product cassification aowed to focus on certain sectors of particuar interest for Costa Rica. Finay, the fact that its methodoogy is derived from a SUT framework, shoud provide a good fit for the data provided by Costa Rica. 5. Converting the domestic IOT into the format used by the WIOD (a) Prices Currency The domestic IOT is in Costa Rican coones, whereas the WIOD is in doars. 2 For the case of the IOT, the Costa Rica Centra Bank provided the annua average exchange rate used in the nationa accounts. Current vs constant prices Both the domestic IOT and the WIOD are in current prices. Whie the WIOD has a time series incuding severa years in constant prices, the tabe for individua years is in current prices. Basic vs producer prices Both the domestic IOT and the WIOD are in basic prices. (b) Sector categorization The WIOD uses an industry-by-industry approach to define the sectors in the tabe, whereas Costa Rica s domestic IOT for 2011 was constructed using a product by product approach, even though the Suppy and Use Tabe (SUTs) contain the necessary information. The survey data used to construct the IOT reveaed that in the case of Costa Rica firms rarey engage in secondary activities, which means that the sector categorization according to products and industries is very simiar. Future versions of the domestic IOT wi use an industry-by-industry approach. In order to ensure consistency, the Centra Bank provided the domestic IOT in the ISIC rev 2 cassification used by the WIOD. (c) Year When the research project begun, the ast year avaiabe in the WIOD time series was In order to insert the domestic IOT for 2011 in the 2009 WIOD, we first backdated the Costa Rican tabe through the foowing steps Aggregated eve data (tota intermediate use by industry, househod consumption, government consumption and gross capita formation and exports) incuded in the tabe was obtained from the information avaiabe in nationa 2 The WIOD uses the exchange rates provided by the IMF to convert nationa currencies to doars

11 accounts, specificay from the 2009 SUT. A new matrix for 2009 was created inputting SUT data in the row and coumns totas, with banks in the rest of the information. The data on intermediate demand for 2009 was obtained using a RAS procedure. This agorithm made use of the interna shares of the intermediate demand matrix for 2011, and the tota intermediate demand and intermediate suppy from the SUT for 2009, appying an iterative procedure adjusting coumn and row vaues. This method is the most widey known and commony used, automatic baancing procedure. In addition, this bivariate method corrects rows and coumns on one input-output tabe at the same time. In order to guarantee that the RAS agorithm woud run smoothy, a zero vaues found in the vectors of tota intermediate consumption and tota production, were adjusted to become positive vaues cose to zero. At the end of the backdating process, the intermediate consumption matrix obtained through the RAS method had to match these totas. These adjustments were minor reative to the tota vaue contained in the symmetric input-output tabe, and fa within the norma statistica error Spitting vectors from the domestic IOT to prepare for insertion in the WIOD 5.2. Exports matrix The domestic IOT contains and exports vector by product category, however it does not specify the country or sector destinations. In order to integrate the domestic IOT into the WIOD, the export vector must first be spit according to the country and sector destinations. Merchandise exports and service exports were treated separatey Merchandise exports Spitting by country In order to spit the export vector by country destinations there are two fundamenta methodoogica decisions, presented in the tabe beow Origin s export data Destination s import data Shares using origin s data Shares using destination s data Consistent with SNA. Does Consistent with SNA. Can better refect not capture fina destination fina destination, use and goods vaues, and can miss the actua but CIF vaues and trade discrepancies vaue of exported goods. can distort trade structure. Not consistent with SNA, can better Not consistent with SNA, refect fina destination, use and goods trade structure does not vaues. But, CIF vaues and trade refect fina destination or discrepancies can distort trade goods vaues. structure.

12 US$ Biion The first is whether to use the country of origin s export data or the country of destination s import data. In genera the destination s import data is considered more reiabe because it makes better adjustments for re-exports and because the recorded vaues of traded products tends to be more accurate. However, the disadvantage of using the destination s import data is that if there are arge discrepancies between the origin s export data and the destination s import data, then the numbers used wi differ significanty from those used in the nationa account data utiized to cacuate the country s GDP. The second is whether to cacuate the shares using the origins data or the shares using the destination s data. According to seected partners data 3, by 2012 Costa Rica exported more than US$ 37.9 biion; neary 3.7 times the amount reported by Costa Rica. Discrepancies are aso growing in time for the period , mirror data shows a Compound Average Growth Rate (CAGR) of 24%, whie nationa data shows a CAGR of 4.4%. By 2012 the difference between nationa data and mirror data attained US$ 27.7 biion (see next figure). Figure 3 Exports of Costa Rica according to source, Exports of Costa Rica (according to partner) Exports from Costa Rica (according to CR) Source COMEX with data from WITS, Eurostat, USITC, PROCOMER and Trade Map. Note trade data incudes ony countries with reported positive imports from Costa Rica by Even though discrepancies are growing, they are concentrated in a few partners and products. By 2012, three partners accumuated 72% of the accounted differences United States, European Union and China. Certain tariff ines mosty reated to microprocessors and computer parts show arge discrepancies due to the way in which inteectua property is assigned 3 Seected countries represented an average of 91% of Costa Rica s exports between

13 Tabe 2 Products exported by Costa Rica with the most important discrepancies US$ miion HS-6D Description Partner Difference Particip. % in Dif.* Eectronic integrated circuits as processors and controers Parts & accessories of computers Partner CR Partner CR Dif Dif USA 6, , ,608 7,027 81% 85% China 3, , ,555 4,846 98% 98% EU 6, , ,178 6,945 89% 91% Source COMEX with data from WITS, Eurostat, USITC, PROCOMER and Trade Map. Preiminary data. *Difference participation estimated using the tota of positive differences. For this reason we chose to use the origin`s export data in order to maintain the consistency with nationa accounts. However in order to refect the fact that in most tariff ines, the mirror data is ikey to be more accurate, we chose to assign shares of exports by destination according to mirror data shares. 4 Formay, et e i,k indicate the share in exports of product i from country k to a particuar country defined as e i,k = E i,k E i where E i,k is the tota vaue from products that are cassified by WIOD product group i exported by country k, and E i the tota export vaue of WIOD product group i. such that (1) E = E i e i,k (2) i Spitting by sector The chaenge here is how to combine the information avaiabe from the biatera trade statistics of Costa Rica with each of its partners; and the information aready avaiabe in the WIOD. ITS information, combined with BEC, heps to obtain shares at three use categories Intermediate consumption, fina consumption and capita goods. However, WIOD aready provides detaied information about the use made by each country of its imports from the rest of the countries into 40 categories 35 categories for intermediate consumption; three for fina consumption; one for capita and another one for change in 4 The best choice woud be to use the structure obtained from mirror data and appy it to the exports vector with nationa data. Certainy, it impies that as nationa vaues are used, Costa Rica s exports coud be underestimated for certain countries, but this effect can be baanced with those countries with a bigger share in mirror data

14 inventories. There are three steps that were foowed to make use of both sources of information First, for each country, imports in the WIOD were aggregated into one vector per sector. This generates a matrix that refects the genera structure of imports for each of the countries in the WIOD. This matrix contains the three arge broad economic categories from BEC- which are subdivided in turn into the 40 stated above. Change in inventories is incuded into the matrix and incuded as part of capita, as it modifies the tota output and needs to be taken into account. A share is obtained within each of the three broad groups. Formay, et m i,j indicate the share of product i used in product j within group (intermediate consumption, fina demand or capita goods), by a particuar country m i = M i.j (3) M i. where M i is the tota vaue of imports that are cassified by use category and WIOD product group i, and M i the tota vaue of WIOD product group imported. Interna shares I i were cacuated within each of these 3 broad groups, such that the sum of the vaues in every category woud add to 1. Formay, et n i,j indicate the share of the use j within each group of product i. where N i is the tota vaue within each sector that is cassified by use category (intermediate consumption, fina demand and capita goods) and WIOD product group i, and N i the tota vaue of interna share of WIOD product group i. A WITS database was downoaded to capture a the countries in the word s imports from Costa Rica for 2009 in HS2007 at the 6 digit eve. In order to make this database usabe, the HS vaues were converted into the WIOD s 35 industry categories and the three BEC categories discussed above. In the case of the BEC categories motor spirits, cars, and others, the foowing basic assumptions were made to distribute them in the other categories oi was distributed 50 % in intermediate consumption and 50% in fina consumption, cars and others were distributed 50% in fina consumption and 50% in gross fixed capita formation. The aggregated WIOD and the WITS database were then used to create the export matrix by destination country and sector. The BEC shares from the WITS database were mutipied by the sector shares from the aggregated WIOD, to create the distribution by destination usage (BEC and sector) where, Export matrix equas Where, M E M = M i j m i,k

15 Formay, et m i,k indicate the share of use categories (intermediate consumption, fina demand and capita) in imports from product i by a particuar country from country k defined as m i,k = M i,k M i where M i,k is the tota vaue from import products that are cassified by use category and WIOD product group i imported from country k = Costa Rica, and M i the tota vaue of WIOD product group i imported from Costa Rica Services exports a) Spitting by country In the case of Costa Rica, systematic survey data on the origin of services imports and exports destinations is not avaiabe. However, where possibe other data sources were utiized to estimate the data for each service sector, and for this reason each sector wi be treated separatey Renting of Machinery and equipment; and other services (this sector represents 15.6% of the tota exports of Costa Rica). Unfortunatey, there is no additiona data regarding the destination of the exports reated to this category. However based on the SUT, this sector consists mosty of offshore services. Furthermore four aternates were considered i. To use the exports structure of a country in the region (time zone) with a significant amount of FDI in services. Mexico and Canada. ii. To use the shares of origin of FDI infows in services, which woud assume that the services are aways exported to the country of ownership. iii. To use certain goods that are compementary to services. iv. To use the genera structure of goods Ideay, there woud be enough data to evauate the degree to which each of the underying assumptions above is most vaid in order to decide which to use. A second best approach woud be to compare the three options using sensitivity anaysis. Hotes and Restaurants This sector represents 11.7% of the tota exports of Costa Rica. For this sector the distribution of incoming tourists during 2009 based on migration data coected at a entry points was considered 5. This data was weighted according to the expenditures per capita by tourists from each country obtained from surveys to exiting tourists by the Costa Rica Tourism Board. Education, heath and socia-work This sector represents 0.6% of the tota exports of Costa Rica. Data from Centra Bank of Costa Rica based on sectoria consutations. Whoesae trade and commission trade, except motor vehices and motorcyces. This sector 5 This data was ceaned foowing the Chamber of Tourism approach, by removing tourists from Nicaragua who entered by and 5, as they are mosty seasona migrants

16 represents 2.4% of the tota exports of Costa Rica. In this case, the structure for merchandise exports of the goods was used, given the ack of data on export destinations in this particuar sector. Air transport. This sector represents 2% of the tota exports of Costa Rica. In this case there are no significant domestic airines which provide internationa services, which means that a services in this sector are airport services. The Genera Directorate of Civi Aviation provided the frequency of fights from different airines, and these airines were identified by the country of ownership in order to distribute the airport services according to destinations. However, data avaiabe was not abe to distinguish between cargo panes and passenger panes. Inand transport This sector represents 1.9% of the tota exports of Costa Rica. Inand transport services coud correspond to imports or exports of goods to or from Costa Rica. The foowing assumptions were made -The transportation service is aways paid for by firms in the destination country -A goods exported through and borders are transported by companies with a ega presence in Costa Rica. Given these assumptions we used customs data regarding and exports to identify the distribution of fina destinations. Other supporting and auxiiary transporting activities; activities of trave agencies This sector represents 1.1% of the tota exports of Costa Rica. Given the ack of data, exports were distributed according to the structure of exports of merchandise. Other sectors with non-zero vaues The remaining categories have ess than 0.5% each, and it was therefore considered that the basic assumption under which those sectors have the same structure as the exports of goods was appropriated. The foowing sectors had non-zero vaues post and teecommunications; financia intermediation; rea estate activities; and other community, socia and persona services. Sectors with no registered export vaues construction, sae, maintenance and repair of motor vehices and motorcyces; retai trade, water transport; pubic administration and defense; and private househods with empoyed persons. b) Spitting by sectors In the case of services two steps were foowed -The same aggregated WIOD matrix described in the case of merchandise was created for services sectors. -In the case of services, data was not avaiabe to determine the BEC categories of exports. Therefore, it was assumed that service imports (exports) M from Costa Rica were used in the same proportion as the imports (exports) from the tota of the other countries, where E = S i j e i,k Formay, et S i Costa Rica tota export of services e i,k indicate the share of use categories (intermediate consumption, fina demand and capita) in exports to product i by a particuar country from country k.

17 5.3. Import matrix Besides the domestic IOT, Costa Rica Centra Back prepared an import matrix that contains imports with the sector of origin and the sector of destination in CIF vaues for the merchandise reated sectors, but adjusted in the totas to turn it FOB. In order to integrate the domestic IOT into the WIOD, the import matrix must be spit according to the country origins in FOB vaues. Merchandise exports and service exports were treated separatey Merchandise imports a) Preparing the data to cacuate ratios We obtained a database from the Genera Customs Directorate which contains imports by HS category by country of origin and mode of transportation in CIF and FOB vaues. We added to the data base the WIOD categories and BEC categories. M = E i j e i,k (1) Formay, et e i,k indicate the share of use categories (intermediate consumption, fina demand and capita) in imports of product i by a particuar country from country k defined as m i,k = M i,k M i where M i,k is the tota vaue from products that are cassified by use category and WIOD product group i imported to country k, and M i the tota vaue of WIOD product group i imported from a country. b) Spitting by country We first cacuated the share that corresponds to each country of origin within each WIOD sector using the FOB vaues. These shares were used to subdivide each of the origin sectors in the import matrix by countries of origin, incuding change in inventories. m i,k = M i,k M i c) Spitting by BEC Secondy, we cacuated the share of each BEC category within each WIOD sector by country using the FOB vaues. These ratios were used to subdivide each of the origin sectors (by country) by BEC category. This guarantees that the characteristics of each country and each BEC category within countries are maintained.

18 d) Converting from CIF to FOB vaues We used the CIF-FOB database to cacuate the ratio of FOB vaues to CIF vaues for each of the origin sectors by origin country and BEC categories. Each of the vaues in the spit import matrix were mutipied by the FOB /CIF ratios that corresponded to their BEC category. The ony assumption made is that the insurance and freight margins for the products with the same country sector of origin and BEC usage to not depend on the specific sector of usage. The trade insurance and freight margins themseves are incuded in imports of inand, water and air transport services and financia intermediation services sectors Services imports Since there is no data on imports that specifies origins, each sector was treated separatey. (a) Inand transport, air transport and financia intermediation Before addressing each of these sectors separatey, some previous work was competed to incude the insurance and freight margins extracted from the CIF vaues above into their respective sectors. We used the mode of transport at the port of entry in the same expanded database from Genera Customs Directorate to subdivide the freight costs to and, sea and air freight. Since the domestic IOT merges sea and air transport, these two categories were combined as air freight. This information was used to cacuate the air freight to CIF ratio, the and freight to CIF ratio and the insurance to CIF ratio for each country of origin and destination sectors. Each of these ratios was mutipied by the import matrix by countries, sectors and BEC to construct an import matrix for inand transport, air transport and financia intermediation. Each of these matrices were coapsed into a vector specifying each destination sector. These vectors were added to the vectors for each corresponding sector in the domestic IOT. The assumption is that a services in the insurance and freight margin are imported. After making these adjustments each sector was treated separatey Inand transport We used customs data to determine the share of countries of origin. The underying assumption is that the transportation service provider is from the country of origin of the import. A observations where the country of origin has no direct and contact with Costa Rica were eiminated in this cacuation. These shares were used to disaggregate the vector by countries Air transport We used data on incoming fights from the genera Directorate of Civi Aviation, assigning the country of origin of each airine. These shares were used to disaggregate the vector by countries and BEC categories. Using this proxy impies that a fights were given the same weight into the tota share by country. We foowed the Centra Bank decision to use the structure of air transport, since many ogistics firms do not separate the costs of air and sea transport.

19 Financia Intermediation The data on internationa financia transaction in Costa Rica is not avaiabe to researchers. As a second best, we decided to use the structure of countries of origin of financia intermediation imports for two simiar countries, Mexico and Canada. The country shares for the financia intermediation imports were cacuated for both countries, they were then averaged and re-indexed to obtain an estimate of the shares for Costa Rica. These shares were used to disaggregate the vector by countries. (b) Other supporting and auxiiary transporting activities, activities of trave agencies; post and teecommunications and renting of Machinery and equipment and other services In the case of these sectors, there is no avaiabe data or the sources are confidentia. As in the case of financia intermediation, we decided to use the structure of countries of origin of each sector s imports for two simiar countries, Mexico and Canada. The country shares for the sector s imports were cacuated for Mexico and Canada, they were then averaged and re-indexed to obtain an estimate of the shares for Costa Rica. These shares were used to disaggregate the vector by countries. (c) Sae, maintenance and repair of motor vehices and motor cyces Inquiries made by the Centra Bank throughout the surveying process suggest that 100 percent of trade in this sector was with Centra America. Since no other Centra American countries are in the WIOD, these exports were assigned to the Rest of the Word. These shares were used to disaggregate the vector by countries. (d) Hotes and Restaurants Data on Costa Rican tourists reported by destination countries does not often exist because exports to Costa Rican tourists are a very sma proportion of GDP. As an aternative, nationa migration data on destinations of Costa Rican tourists was used. It was assumed that a tourists spend the same amount during their stay due to ack of data. These shares were used to disaggregate the vector by countries. (e) No registered vaues Construction; Whoesae trade and commission trade except motor vehices and motorcyces; retai trade; water transport; rea estate activities; pubic administration and defense; Education; heath and socia-work; Other community, socia and persona services; private househods with empoyed persons 6. Insertion of domestic IOT in WIOD 6.1. Insertion In order to insert Costa Rica into the WIOD Costa Rica was extracted from the rest of the word both in terms of countries imports and countries exports, and interactions within row. Extract Costa Rica s exports to each country from exports of ROW to each country in the WIOD, incuding ROW. This first invoved inserting Costa Rica s spit export

20 matrix in a new row in the WIOD to capture Costa Rica s exports to each country. Then these vaues were subtracted from the exports of ROW to each country, in order to create the exports of ROW* to each country, where ROW* is ROW excuding Costa Rica. Extract Costa Rica s biatera imports from each country from imports of ROW to each country in the WIOD, incuding ROW. This first invoved inserting Costa Rica s spit import matrix in a new coumn in the WIOD to capture Costa Rica s imports to each country. Then these vaues were subtracted from the imports of ROW from each country, in order to create the imports of ROW* from each country, where ROW* is ROW excuding Costa Rica. Extract Costa Rica s domestic matrix from ROW s domestic matrix This first invoved inserting the intermediate and fina demand sections of Costa Rica s domestic IOT in the space in the WIOD representing interactions between Costa Rica and Costa Rica. Then these vaues were subtracted from the matrix of interactions between ROW and ROW. The vaue added matrix from the domestic IOT was inserted in the vaue added section of the WIOD for Costa Rica Deaing with negative vaues in ROW In some cases the vaues eft for exports to ROW* or imports from ROW* were negative. This occurred when the vaue imported or exported by Costa Rica is arger than the vaue which ROW imported or exported. In each of these cases the data used and assumptions made were revised and improved (where possibe) to reduce the incidence of negative numbers as much as possibe. In order to ensure that these cases do not affect anaysis of the matrix adversey, they were removed and paced in a separate category. The negative numbers that emerge from ROW s exports are grouped in a separate row and the negative numbers that emerge from ROW`s imports are grouped in a separate coumn. We engaged in further anaysis to quantify the extent to which there are discrepancies in the data and the sources of error. Negative numbers can emerge due to four reasons Costa Rica s data was ess accurate that the data used to construct the WIOD, even before spitting by countries and usages. Even though the aggregate data on exports or imports for each sector in Costa Rica is accurate, the underying assumptions about how to spit various vectors by country and usage, were inaccurate. Data provided by key countries were ess accurate that the data used by Costa Rica, even before spitting by countries and usages. Even though the aggregate data on exports or imports for each sector in partner countries is accurate, the underying assumptions about how to spit various vectors by country and usage were inaccurate. 7. Summary of methodoogica insights Internationa Input-Output Tabes are a vauabe too for countries to understand their interactions with the goba economy. For this reason, many deveoping countries that strive enhance their participation in goba vaue chains may be interested in construction their

21 own IIOT. This methodoogica paper has described the process through which Costa Rica was incuded in an Internationa IOT for 2009 based on the WIOD, with the objective of describing the process that other countries woud have to foow in order to buid their own tabe. The process that was foowed draws upon the guideines used to buid the WIOD. However, each country faces its own set of data constraints and case-specific assumptions must be made that everage existing data and oca contextua knowedge. There are, however, common chaenges that many countries are ikey to face and this paper addresses three which may be of particuar interest 1-Lack of access to partner country data Ideay, the process to buid an IIOT, woud use each of the country s origina data sources to construct and then baance the tabe. This is the approach that has been used to buid the WIOD, the OECD s ICIO and other efforts. However, from the perspective of a nationa statistics office or centra bank this process is inhibitive, mainy because of the difficuties invoved in obtaining data from each partner country. Whie it might be conceivabe to obtain SUTs from each country, the data and context-based knowhow required to make reasonabe assumption in order to incude each country in the IIOT is generay not accessibe. This paper describes a new second-best approach, which strives to incude new countries in a preexisting IIOT. This process invoves new steps that everage existing internationa data, such as the foowing Choosing a base IIOT There are a range of pubicay avaiabe IIOT s which can be used as a base tabe. The avaiabiity of severa options aows researchers to optimize their objectives by weighing the advantages and disadvantages according to three criteria the number of partner countries in the tabe, the eve of disaggregation and the robustness of the assumptions made in combining data sources. Making use of data in the base IIOT The base IOT contains a weath of data regarding the origins of each country s imports and the destination of its exports, both in terms of country and industry. This data can be everaged in cases in which domestic data is acking to estimate the shares to spit the import and export vectors for different products by country. Inserting the domestic IOT in the base IIOT In order to buid the IIOT, a creative approach must be used to extract the new country from the category caed rest of the word (ROW). This invoves extracting the new country s exports from ROW s exports, extracting the new countries imports from ROW s imports and extracting the new countries domestic IOT from ROW s domestic IOT. 2-Discrepancies in country s trade data reative to partner countries mirror data It is we known that there are significant discrepancies between the vaues that trade partners report for their biatera trade. The rise of goba vaue chains has exacerbated these discrepancies due to the ambiguity in the way inteectua property shoud be assigned by country, inaccuracies in the measurement of insurance and freight margins and other issues. Organizations such as the OECD are working with various countries to reduce these discrepancies, however, in the meantime efforts to buid IIOT s from each country s individua data must make consistent assumption for a countries, which oftentimes overestimate or underestimate trade fows for specific countries. The typica assumption made is that import data is more reiabe than export data since there the revenues coected from tariffs from incoming products create a strong incentive to keep tight contros. In cases of countries whose GDP is highy reiant on trade in which discrepancies between data sources are arge, this can ead to significant deviations from nationa accounts data.