An Application to Lao People s Democratic Republic and Thailand. Developing an Interregional Input Output Table for Cross-border Economies:

Transcription

1 ECONOMICS AN RESEARCH EPARTMENT ER OCCASIONAL STATISTICAL PAPER SERIES eveloping an Interregional Input Output Table for Cross-border Economies: An Application to Lao People s emocratic Republic and Thailand Benson Sim, Francisco Secretario, and Eric Suan July 2007

2

3 ER Occasional Statistical Paper Series No. 1 eveloping an Interregional Input Output Table for Cross-border Economies: An Application to Lao People's emocratic Republic and Thailand Benson Sim, Francisco Secretario, and Eric Suan July 2007 Benson Sim is a Statistician, Francisco Secretario a Technical Consultant, and Eric Suan an Assistant Economics and Statistics Analyst in the evelopment Indicators and Policy Research ivision, Economics and Research epartment, Asian evelopment Bank (AB). This paper stems from the report on Estimating An Empirical Interregional Input-Output Table for Mukdahan and Savannakhet prepared under a project carried out by Francisco Secretario for AB. The authors are thankful to the various government agencies in Lao People s emocratic Republic and Thailand for providing the data for the study. The authors are also grateful to Tun Lin, alisay Maligalig, and Fan Zhai of the Economics and Research epartment and two anonymous referees for their useful comments and constructive suggestions; and Abuzar Asra for his valuable encouragement and guidance during the inital stage of this study.

4 Asian evelopment Bank 6 AB Avenue, Mandaluyong City 1550 Metro Manila, Philippines by Asian evelopment Bank July 2007 ISSN The views expressed in this paper are those of the author(s) and do not necessarily reflect the views or policies of the Asian evelopment Bank.

5 Foreword The ER Occasional Statistical Paper Series is based on papers prepared by staff of the Economics and Research epartment, their consultants, or resource persons. The purpose of the Series is three fold. First, it deals with studies on the quality and comparability of the data compiled by the Asian evelopment Bank from national sources in its developing member countries and data prepared by international organizations. Second, it attempts to fill information gaps by deriving new statistical series from available data. Third, it presents new applications of statistical methods for use in the AB s operational work and member countries. The Series papers are prepared mainly for internal readership within the Asian evelopment Bank, but may be accessed by interested external readers.

6

7 Contents Abstract vii I. Introduction 1 II. Literature Review 1 III. The Mukdahan Savannakhet IRIO Framework 2 A. The Ideal Mukdahan Savannakhet IRIO Model 2 B. Modified IRIO Table for Mukdahan and Savannakhet 5 IV. eveloping The Mukdahan Savannakhet Modified IRIO Table 8 A. Phase 1: Estimation of Inputs, Final Local emand, and External Trade 8 B. Phase 2: Conversion of Competitive Flows to Noncompetitive Flows and Estimation of Regional Input Coefficients 14 C. Phase 3: Reconciliation and Revalidation of the Integrated IRIO Table 15 V. Main Results and Applications 15 A. Analysis of Economic Structure 15 B. Interregional Analysis and Applications 23 C. Impact Analysis 29 VI. ata Issues 35 VII. Conclusion 36 Appendix 1: List of Input Output Tables 38 Appendix 2: Correspondence between 20 Industries, 10 Major Industries, and 3 Major Sectors 39 Selected References 40

8

9 Abstract This paper constructs a modified interregional input output (IRIO) table to link the economies of the province of Mukdahan in Thailand and the neighboring Lao People's emocratic Republic province of Savannakhet using the Chenery- Moses model in conjunction with popular data reduction methods such as the simple location quotient method in the input output (IO) tables literature. A hybrid approach was used to construct the IRIO table. This involved using primary data collected from a specially conducted survey to develop the Savannakhet portion of the table and indirect methods to construct the Mukdahan portion of the table as well as the trade flows. Results showed that the value of trade of these provinces with the rest of the world was much higher than the trade between them. The results also showed that in both provinces, industries in the services sector were found to have generally higher value-added multipliers than industries in the manufacturing sector. The agriculture and forestry industry in Savannakhet and the manufacturing industries in both provinces had high backward and forward linkages. Exports to the rest of the world and consumption were found to have the highest employment multipliers in Mukdahan and Savannakhet, respectively. Mukdahan was also found to have higher net foreign exchange earnings, implying that the Thai province may be able to add greater value to its exports than the Lao PR province. Finally, the study outlines some key data and methodological issues that could bias the quality of the findings, and suggests solutions to enhance the accuracy and reliability of the results.

10

11 I. Introduction Economic interdependence between Lao People s emocratic Republic (Lao PR) and Thailand has increased substantially since the latter formally opened its border with the former in the late 1980s. The level of economic interdependency between these two countries was further boosted in 1992 when they and other countries of the Greater Mekong Subregion (GMS) embarked on a program of economic cooperation (known as the GMS Program) to promote development through closer economic linkages; and in 1997 when Lao PR became a full member of the Association of Southeast Asian Nations (ASEAN). 1 This paper measures the extent of economic interdependency between Lao PR and Thailand at the subnational level by constructing an interregional input output (IRIO) table that links the cross-border economies of Mukdahan province in Thailand with Savannakhet province in Lao PR using the Chenery-Moses model in conjunction with popular data reduction methods such as the simple location quotient method in the input output (IO) tables literature. An IRIO table provides a fairly compact, comprehensive, and detailed accounting framework to model and measure the economic interrelationships between industries located in the study regions. Similar to a single-region IO table, an IRIO table can be used to estimate the magnitude of an external shock on major macroeconomic indicators such as output, value-added, income, and employment. However, unlike its single-region counterpart, an IRIO table can also be applied to capture and assess the interregional spillover and feedback effects arising from an exogenous change in demand for the output of any one of the study regions. In other words, constructing an IRIO table will not only allow us to estimate the stimulus to production outside the study region benefiting from, say, an increase in foreign The countries of the GMS comprise Cambodia, People s Republic of China, Lao PR, Myanmar, Thailand, and Viet Nam. demand for its output, but also the resultant impact on its output arising from the production stimulus it causes in the other study regions. The paper is structured as follows. Section II provides a brief survey of the development of IO tables. The accounting framework used to develop the IRIO table for Mukdahan and Savannakhet is outlined in Section III. The methods and data used to construct the modified IRIO table for Savannakhet and Mukdahan are described in Section IV and main results are discussed in Section V. In Section VI, the major data issues that could affect the results are presented, while Section VII concludes. II. Literature Review The basic framework for IO analysis has its antecedents in the Tableau Economique developed by the French economist Francois Quesnay nearly 250 years ago. Quesnay s tableau describes the relationships between sales and purchases of the various industries of an economy. More than a hundred years later, Leon Walras adapted his model to provide a more concise theoretical formulation of an economic system including consumer purchases and economic representation of technology. However, it was not until the 20th century that Leontief (1936) greatly simplified Walras theoretical formulation by assuming the fixity of both technology and trading patterns over time to develop an IO model of the 1919 United States (US) economy to estimate the effects of the end of the First World War on national employment. The field of IO analysis was advanced further when Isard (1951) extended the work of Leontief to analyze IO tables at the subnational level by publishing a seminal paper on the theoretical structure of interregional IO tables under the assumption that the sectoral and geographical origin of each delivery can be specified. His model has been described as ideal by Richardson (1972) because identical sectors in different regions are treated as distinctly separate

12 ER Occasional Statistical Paper Series Issue No. 1 industries, so that interregional trade flows by region of origin and destination and by industry of source and purchasing sector are fully specified. However, data at this very detailed level of disaggregation are mostly not readily available. Thus, detailed and comprehensive surveys to collect data on regional purchases by sector and regional sales by sector will need to be undertaken. Unfortunately, there is usually not enough time and/or financial resources to conduct this kind of data gathering exercise. Consequently, the ideal model has rarely been implemented empirically in its entirety. This has resulted in the development of models that require considerably less data to be statistically implemented. Examples include the Chenery-Moses model, intranational model, and gravity model, all of which involve the calculation of interregional trade flows without having to trace the individual firm s nonlocal sales and purchases. 2 In addition, data reduction methods using secondary data (such as national IO coefficients and other published information) to derive local IO coefficients have also been developed for use in conjunction with these models. Examples include the use of procedures such as the simple location quotient method and commodity balance approaches to split national coefficients into local coefficients and imported inputs. Moreover, to reduce the enormous time and resources needed to conduct full-scale surveys, a hybrid or mongrel approach to constructing IRIO tables has also been utilized. This essentially involves conducting surveys to collect the necessary data for the most prominent or important sectors of the economies concerned and using various nonsurvey methods to obtain the estimates for the other sectors. 3 The development of these simplifying models and procedures has resulted in a vast amount of IO work at the interregional level. For example, in recent years, IRIO tables have been developed for European countries such as Austria (Fritz et al. 2006), Finland (Piispala 2000), Italy (Benvenuti and Paniccià 2003), and Spain (Verdura 2000). In the Asian context, various separate studies have been conducted recently to develop IRIO tables for countries such as Japan (Ishikawa and Miyagi 2004), People s Republic of China (PRC) (Okamoto et al and Okuda et al. The Chenery-Moses model has also been described as the multiregional IO model by Isard et al. (1998) as well as Miller and Blair (1985). A very good description of the models and techniques can be found in Richardson (1972). 2 Benson Sim, Francisco Secretario, and Eric Suan 2004), Philippines (Secretario 1994 and Secretario et al. 2002), and Viet Nam (Secretario et al. 2003a and 2003b). At the supra-national level, the Institute of eveloping Economies Japan External Trade Organization has been compiling international IO tables linking countries in Asia and the Pacific region since the 1970s. Currently, these tables are available for reference years 1975, 1985, 1990, 1995, and A rather unique study covering regions located in different countries was also conducted by Ichihashi (2000) to assess the economic dependency between Hiroshima (a prefecture in Japan) and Heilongjiang (a province of the PRC). To our knowledge, this study appears to be the only one that has been conducted so far to measure economic dependence between regions located in different countries in Asia. However, the study only made use of the separate IO tables for each region to assess economic dependence in terms of net exports. There was no attempt to develop a fullfledged set of IRIO tables for analysis. The paucity of studies to construct multicountry IRIO tables at the subnational level could be mainly due to the severe problems and constraints inherent in collecting and assembling consistent and reliable subnational data for the regions concerned. Thus, our study to develop an IRIO table for Mukdahan and Savannakhet constitutes a very useful contribution to the literature of IRIO tables. III. The Mukdahan Savannakhet IRIO Framework A. The Ideal Mukdahan Savannakhet IRIO Model Assuming the availability of an ideal data set, the economic flows between Mukdahan and Savannakhet can be presented in the form of the Isard-type IRIO table in Figure 1. The flows in the model are valued at producers prices (i.e., prices net of trade and transport margins, but gross of product taxes), while the reference year for the model is The model contains 20 industries, four final local demand components, and four primary inputs or value-added 2003 is used as the reference year for the model to ensure consistency with the reference year of the Savannakhet IO table developed by Asra et al. (2006). As will be seen later, the Savannakhet IO table will be heavily used in the development of the Savannakhet portion of the modified IRIO table in our study.

13 eveloping an Interregional Input Output Table for Cross-border Economies: An Application to Lao People's emocratic Republic and Thailand components in each province. The model also contains a third region rest of the world (ROW) which groups, for convenience, all areas (rest of Thailand, rest of Lao PR, and all foreign territories) outside the two provinces. The notations used in Figure 1 are defined as follows: X is a matrix, where each element is a flow from one of the 20 industries to another of the 20 industries in Mukdahan. X S is a matrix containing analogous interindustry flows, but from Mukdahan to Savannakhet. Y is a column vector of 20 elements, where each element represents a flow from one of the 20 industries in Mukdahan to final local demand (i.e., private consumption expenditure, government consumption expenditure, gross fixed capital formation, and change in inventories) in the same province. Y S is a column vector of 20 elements containing analogous final demand flows, but from Mukdahan to Savannakhet. E is a column vector of 20 elements, where each element is an export of Mukdahan by industry to the ROW. X is a column vector of 20 elements, with each element representing the gross output of each of the 20 industries in Mukdahan. X S is a matrix, where each element is a flow from one of the 20 industries in Savannakhet to another of the 20 industries in Mukdahan. X SS is a matrix containing analogous interindustry flows, but within Savannakhet. Y S is a column vector of 20 elements, where each element represents a flow from one of the 20 industries in Savannakhet to final local demand (i.e., private consumption expenditure, government consumption expenditure, gross fixed capital formation, and change in inventories) in Mukdahan. Y SS is a column vector of 20 elements containing analogous final demand flows, but within Savannakhet. E S is a column vector of 20 elements, where each element is an export of Savannakhet by industry to the ROW. Figure 1. Accounting Framework for 2003 Mukdahan Savannakhet Interregional IO Table, Noncompetitive Type From To Intermediate emand Final emand MK SVK MK SVK ROW 1 j 20 1 j 20 E M Total Gross Output i n t e r m e d i a t e I n p u t s M K S V K R O W 1 : i : 20 1 : i : 20 1 : i : 20 X X S Y Y S E 0 X X S X SS Y S Y SS E S 0 X S I M W I M WS F M W F M WS 0 M T 0 Gross Value Added V ' V S' T T Total Gross Input X ' X S' MK means Mukdahan; SVK means Savannakhet; ROW means rest of world; E means exports to ROW; M means imports from ROW. Source: eveloped by the authors. July 2007

14 ER Occasional Statistical Paper Series Issue No. 1 X S is a column vector of 20 elements, with each element representing the gross output of each of the 20 industries in Savannakhet. I M W is a matrix, where each element is a flow from one of the 20 industries from the ROW to another of the 20 industries in Mukdahan. I M WS is a matrix containing analogous flows from the ROW to Savannakhet. F M W is a column vector of 20 elements, where each element is a flow from one of the 20 industries from the ROW to the final local demand in Mukdahan. F M WS is a column vector containing analogous flows from the ROW to Savannakhet. M T is a column vector of 20 elements, with each element representing the total imports (with a negative sign) by industry of Mukdahan and Savannakhet. V ' is a row vector of 20 elements, with each element defining the primary inputs (i.e., compensation of employees, indirect taxes less subsidies, depreciation, and gross operating surplus) by industry flowing into Mukdahan. V S' is a row vector of 20 elements containing analogous flows into Savannakhet. T T is the total value of tariff or import duties and taxes generated in both Mukdahan and Savannakhet. X ' is a row vector of 20 elements defining total input by industry for Mukdahan. X S' is a row vector of 20 elements defining total input by industry for Savannakhet. The outlined IRIO model is of the noncompetitive, open, and static variety. It is noncompetitive because it makes an explicit distinction between regionally produced and nonregionally produced goods and services. Such a distinction provides a better reflection of the use of local production technology and inputs in the production of output in each province. The openness of the model is derived from the fact that economic activities are split into the intermediate and final demand categories. The transactions in the former category can be explained by the model, while the latter category contains exogenous transactions that must be initially known or given. The static nature of the model is a consequence of the absence of a time dimension from it. A system of IRIO tables is balanced, implying that the supply and demand sides are equal. Using Figure 1, this equality can be translated into the following accounting identities: xi = i = 1 j = 1 x j (i.e., total output of industry i summed over all the 20 industries in Mukdahan is equal to total input of industry j summed over the same industries), and S xi = i = 1 j = 1 x S j (i.e., total output of industry i summed over all the 20 industries in Savannakhet is equal to total input of industry j summed over the same industries). Figure 1 can also be used to form the following balancing equations in matrix form: X = X i + X S i + Y + Y S + E (1) X S = X S i + X SS i + Y S + Y SS + E S (2) In both equations, i represents a column vector of appropriate ones. The first term on the right hand side of equation (1) represents interindustry sales in Mukdahan; the second term denotes the interregional trade flows from Mukdahan to Savannakhet; the third and fourth terms represent the sales of the output of Mukdahan to final local demand in Mukdahan and Savannakhet, respectively; while the last term represents the exports of Mukdahan to the ROW. An analogous explanation applies to equation (2). Using Leontief s assumption of linearity or firstorder homogeneity in the production functions, we can define the following regional input coefficients in matrix form: A A S = X = X S 1 Xˆ (3) 1 S Xˆ (4) Benson Sim, Francisco Secretario, and Eric Suan

15 eveloping an Interregional Input Output Table for Cross-border Economies: An Application to Lao People's emocratic Republic and Thailand A A S SS = X = X S SS ˆ X 1 (5) 1 S Xˆ (6) Equations (3) and (6) represent the matrices of intraregional direct input coefficients, while equations (4) and (5) stand for the matrices of interregional trade coefficients. Substituting these structural equations into equations (1) and (2), we have: X = A X + A S X S + Y + Y S + E (7) X S = A S X + A SS X S + Y S + Y SS + E S (8) X X S Combining equations (7) and (8), we have: A = A S A A S SS X X S Y + Y S (9) where Y = Y + Y S + E and Y = Y S + Y SS + E S. In equation (9), Y and Y S are column vectors of 20 elements each, with each element representing the sum of flows from one of the 20 industries in Mukdahan and Savannakhet, respectively, to final demand in Mukdahan and Savannakhet and the ROW. Simplifying equation (9), we have: 1 X S I 0 A A Y = S X 0 I S SS S A A Y (10) Equation (10) can be further simplified and shown its generalized form as: X = BY (11) where X is the matrix of regional outputs, X S X ; Y is the matrix of final demand, Y S ; and B is the Y interregional Leontief inverse matrix, I 0 A 0 I A S A A S SS 1. In order to measure the spillover and feedback effects caused by interregional trading, equation (9) can first be expanded to obtain the following simultaneous equations: ( ) ( + ) 1 ( ) ( + ) 1 S S X = I A A X Y S SS S S X = I A A X Y (12) (13) Equations (12) and (13) can then be further expanded and rearranged to obtain the following matrices: X X S F 0 = 0 F S I S S S I S where M = I A, M I A S S S S S S = M A, S = M A F = I S S M 0 0 M ( ) = ( ) 1 1 S SS ( ) S S S Y Y ( ) 1 1 S S S, F = I S S S (14) The unknowns M and M S, S S and S S, and F and F S account for the intraregional linkages and interregional spillover and feedback effects, respectively. B. Modified IRIO Table for Mukdahan and Savannakhet Unfortunately, the data needed to implement the ideal framework outlined in Figure 1, in particular, the interregional interindustry flows, are mostly not readily available. This means that equation (11) cannot be readily used to estimate the Leontief inverse. Instead, we will rely on the Chenery-Moses model independently developed by Chenery (1956) and Moses (1955) to develop a modified IRIO table to link the economies of Mukdahan and Savannakhet in our study. 5 The starting point for the development of the modified IRIO table is the intraregional IO framework in Figure 2. The notations used in Figure 2 are defined as follows: 5 An early comparison of the Isard and Chenery-Moses models can be found in Hartwick (1971). The Chenery-Moses model overcomes the severe data needs of the Isard model, which requires data on the shipments of goods between sectors and between regions, by requiring only data on the flow of goods by sector from the origin to the destination region. July 2007

16 ER Occasional Statistical Paper Series Issue No. 1 X. is a matrix, where each element is a flow from one of the 20 industries (regardless of origin) to another of the 20 industries in Mukdahan. Y. is a column vector of 20 elements, where each element represents a flow from one of the 20 industries (regardless of origin) to final local demand (i.e., private consumption expenditure, government consumption expenditure, gross fixed capital formation, and change in inventories) in Mukdahan. E S is a column vector of 20 elements, where each element is an export of Mukdahan by industry to Savannakhet. E W is a column vector of 20 elements containing analogous flows to the ROW. X is a column vector of 20 elements, with each element representing the gross output of each of the 20 industries in Mukdahan. X.s is a matrix, where each element is a flow from one of the 20 industries (regardless of origin) to another of the 20 industries in Savannakhet. Y.s is a column vector of 20 elements, where each element represents a flow from one of the 20 industries (regardless of origin) to final local demand (i.e., private consumption expenditure, government consumption expenditure, gross fixed capital formation, and change in inventories) in Savannakhet. E S is a column vector of 20 elements, where each element is an export of Savannakhet by industry to Mukdahan. E SW is a column vector of 20 elements containing analogous flows to the ROW. X S is a column vector of 20 elements, with each element representing the gross output of each of the 20 industries in Savannakhet. M S is a column vector of 20 elements defining imports (with a negative sign) by industry from Savannakhet to Mukdahan. M W is a column vector of 20 elements containing analogous flows from the ROW into Mukdahan. M S is a column vector of 20 elements defining imports (with a negative sign) by industry from Mukdahan to Savannakhet. M WS is a column vector of 20 elements containing analogous flows from the ROW into Savannakhet. V ' is a row vector of 20 elements defining the primary inputs (i.e., compensation of employees, indirect taxes less subsidies, depreciation, and gross operating surplus) by industry flowing into Mukdahan. V S' is a row vector of 20 elements containing analogous flows into Savannakhet. T is the total value of import duties and taxes generated in Mukdahan. T S is the total value of import duties and taxes generated in Savannakhet. X ' is a row vector of 20 elements, with each element defining total input by industry for Mukdahan. X S' is a row vector of 20 elements, with each element defining total input by industry for Savannakhet. Intermediate emand Figure 2. Accounting Framework for 2003 Intraregional IO Table Mukdahan Final emand Local Exports Imports Intermediate emand Savannakhet Final emand Exports Imports 1 2 : i : G V A 1... j 20 SVK ROW SVK ROW TGO 1... j 20 Local MK ROW MK ROW X. Y. E S E W M S M W X X. s Y. s E S E SW M S M WS X S V ' T V S' T S TGI X ' X S' TGO means total gross output; TGI means total gross input; GVA means gross value added; ROW means rest of world; MK means Mukdahan; SVK means Savannakhet. Source: eveloped by the authors. TGO Benson Sim, Francisco Secretario, and Eric Suan

17 eveloping an Interregional Input Output Table for Cross-border Economies: An Application to Lao People's emocratic Republic and Thailand As can be observed from Figure 2, the basic idea of the modified IRIO table is that intraregional IO tables that make use of transactions data that are more readily available or derivable are developed first. For instance, information on the amount of input from industry i to be used by industry j in, say, Mukdahan (i.e.,. x ij ) is likely to be more easily obtainable than information on the amount of input from industry i from Savannakhet used by industry j in Mukdahan (i.e., S x ij ). Using the data in Figure 2, we can define the following regional technical coefficients in matrix form:.. A = X ( X $ -1 ) (15).S.S S A = X ( ) X $ 1 (16) The interconnections among the regions in the modified IRIO table framework are captured in the form of an interregional trade coefficients matrix in a very different manner from the ideal IRIO framework. In the modified IRIO table framework, trade flows are estimated by product in order to exploit the types of data likely to be available. The elements of the interregional trade coefficient matrix, denoted by ci RS (R, S = Mukdahan and Savannakhet), show the proportion of all of product i used in region S that comes from each region R. 6 Interregional trade coefficients are obtained from the flow from R to S divided by the total inflow into the latter and are defined as: c RS i = RS ti. n RS ti. R where (n = Mukdahan, Savannakhet) (17) RS In equation (17), t i. denotes the amount of product i that is shipped from region R to region S, irrespective of the industry of destination in the latter. These flows will include shipments to the producing sectors as well as final demand in the destination region, implying that the proportion of the supply of goods and services in each sector in the destination region that comes from the shipping region is the same. The interregional trade coefficients can be rearranged in the form of a column vector for each origin-destination pair of regions, which can also be transformed into a diagonal matrix as follows: (18) Using the matrices for the regional technical coefficients and interregional trade coefficients, we can estimate the regional input coefficients in equation (9) as follows: A A A A C C$ S S SS SS c. =.. cn C$. C $ A (20) S.S C $ A (21) S. C $ A (22) SS.S, C S c. =.. cn S 1 1 S c K 0 = M O M 0 L cn SS c K 0 = M O M 0 L cn, C C$, c. =.. cn, C C $ A (23) As the trade coefficient information used in our modified IRIO framework is of an origin-todestination format, it also includes shipments not only to industries, but final demand as well. Consequently, in the modified IRIO framework, a final demand term, such as Y., is taken to be satisfied from a pool of goods, i.e., partly by purchases from industries in the region and partly by purchases from industries outside the region. SS 1 1 S SS C$, 1 1 S S c. =.. cn SS S 1 1 SS S c K 0 = M O M, S 0 L cn S S c K 0 = M O M ( n = 20) S 0 L cn (19) 6 As there are 20 products in our model, there will be 20 such matrices in all. July

18 ER Occasional Statistical Paper Series Issue No. 1 Given this view, the fundamental equations in our modified IRIO table can be written in matrix form as X = C$ C$ S S X C $ C$ E E W SW M M W WS S SS which can be reduced to. A 0 X + C$ C$. S S A X 0 S C $ C$ S SS Y Y.. S + (24) X I 0 A. S A.S = S X 0 I S A. SS C$ C$ A.S C $ C$ (25) S C$ C$. W Y E W M + S SS. S SW WS C $ C$ Y E M To obtain the interregional and spillover effects, equation (24) can first be expanded to obtain the following simultaneous equations: 1 X I A S S S A X Y S = Y S E M C$. C$. + C$ + C$ + (26) 1 S SS S S X I A A X SS Y S S = Y E S M S C$. C$. + C$ + C$ + (27) Equations (26) and (27) can be further expanded and rearranged to obtain the following matrices: X X S where, F% = 0 C$ C$ S C $ C$ S SS 1 0 S I S% M% 0 S S S F% S % I 0 M% Y Y.. S + E E W SW M M W WS 1 1. S SS.S M% = I C$ A M I C$ A, % = S% = M% C$ S S.S A, S% = M% C$ S S S. ( ) = ( ) F% = I S% S%, F% I S% S% S S 1 S S S 1 A (28) Equation (25) and, by extension, equation (28) are the equations that can be estimated using more readily available or derivable data. In equation (28), the unknowns M % and S M %, S S % and S S %, and F% and F% S account for the intraregional linkages and interregional spillover and feedback effects, respectively, in the modified IRIO framework. IV. eveloping the Mukdahan Savannakhet Modified IRIO Table In many countries, IO tables are typically developed using the cost or expenditure method. This means that the tables are formed by merging the cost or input structures of industries as well as the consumption patterns of the components of final demand. Such a column-wise method of constructing IO tables is used because of the dearth of data that trace the flow of outputs from producer to consumer. Likewise, the modified IRIO table in our study was developed using the cost method. The construction of the modified IRIO table was generally carried out using the 1968 System of National Accounts guidelines in three major phases: Phase 1: Estimation of Inputs, Final Local emand, and External Trade Phase 2: Conversion of Competitive Flows to Noncompetitive Flows and Estimation of Regional Input Coefficients Phase 3: Reconciliation and Revalidation of the Integrated IRIO table As the source data were valued in the respective currencies of Thailand and Lao PR, they were converted into US dollars using the nominal average exchange rates for 2003 during the process of developing the modified IRIO table. 7 A. Phase 1: Estimation of Inputs, Final Local emand, and External Trade This phase involves the development of separate intraprovincial competitive IO tables (as shown in Figure 2) for Mukdahan and Savannakhet by deriving estimates of inputs, final local demand, and external trade. The bulk of the work in this phase was devoted to the construction of the intra-mukdahan IO table as a survey-based IO table for Savannakhet has already 7 The exchange rates used were US$1 = Baht and US$1 = 9,614 Kips. Benson Sim, Francisco Secretario, and Eric Suan

19 eveloping an Interregional Input Output Table for Cross-border Economies: An Application to Lao People's emocratic Republic and Thailand been developed in a separate unpublished study by Asra, Secretario, and Suan (2006) Compilation Methodology and ata Sources The more appropriate method of estimating the inputs is to conduct specialized IO surveys of enterprises and use public administrative/regulatory reports to obtain primary data on production costs. This method was used to construct the IO table for Savannakhet. However, due to time and budgetary constraints, no surveys were conducted in the case of Mukdahan. Instead, a nonsurvey method was used to construct the intra-mukdahan IO table using solely the following secondary statistical information provided by various government agencies such as the Thai National Statistical Office and Thai National Economic and Social evelopment Board. 9 (i) National Input Output Table for Thailand for Year 2000 (ii) Gross Provincial Product (GPP) by Sector for Mukdahan for year 2003 (iii) 2003 Survey of Establishments for the Industry and Services Sectors for the Northeastern Region (iv) 2003 Survey of Household Income and Expenditure for the Northeastern Region (v) Exports/Imports ata from Mukdahan Customs Office (vi) Exports/Imports ata from Savannakhet Customs Office (vii) Thailand Foreign Trade Statistics (viii) Other relevant data from administrative reports such as cost and returns data for agriculture, government budget, etc. 2. The Intra-Mukdahan IO Table Given the above statistical information, the competitive intra-mukdahan IO table was developed using the following steps: 8 More details on the construction of this table can be obtained from the study. 9 Since a survey-based approach was used to develop the IO table for Savannakhet, this means that a hybrid approach involving the use of surveys and nonsurvey methods was used to develop our modified IRIO table. Step 1. Estimation of Inputs (i.e., X. and V ) Step 2. Estimation of Final Local emand Vectors (i.e., Y. ); Step 3. Estimation of External Trade Vectors Step 4. Balancing the Intra-Mukdahan IO Table (i) Step 1: Estimation of Inputs (X. and V ) Inputs for the 20 industries in Mukdahan were estimated by assuming that their input structure can be proxied by the corresponding IO coefficients from the Thailand IO table. As the national IO table is split into 180 industries, they were merged and combined to conform to the sectoral breakdown of the Mukdahan IO table before the national IO coefficients were applied. The national IO coefficients for 14 of the 20 industries as outlined in ata Set A were further modified using data from the sources listed in order to generate a set of input ratios that were more reflective of Mukdahan s production cost structure. Intermediate and value-added ratios for the crops and livestock, poultry, and fishery industries were derived from costs and returns survey data for crops cultivation and livestock raising provided by the Bureau of Agricultural Economics, Ministry of Agriculture. As Mukdahan is one of the provinces in the northeastern region of Thailand, the 2003 Survey of Establishments for the Industry and Services Sectors for the Northeastern Region (SE) was used to modify the national coefficients for the manufacturing (05-11) as well as the private services industries (16-18 and 20) on the basis that the production cost structure for the province could be proxied by the regional one. Where necessary, the industries in the survey were first grouped to fit into the sectoral breakdown of the Mukdahan IO table before the national coefficients were modified. In the case of public administration, input indicators derived from the government budget report of the Mukdahan provincial government were used. For the remaining six industries listed below, the corresponding national IO coefficients were adopted as proxies due to the unavailability of appropriate subnational indicators: July 2007

20 ER Occasional Statistical Paper Series Issue No. 1 Industry 03: Forestry and logging Industry 04: Mining and quarrying Industry 12: Electricity and water supply Industry 13: Construction Industry 14: Transportation Industry 15: Post and telecommunications ata Set A. ata Sources for Estimating Mukdahan s Input Structure Industry ata Source 01 Crops Bureau of Agricultural Economics, Ministry of Agriculture 02 Livestock and poultry, fishery Bureau of Agricultural Economics, Ministry of Agriculture 05 Food, beverage, and tobacco 2003 Survey of Establishments for the Industry and Services Sectors for the Northeastern Region 06 Textiles, garments, and leather products 2003 Survey of Establishments for the Industry and Services Sectors for the Northeastern Region 07 Wood and paper products; printing/ publishing 2003 Survey of Establishments for the Industry and Services Sectors for the Northeastern Region 08 Chemical products; petroleum 2003 Survey of Establishments for the Industry and Services Sectors for the Northeastern Region 09 Nonmetallic mineral products 2003 Survey of Establishments for the Industry and Services Sectors for the Northeastern Region 10 Metal products, machinery, equipment, and spare parts 2003 Survey of Establishments for the Industry and Services Sectors for the Northeastern Region 11 Other manufactured goods 2003 Survey of Establishments for the Industry and Services Sectors for the Northeastern Region 16 Wholesale and retail trade 2003 Survey of Establishments for the Industry and Services Sectors for the Northeastern Region 17 Banking, insurance, business services 2003 Survey of Establishments for the Industry and Services Sectors for the Northeastern Region 18 Real estate and ownership of dwellings 2003 Survey of Establishments for the Industry and Services Sectors for the Northeastern Region 19 Public administration Government Budget, Mukdahan Provincial Government 20 Personal, social, and community services 2003 Survey of Establishments for the Industry and Services Sectors for the Northeastern Region Source: Compiled by the authors. a. Modifying the National IO Coefficients The input structures for the 14 industries listed in ata Set A were estimated as follows. Intermediate input ratios were obtained from the following formula: a. ij TIIR = TIIR SE j N j N aij (29) where a. ij is the estimated coefficient for Mukdahan s consumption of product i by industry j; TIIR SE j is the ratio of the total intermediate input of industry j to its corresponding revenue (or output), as derived from the SE; TIIR N j is the ratio of the total intermediate input of industry j to its corresponding output, as recorded in the national IO table; and a N ij is the coefficient for the nation s consumption of product i by industry j. Primary input (value-added) ratios were estimated using the following relationship: b pj SE g j N = var b N pj g var (30) j where b pj is the ratio of Mukdahan s value added, p, for industry j to its corresponding gross output, x j ; SE g var j is the ratio of gross value added (GVA) of industry j to its total revenue (or output), as derived from the SE; N g var is the ratio of the GVA of the nation s industry b N pj j j to its corresponding gross output, and is the national GVA coefficient of industry j for primary input p. This modified set of input coefficients, when multiplied by an estimated vector of gross outputs, gives the monetary value of input transactions in Mukdahan. Mathematically, the sum of the intermediate and primary input coefficients should be equal to one. In cases where the sum of these coefficients was not equal to unity, proportional adjustments were made to ensure that they added up to one. 10 Benson Sim, Francisco Secretario, and Eric Suan

21 eveloping an Interregional Input Output Table for Cross-border Economies: An Application to Lao People's emocratic Republic and Thailand b. Estimation of Gross Output Given estimates of GPP or gross value added by industry for Mukdahan for the year 2003 and the input coefficient matrix, gross output was estimated using the following formula: x j GVAj = GVAR j (31) where x j is the estimated gross output of industry j in Mukdahan for j = 1,2,...,20 GVA j is the given gross value added (or total primary input) of industry j in Mukdahan; and GVAR j is the gross value added ratio of industry j in Mukdahan. c. Calculating the Input Transactions The estimates of inputs were then calculated as follows. Intermediate input transactions were estimated using the following relationship: X = A X$ (32).. where X. is the matrix of intermediate input transactions of the competitive-import type for Mukdahan; A. is the matrix of regional technical coefficients for Mukdahan; and X ˆ is the diagonal matrix of sectoral gross outputs for Mukdahan. Estimates of primary input transactions were obtained using the following relationship: V B X ' (33) = ( ˆ ) i where V is the primary input or value added column vector for Mukdahan; B is the matrix of primary input coefficients for Mukdahan; X ˆ is the diagonal matrix of sectoral gross outputs for Mukdahan; and i is a column vector of appropriate ones. (ii) Step 2: Estimation of Final Local emand Vectors The Mukdahan IO model classifies exogenous final demand into four components of final local (domestic) demand and four categories of external trade. The estimation of the final local demand components was done independently using the data sources in ata Set B. a. PCE Vector Total PCE was first estimated by multiplying the average household expenditure for Mukdahan, as reported in the 2003 Survey of Household Income and Expenditure for the Northeastern Region, by the estimated number of households in the province for the same year. The PCE vector was then obtained by multiplying the estimated total PCE for Mukdahan by a vector of expenditure patterns derived from the same household survey. As the PCE data are disaggregated in terms of commodities, they were allocated to the appropriate industry in the Mukdahan IO table before the vector of expenditure patterns was obtained. Moreover, as the resulting PCE vector was initially valued at purchaser s price, it was converted to producer s price by netting out the trade and transport margins on PCE on goods at purchaser s price. Trade and transport margin ratios derived from the national IO table were used in the netting out process. On the other hand, no margins were applied to household expenditure on services. ata Set B. ata Sources for Estimating Mukdahan s Final Local emand Final Local emand ata Source B-1. Private consumption expenditure (PCE) 2003 Survey of Household Income and Expenditure for the Northeastern Region B-2. Government consumption expenditure (GCE) Mukdahan Provincial Government Budget Report, 2003 B-3. Gross fixed capital formation (GFCF) 2003 Survey of Establishments for the Industry and Services sectors for the Northeastern Region B-4. Change in inventories (CI) 2003 Survey of Establishments for the Industry and Services sectors for the Northeastern Region Source: Compiled by the authors. July

22 ER Occasional Statistical Paper Series Issue No. 1 b. GCE Vector No entries were recorded in the GCE vector, except for the cell related to public administration. In this cell, the total gross output that is consistent with national accounts concept was recorded. c. GFCF and CI Vectors The GFCF and CI vectors were approximated based on regional ratios derived from the 2003 Survey of Establishments for the Industry and Services Sectors for the Northeastern Region. For example, capital expenditure-to-output ratios by industry and by type of durable equipment as estimated from the survey were applied to the corresponding Mukdahan output estimates to obtain an estimate of expenditure for durable equipment. To complete the GFCF vector, the total value of output of construction less expenditure by industries and households for housing repair and maintenance was recorded in this GFCF vector. Similarly, the estimates of CI were obtained by applying the ratio of change in stocks to total output or revenue for each industry from the survey to the corresponding total output. As these estimates are perceived to be weak, the CI estimates were later accordingly adjusted during the reconciliation and revalidation phase. Merging the results of Steps 1 and 2 with a residual vector of net exports resulted in the following equation:.. X = X + Y + E M ( ) (34) The next step was to split the residual net exports vector into its four components as follows: ( ) = + ( ) ( + ) (35) S W S W E M E E M M (iii) Step 3: Estimation of External Trade Vectors The following data on Mukdahan's external trade were available (see ata Set C): ata Set C. ata Sources for Estimating Mukdahan s External Trade External Trade ata Source C-1. Mukdahan exports to Available by product from Mukdahan Savannakhet, E S Customs and Immigration Office C-2. Mukdahan imports Available by product from Mukdahan from Savannakhet, M S Customs and Immigration Office C-3. Mukdahan exports to Available from Mukdahan Customs ROW, E WS and Immigration Office C-4. Mukdahan imports Available from Mukdahan Customs from ROW, M W and Immigration Office Source: Compiled by the authors. Breaking up the net exports vector into its four components required an assessment of the raw data collected from the Mukdahan Customs Office as external trade data compiled by line agencies do not usually conform to the producer-to-user concept in IO accounting. The need to distinguish direct users from traders or middlemen as well as foreign trade transactions such as in-transit/cross hauling deliveries should also be taken into account and rationalized before the development of the external trade vectors. An assessment of the four data sets found that data sets C-1 and C-2 were useful for measuring the external trade between Mukdahan and Savannakhet. A better understanding of the nature of these two datasets was obtained after a series of interviews and discussions with trade experts from Mukdahan s Customs Office. In contrast, data sets C 3 and C-4 (Mukdahan trade data with ROW) were found to be not so useful as the product classification was too aggregated and information on the origin and destination of the products by province was too vague. 10 Making data sets C-3 and C-4 useful would require time-consuming and costly special surveys to be conducted or special tabulations of export/import customs manifests to determine the origin-destination of trade flows at the subnational level. ue to the lack of budgetary resources, the missing data were estimated using the simple location quotient (SLQ) method. 10 This is understandable since in many countries, foreign trade statistics are normally compiled at the national level only. 12 Benson Sim, Francisco Secretario, and Eric Suan

23 eveloping an Interregional Input Output Table for Cross-border Economies: An Application to Lao People's emocratic Republic and Thailand a. SLQ Method The first step in splitting the net exports vector was to calculate by the SLQ method Mukdahan s total imports, whether sourced from Savannakhet or from ROW. The SLQ method assumes that the needs of region R for output i in each industry relative to the needs for output i in each of these industries nationally are the same as the ratio of the total regional to the total national output. In equation form, the SLQ is calculated as: SLQ R i R Qi T = N Qi T R N (36) where SLQ i R is the simple location quotient of product i in region R; Q i R is a measure of economic activity of industry i in region R; Q i N is a measure of economic activity of industry i in the nation; T R is a measure of aggregate economic activity in region R; and T N is a measure of aggregate economic activity in the nation. An SLQ is a measure of a region s self-sufficiency in production. From equation (36), it can be inferred that if SLQi R is less than 1, region R is presumed to import some of the output of product i from elsewhere. Conversely, if SLQi R is equal to or greater than 1, region R is viewed as self-sufficient with respect to its output i. The SLQ method was used to quantify the total imports of Mukdahan by product, regardless of the geographic source. SLQs were estimated using sectoral GVA as measures of economic activity since such data are available at the provincial and national levels. b. Estimating the Import Vectors By the SLQ approach, the import content of the intersectoral transactions in the IO table of the competitive type was calculated as: ( ) (37) tm = 1 SLQ x ij i ij where tm ij is the total import content value of product i consumed by industry j, (1 SLQ i ) is the import content ratio of product i, and x ij is the total value of product i consumed by column industry j. If SLQ i 1.0, then tm ij = 0. Using equation (37), a total imports table for Mukdahan was constructed, the row sums of which represented an initial estimate for the column vector of total imports, M T, which in the case of Mukdahan s economy, is equal to Mukdahan s imports from Savannakhet plus imports from the ROW, i.e., M T = M S + M W. Given the available data for column vector M S, the unknown ROW imports vector, M W, was then residually calculated as: M W = M T M S (38) Solving equation (38), the remaining unknown vector, E W, was residually obtained from equation (35) as: E W = (E M) E S + (M S + M W ) (39) In general, a similar approach was also used to split Savannakhet s total exports and total imports into their respective destinations and origins. c. Reconciling and Revalidating the External Trade Vectors The reconciliation process took into consideration the following accounting identities in a bi-region IRIO framework: (i) Mukdahan s outflows to Savannakhet = Savannakhet s inflows from Mukdahan, i.e., E S = M S (ii) Mukdahan s inflows from Savannakhet = Savannakhet s outflows to Mukdahan, i.e., M S = E S Moreover, overall intersectoral adjustments had to be made to correct counterintuitive results such as negative entries in the residual exports vector as July