Regional Input-Output Tables of Czech Republic reviewing the CHARM method. Marek Radvanský - Ivan Lichner 1

Transcription

1 Regional Input-Output Tables of Czech Republic reviewing the CHARM method Abstract 2 Marek Radvanský - Ivan Lichner 1 May 2018 The main scope of the paper is to review the construction method of (Multi-)Regional input-output tables in Czech conditions at the NUTS 3 regional level. We compare the estimation based on modified CHARM method of estimation of regional IO tables applied by authors with mixed survey and non-survey applied by Department of Statistics at the University of Economics in Prague (Sixta and Vltavská, 2016) based on the reference year The paper shows that despite the lack of official published RIOTs by state statistical offices, the existing methodology is a plausible option to create more or less precise RIOTs suitable for further analytical research. We show that the main concerns still rely on the estimation of intra-regional trade flows. Introduction Regional IO tables are scarcely produced by national statistical offices mainly due to regional data limitations and data protection. Exemptions could be found in case of sub-national level within bigger countries. Nevertheless, in recent years we can observe an increased focus on the analytical creation of regional input-output tables (RIOTs) by researchers mainly due to rising demand for regional analyses. From the analytical point, two main methods can be utilized for their construction: surveybased and non-survey method. Survey-based method of constructing RIOTs is prohibitively expensive (Tὄbben, Kronenberg (2015)) and thus was utilised only in several countries including Denmark, Netherlands, Italy, Finland and Canada. In Czech conditions, there was recently estimated RIOTs using mixed method (Sixta, Vltavská, 2016), which were utilized to a comparison of estimates produced with the application of the non-survey method on the basis of available national and regional. Despite the recent methodological development of non-survey methods (e.g. Bonfoglio and Chelli, 2008; Flegg and Tohmo, 2013; Lehtonen and Tykkylainen, 2014), the biggest problem in the estimating RIOTs remains in the importance of cross-hauling, thus estimation of intra- and interregional trade flows. There are two main non-survey methods of dealing with this problem. First is dealing with this issue by using FLQ - Flegg location coefficients (Flegg and Weber,1997; Flegg and Tohmo, 2013, Kowalewski, 2015), and the methodological extension named CHARM - Cross-Hauling Adjusted Regionalization Method (Kronenberg, 2009, Kronenberg and Tobben, 2011; Flegg et al., 2015). Both methods are used to estimate trade flows between each particular region to other regions within an analysed country (internal flows) and trade abroad. For more detailed analyses, the iterative proportional fitting (RAS method) to estimate bilateral flows between all pair of regions is typically applied. In this paper, we utilize estimation by CHARM non-survey method for construction of regional input tables (RIOTs) in the Czech Republic (Kronenberg and Tobben, 2011 and Tobben and Kronenberg, 2015) with several methodological adjustments. This method utilizes information from publicly 1 Institute of Economic Research, Slovak Academy of Sciences, Šancová 56, Bratislava, Slovakia, marek.radvansky@savba.sk, ivan.lichner@savba.sk 2 The research was supported by the Slovak Research and Development Agency under the contract No. APVV

2 available data sources on regions and enables researchers to construct regional input-tables at relatively low costs. Those data include a national input-output table, national supply and use tables, regional labour market data, regional national accounts data and foreign trade data. The structure of available information is also limiting the number of considered sectors (to 11 on first digit NACE2 sectors). For our application we choose the Czech Republic, because it is possible to compare obtained results with results achieved in the research of Vltavska, Sixta (2017) 3 in which researchers had access to more detailed statistical data provided within close cooperation with Czech statistical office. In detail, production coefficients of each region are analysed and cross-compared with their national counterparts. Due to the comparability of results, we consider all IO tables with base the year Modelling approach and Methodology The construction process of regional input-output tables in our research used for further analysis based on the CHARM method. Applied methodology was in detail described in Kronenberg and Tobben (2011) and Tobben and Kronenberg (2015). In this paper Multi-regional input-output tables (MRIOTs) for 14 Czech (NUTS 3) level regions were constructed. For the creation of MRIOTs, national symmetric input-output table by CPA (product x product), supply table (product x activity (NACE)) and Regional National Accounts data for the year 2013 were utilized. As a result, multi-regional input-output table of type E was created. Table 1: Type E input-output table Source: Tobben Kronenberg (2015) 3 Respective tables are published at

3 Initially, due to limited structure of regional data provided by the Czech Statistical Office, aggregation of detailed national data w to a more compressed structure for the purposes of creation of MRIOTs for the Czech Republic with 11 products took place. Underlying datasets comprised from 88 economic sectors by product in case of supply table and 82 sectors in case of a symmetric national input-output table, respectively. The final sectoral structure of produced tables is as follow: A B+D+E C F G+H+I J K L M+N O+P+Q R+S+T+U Products and services of : Agriculture Mining, quarrying and energy Manufacturing Construction Storage, transport and accommodation Telecommunications and IT Financial services Real Estate Other market services Public services Other services Table 2: Aggregation of the production sectors Source: Authors In the following part, we will closely follow the approach of (Kronenberg and Tobben, 2011) and highlight the main differences in the applied assumptions. In the first step of creation of the multiregional input-output tables for Czech Republic utilizing the - Cross-Hauling Adjusted Regionalization Method, national data on total output by given sector from supply table were transformed in to the regional output by sector on the basis of regional compensations of employees by sector (as its share on national level) (1). These data represent a most reliable approximation of total factor productivity available on a regional level. The assumption that compensations represent a reasonable approximation of the productivity of a given sector in all regions was applied. r r i,j = w j r n n r w i,j j where r r i,j represent output othe f product i by sector j in region r, w r j is amount of compensations paid to employees in the sector j in the region r, w n j represents amount of compensations paid in sector j on national level and a a r n i,j stands for total national output of product i by sector j (from national supply table). In the specific cases of sectors providing financial and real estate services in three regions (Central Bohemian Region, Ústí nad Labem Region and Vysočina Region) application of this approach was slightly modified due to resulting negative intermediate inputs 4. As a result, we can enumerate the estimated regional output by product denoted as x i as a total procuction of product i in whole economy. (1), Products (CPA) Industries (NACE) Total output j by products (CPA) 1 r 1,1 r 1,2... r 1,j x 1 2 r 1, x Coefficents representing share of regional wages to national weges were rescaled to values of regional value added on national value added for given sector in a given region.

4 i r i, r i,j x i Total output r 1 r 2 r j Table 3: The structure of the supply table Source: authors Contrary to (Kronenberg and Tobben, 2011), estimated regional outputs x i already includes the value of the output produced for own final use (item P.12 by ESA methodology), thus there is no need for additional adjustments and we can transpose the estimated values of output from regionalized supply tables directly to regional IO tables. The relatively similar approach is applied in case of gross value added (GVA) estimation. In this step, GVA data published on a regional level by NACE classification are adjusted to aggregate values from the national input-output table by CPA classification (2). r,nace v j r,cpa n = v v j j n,nace (2), v j where v jr,cpa is approximatethe d value of GVA for sector j in region r, v j n represents national value of the GVA of sector j by product, v j r,nace is value of GVA of sector j in regthe ion r and v j n,nace represents national value of GVA of sector j. Following the same logic, regionalisation of compensations of employees (denoted as w) was enumerated including the transformation of values from NACE to CPA (3). Based on information about the compensations of employees, the value of wages and salaries (ws) are calculated utilizing their fixed shares on national data (4). w j r,cpa n,cpa = w w r,nace j j n,nace (3), w j ws j r,cpa n, CPA = ws w j r,cpa j n,cpa (4) w j Outstanding parts of gross value added (net operating surplus (π), net taxes on production (t) and consumption of fixed capital (α)) are calculated as fixed share 5 on the value of GVA reduced by the previously calculated value of compensations of employees. π j r = (v j r w j r ) t j r = (v j r w j r ) α j r = (v j r w j r ) π j n (v j n w j n ) t j n (v j n w j n ) α j n (v j n w j n ) (5) (6) (7) Total intermediate consumption (za) in purchasers prices was in the next step calculated as the difference between output and gross value added. za r j = x r r j v j (8) 5 Based on national data.

5 To get intermediate consumption in basic prices value of net taxes on products (taxes less subsidies - ts) was calculated on the basis of a sector-specific national share of this item on the intermediate consumption in purchasers prices. ts j r = za j r ts j n za j n (9), and finally, we can enumerate total intermediate consumption in basic prices as z r j = za r r j ts j (10). In the following step, values of elements of intermediate consumption (z) in the inter-industry matrix are calculated on the basis of following formula: z r i,j = z n i,j z j r n (11), z j where z r i,j represents the value of intermediate consumption of sector i produced by sector j in the region r, z n i,j is national value of intermediate consumption of sector i produced by sector j, z r j is value of total intermediate consumption (basic prices) of sector j in region r and z n j represent value of total intermediate consumption (basic prices) of sector j on national level. Application of this approach leads to difference between regional and national IO coefficients, but on the other hand assumption that the share of intermediate input in total intermediate use of branch j is constant is applied this is called as weak version of equal technology assumption (Kronenberg and Tobben, 2011). Estimation of the regional final use In the next step elements of final demand are calculated. In contrast to (Kronenberg and Tobben, 2011), we have used more simplified approach 6. To approximate final consumption of households available information about total regional household s income was utilized in the following form: hc r i = hc n i hir hin (12), where hc r i represents final consumption of products of sector i in region r by households, hc n i is final consumption of products of sector i by households, hi r is total households income (net disposable income B.6n by ESA10) in region r and hi n represents total national households income. Application of this formula maintains proportional structure of consumption across regions, which in reality represent a strong assumption whilst significant interregional differences are more likely. Structure of non-profit institutions serving households consumption is based on its national shares on households consumption of sectoral demand of households resulting from previous step. npish r r j = hc npish j n j n (13). hc j Final consumption of government (gc) was calculated also on the basis of modified approach, where shares of intermediate consumption of public administration (NACE - O), education (NACE - P) and health (NACE - Q) in the given region on national intermediate consumption of those sectors were utilized. gc r i = gc n i z r r r i,j=o+z i,j=p +z i,j=q n n n = gc n i z r i,j={o,p,q} i,j n (14), z i,j=o +z i,j=p +z i,j=q i,j={o,p,q} z i,j and final consumption expenditure is the sum of respective rows of its elements fc r i = hc r i + npish r r i + gc i (15). 6 There exist publicly available data on regional household conumption by use structured by 10 groups of COICOP expenditure and NUTS 2 level. This paper use regional structure on NUTS 3 level and 11 groups of expenditures by products (CPA). There is possibility to do further research on using individual data provided by HBS (household budget survey) in the future to obtain more precise estimates of household consumption, despite that limitation of these data was already discussed in the cited work (Kronenberg and Tobben, 2011).

6 To approximate regional level of gross fixed capital formation by CPA classification available regional data by NACE were transformed in a similar manner as in case of GVA in (2). r,nace gfcf r,cpa i = gfcf i n,cpa n,nace gfcf gfcf i (15) i Regionalization of inventories data is proportional to the respective shares of the regional gross fixed capital formation. inv r i = gfcf r i inv i n n (16) gfcf i And finally, it is possible to enumerate gross capital formation (gcf) as the sum of gross fixed capital formation and inventories gcf r i = gfcf r r i + inv i (17). Now we have all the information to enumerate final domestic use (fdu) as fdu r i = fc r r i + gcf i (18) Estimation of trade flows In the following steps, data about exports and imports are estimated. On a regional level, we have to take into consideration both international and inter-regional trade flows. To calculate international trade flows, known data about foreign exports and imports are regionalized 7. In case of exports, we assume that the share of foreign export is equal to the share of regional output on the national level of output by branch e r i = x i r n n e x i (18). i We assume, that regional foreign import share is equal to the volume of final domestic use and intermediate consumption of the given region on a national level Impsh r i = fdu i r r +z i fdu n n (19) i +z i Now we can transpose import share back to respective columns and calculate regional foreign import m r j = Impsh r n i m j for all i=j(20). The crucial part is related to the estimation of interregional trade using the cross-hauling method. Inter-regional trade flows are calculated in the applied methodology based on cross-hauling potential. Firstly, national cross-hauling is calculated in line with the following formula: q n j = (e j + m j ) e j m j = tv j b j (21), where q n j represent amount of national cross-hauling for sector j, e j is total international exports of sector j, m j represents amount of international imports of sector j, tv j denotes trade volume of sector j and b j is trade balance of sector j on national level. In the next step, national cross-hauling potential is enumerated as: n h jn q j = 2 min(x n j,z n j +fdu n j ) where h in represents national cross-hauling potential of sector j, x j n represents output of sector j on national level, z j n represent national level of intermediate consumption of sector j and fdu j n denotes final domestic use for production of sector j. Based on the information about the national cross-hauling potential amount of regional cross-hauling is estimated using the following formula: q jr = 2 h jn min(x j r e j r, fdu j r m j r, x j roc e j roc, fdu j roc m j roc ) (23), where superscript n denotes national value, superscript r represent regional level and roc is used for values summing remaining regions (rest of the country). Minimizing conditions are aimed at finding a (22), 7 In some countries, regional external trade flows are covered by national statistics. However, the trade statistics due to methodological differences provides slightly different information related to flows. There can be also utilized other proxies of intraregional trade flows, such as transportation statistics.

7 feasible solution, thus relations between output to external export and domestic use to external import. Next, the commodity balance on a regional level is calculated by using the classical approach: b r i e r i m r i = x r i z r r i d i (24). 8 And finally, based on previous steps inter-regional exports (ei) and imports (mi) by sector are estimated using the following formulas: ei r i = tv i r r +b i (25), 2 = q i r + b i r +bi r 2 mi i r = v i r b i row 2 = q i roc + b i roc bi roc 2 (26). Finalisation As final steps of creation of MRIOTs, we can enumerate total exports (te) of the region as the sum of internal and external export and total imports in the same way te r i = ei r r i + e i (27) tm r i = mi r r i + m i (28) Now we can fill the last missing information in regional IO tables, such as the final use fu r i = fdu r r i + te i (29) and total use with the addition of the intermediate use tu r i = fdu r i + te r r i + z i (30) We can alternatively estimate import form other regions as the gap between total use, production and foreign import mi r i = tu r i x r i m r i (31). To conclude, we just need to enumerate total regional supply as the sum of output and import tsup r j = x r r j + ti j (32). Single region input-output tables are then combined into the multi-region input-output table and balances are checked. To enumerate all interregional trades, we should apply the optimization RAS approach 9. In the following part, the obtained results are compared to published by VŠE 10 and based on the methodology of Sixta and Vltavska (2016). Results Input-output tables comprise detailed information about relations between economic sectors within a defined region (national/regional level) thus only selected indicators of two available approaches are compared in this chapter. From the Gross domestic product (GDP) point of view, both analysed approaches provide relatively coherent results. Largest deviation (3.4 p.p.) between methods is visible in the case of Prague region, in which CHARM method provides more accurate estimates when compared to official statistical data. On the other hand, significantly more similar results were 8 In this paper, we use modified CHARM formula, which partially resolves the problem related to re-export of various commodities between regions. Some concerns in this respect remains and are discussed in (Tobben and Kronenberg, 2015). However, in terms of re-export, it s only effect is related to the size of regional trade, because re-export affects similarly export and import in particular region, thus it does not significantly affect the regional trade balance. 9 For more details see e.g. Trinh and Phong (2013). 10 Department of Economic Statistics, University of Economics, Prague (

8 obtained by the method of Sixta and Vltavska in case of Zlín and Ústí regions. Almost identical results in respect to the value of regional GDP were obtained in the case of Olomouc region. In general, both applied methods provide underestimated values of regional GDP compared to available official statistical data. Figure 1: Regions of the Czech Republic NUTS 3 Source: Transport yearbook of the Czech Republic, Ministry of Transport, 2013 Pha Stc Jhc Plz Kar Ust Lib Krh Par Vys Jhm Olm Zln Mrs Real CHARM Sixta, Vltavska

9 Figure 2: Regional gross domestic product11, market prices, mil. CZK, Czech Statistical Office On the national level, official statistical data for real GDP in the year 2013 was at bn CZK, by summation of regional GDP values based on selected regionalization methods following values were estimated: for CHARM bn CZK and for Sixta and Vltavska 4 077bn CZK. This minor difference to the real data on a national level is a result of the statistical revisions. On the other hand, it can be already observed that slight differences on regional level caused by a different approach to regionalisation, occur Pha Stc Jhc Plz Kar Ust Lib Krh Par Vys Jhm Olm Zln Mrs Sixta, Vltavska CHARM Figure 3: Regional output, market prices, mil. CZK Estimated regional output at market prices is relatively more differentiated than in case of GDP estimates. This is because in case of CHARM method output by product was regionalized based on available regional data on compensations of employees, and Sixta and Vltavska (2017) utilized special allocation keys for each industry in all fourteen regions. Largest deviation is visible in case of Central Bohemia region that is geographically neighbouring with metropolitan Prague region, thus methodology for compilation of wage and compensations statistics should be relatively inaccurate due to a significant amount of commuting and strong economic interlinkage with the capital region. On the other hand, significantly higher regional output was estimated by CHARM method in the case of Olomouc region. From the overall point of view the national total output as a summation of regional outputs by Sixta, Vltavska gives the amount of bn CZK which is slightly lower compared to official national statistics of bn CZK for The applied methodology of CHARM that is based on regionalization of national data. Thus, the summation of regional values provides the exact amount of national output. 11 Notes: Pha Prague, Stc Central Bohemia region, Jhc South Bohemia region, Plz Plzen region, Kar Karlovy Vary region, Ust Usti region, Lib Liberec region, Krh Hradec Kralove region, Par Pardubice region, Vys Vysocina region, Jhm South Moravian region, Olm Olomouc region, Zln Zlin region, Mrs Moravian- Silesian region

10 The comparison of the regional trade-flows provides interesting findings specifically in the interregional part. International exports and imports regionalization were in both approaches allocated to regions product-by-product using the similar methodology. Thus, the differences in the values arise only from the deviations in the regional estimates of underlying variables: regional output and regional domestic use (i.e. intermediate consumption and final demand less exports) Pha Stc Jhc Plz Kar Ust Lib Krh Par Vys Jhm Olm Zln Mrs CHARM Sixta, Vltavská Figure 4: Regionalization of international imports Pha Stc Jhc Plz Kar Ust Lib Krh Par Vys Jhm Olm Zln Mrs CHARM Sixta, Vltavská Figure 5: Regionalization of international exports Differences that are more significant occur in the case of inter-regional trade flows because different approaches were applied. In the work of Sixta and Vltavska, RAS method was utilized. This method is described in detail in Trinh and Phong (2013). While estimation of inter-regional trade flows in our work is based on the calculation of cross-hauling potential in line with CHARM method described in

11 Tobben and Kronenberg (2015). From the overall point of view, significantly higher inter-regional trade activity was estimated using the CHARM method. This method utilized by authors estimated on a national level approximately by 70% more intensive domestic trade flows than in the case of Sixta and Vltavska (Figure 6 and Figure 7) Pha Stc Jhc Plz Kar Ust Lib Krh Par Vys Jhm Olm Zln Mrs CHARM Sixta, Vltavská Figure 6: Regionalization of inter-regional imports Pha Stc Jhc Plz Kar Ust Lib Krh Par Vys Jhm Olm Zln Mrs CHARM Sixta, Vltavská Figure 7: Regionalization of inter-regional exports In the next step of our analysis, deviations of regional intermediate consumption from values of the national symmetric input-output table were scrutinized (Figure 8). From the perspective of a share of intermediate consumption on output (sum of technical coefficients), there is observed more significant heterogeneity among products as in the case of output multipliers. CHARM estimation provides similar regional heterogeneity in terms of services with Sixta and Vltavska estimation, whilst estimation in the manufacturing and construction sectors seems to be more homogenous. The biggest difference can be observed in the Real estate services sector, which results from the effects

12 of methodological differences on the smallest sector. In this case, the distribution based on compensations of employees applied in CHARM method does not seem to be the reasonable proxy. CHARM A B+D+E C F G+H+I J K L M+N O+P+Q R+S+T+U Sixta, Vltavská A B+D+E C F G+H+I J K L M+N O+P+Q R+S+T+U Figure 8 (a,b): Deviations of analysed approaches from national shares of intermediate consumption on the output Analysis of output multipliers (type II) has shown much smaller variance in the results of the two compared methods. Largest deviations remain in the case of market services multipliers, while in the production of manufacturing much more similar results were obtained.

13 3 CHARM A B+D+E C F G+H+I J K L M+N O+P+Q R+S+T+U 3 Sixta, Vltavská A B+D+E C F G+H+I J K L M+N O+P+Q R+S+T+U Figure 9: Deviations of analysed approaches from national output multipliers (type II) In Figure 10 we compare the differences in sectoral output multipliers between regions and methods (CHARM and Sixta and Vltavská approach). We can observe that on average, there are estimated slightly lower multipliers by CHARM method particularly in the agriculture sector, manufacturing, transportation and trade as well as in information and communication.

14 25% 20% 15% 10% 5% 0% -5% -10% A B+D+E C F G+H+I J K L M+N O+P+Q R+S+T+U St.Dev Avg. Diff Figure 10: Relative differences in regional Output multipliers between CHARM and Sixta, Vltavská approach On the other hand, we can see that the average differences across all regions remain relatively low (up to 5%), the individual differences in output multipliers in particular regions described by standard deviation are significantly higher, especially in small sectors. The relative differences in output multipliers among individual regions are presented in Table 4.

15 A B+D+E C F G+H+I J K L M+N O+P+Q R+S+T+U Pha -7% -14% -2% -3% -6% -12% 1% 18% -18% 1% -4% Stc -10% 7% -15% 2% -1% -3% 19% 9% 1% 1% -6% Jhc -2% -20% -1% 6% -9% -8% 2% -23% -1% -4% 6% Plz -3% -13% -5% -2% -8% -1% -3% -19% -2% 1% -5% Kar -10% 29% 11% 12% -1% -5% 10% 2% -2% 0% -15% Ust -12% 14% -9% 9% -2% -29% 22% -10% 1% 3% 8% Lib -14% 14% -2% 1% -4% -5% 0% 8% -1% -1% 2% Krh -3% -12% -5% 1% 1% -21% 22% -18% -3% 5% 5% Par 0% 7% -1% 4% 5% -18% -5% 12% 2% 1% 6% Vys 4% -16% -2% 2% -4% -1% 6% 7% 2% 8% 2% Jhm -12% -28% -4% -6% -9% 7% -13% -10% -7% 3% -1% Olm 9% 17% 8% 16% 7% 4% 21% -21% 10% 12% 2% Zln -9% 7% -9% -5% -8% 3% 5% -26% -6% 0% -15% Mrs -10% 34% -6% 7% 0% 2% 11% 13% 6% 7% 13% Discussion Table 4: Relative differences in regional Output multipliers estimated by CHARM and Sixta, Vltavská approach Biggest differences between two compared approaches remain in the method of inter-regional trade volumes estimation and its significance, as well as relative structure. If we consider the volume of intra-regional trade in respect to external trade, the CHARM method provides significantly higher estimation (67% to 40% of Sixta and Vltavská). Moreover, published papers (e.g. Flegg & Tohmo, 2013b) discussed, that even CHARM method tends to underestimate overall regional trade flows. Within this indicator, the manufacturing sector has the biggest weight by the size of trade flows. The main structural differences between compared estimations are highlighted in Table 5. While both estimations suggest, that the main flows of manufacturing output are exported out of the country, only part is used within the country in different regions for final or intermediate consumption (41% estimated by CHARM, only 11% by Sixta and Vltavská). Significant differences can be observed also in Trade, transportation and accommodation services (G+H+I), where CHARM method suggests that majority of trade can be observed within the regions and Sixta and Vltavská estimation suggests that majority of output in these services is aimed towards external trade. A B+D+E C F G+H+I J K L M+N O+P+Q R+S+T+U Total CHARM 147% 111% 41% 571% 126% 160% 565% 4691% 149% 176% 184% 67% SIXTA 79% 138% 11% 865% 89% 87% 753% 3313% 126% 594% 202% 40% Table 5: Comparison of inter-regional trade volume (export) to sectoral export on a national level There are no official statistical data for regional trade volume in monetary values for the Czech Republic publically available. For further analysis, we can only open discussion about the plausibility of obtained results. For this discussion, we utilize the estimation of external and internal trade of most specific capital region (Prague). From the point of export from the Prague region abroad (marked as ext) and trade to other Czech regions (int), we can point out several observations. The share of regional external export on the total export of the country seems to be similar by both approaches. From the point of internal export, there are three sectors achieving extreme values for

16 both approaches, but in line with expectations. In agriculture and construction sectors, there is almost no export from the capital region to other regions. On the other hand, almost 100 % of internal export in financial and real estate services comes out from the capital to the other regions. The biggest difference from the export point of view is in mining, quarrying and energy sector (with overestimation in Sixta approach). This is caused by differences in reporting of output and workers among main energy companies in the country. The second highest difference is observed in other services (with the most significant part related to sector R - arts, entertainment and recreation). A B+D+E C F G+H+I J K L M+N O+P+Q R+S+T+U Total Ext CH 3% 16% 7% 26% 32% 60% 61% 42% 45% 24% 32% 12% Ext S 3% 17% 5% 24% 34% 63% 65% 38% 52% 18% 29% 11% Int CH 1% 12% 5% 4% 66% 76% 93% 96% 68% 6% 70% 32% Int S 1% 37% 4% 1% 68% 70% 88% 82% 66% 10% 49% 42% Table 6: Comparison of results for the Prague region (export) as share on a national level From an import perspective, differences that are more significant occurred among almost all sectors. CHARM method estimates, that quarter of the agriculture products were imported to the capital region, while Sixta and Vltavska are close to 50%. Both of those estimates are not in line with official transport data in tones of goods of agricultural products that were in 2013 unloaded in Prague region, which accounted for only around 2% of interregional trade-flows. This should be partially explained by the fact that one-quarter of interregional trade-flows of agricultural products are transported to Central Bohemia region that is surrounding Prague region and only after undertaking of processing in this region enters the Prague region. A B+D+E C F G+H+I J K L M+N O+P+Q R+S+T+U Total Ext CH 11% 17% 13% 32% 23% 42% 38% 25% 34% 25% 23% 16% Ext S 14% 16% 12% 31% 27% 48% 37% 26% 37% 22% 31% 15% Int CH 24% 17% 10% 79% 24% 33% 5% 1% 22% 40% 17% 17% Int S 51% 13% 24% 58% 12% 12% 8% 7% 20% 23% 18% 20% Table 7: Comparison of results for the Prague region (import) as share on a national level Despite the significant differences, the relative weights of trade especially in service sectors remain small. External trade focuses mainly on goods with a small share of services (10,7 % on imports and 14,5 on export), whilst share of services on internal trade is much more significant 38 % (both imports and exports due to balance condition). A B+D+E C F G+H+I J K L M+N O+P+Q R+S+T+U Trade Int 3.4% 5.1% 50.0% 3.4% 14.6% 5.9% 4.1% 3.6% 7.6% 1.0% 1.3% Imp Ext 1.9% 9.0% 78.1% 0.3% 3.7% 1.8% 0.9% 0.1% 3.8% 0.2% 0.2% Exp Ext 1.6% 2.9% 80.6% 0.4% 7.6% 2.4% 0.5% 0.1% 3.1% 0.4% 0.4% Conclusions Table 8: Weights of export and import on total trade (CHARM estimation) Source: authors Inter-regional trade-flows represent an important source of economic development of the national economy. In the Czech Republic, more than 50% of national output is produced in four (Prague, Central Bohemia, South Moravian and Moravian Silesian region) of fourteen NUTS 3 regions.

17 According to achieved results, approximately 40 % of the exports out of the given region are directed towards other regions of the Czech Republic and additional roughly 25 % of produced output is traded within the local region. Comparison with the more data demanding approach of Sixta and Vltavská (2017), results show that differences are strongly related to region type and distance of its characteristics from the specifics of the national economy. Despite, that there are relatively low average differences in output multipliers, the structural variability measured by standard deviation remains significant (10%). The applied approach allows for detailed analysis of regional development and provides evidence of structural differences among the regional economies of Czech Republic. Due to the unavailability of official data, we could not do many hard conclusions. Therefore, further research in this area needs to be performed. Additionally, as it is often mentioned in related literature, regional data still have many limitations and biases related to applied methodology. For example, regional data are based on the headquarters of the reporting unit, and problem prevails where the administrative unit (company) have non-reporting branches across several regions. Despite the illustrated differences, both approaches described in this paper provides a significant improvement in data sources for experimental (academic) regional modelling. In further steps, we are considering incorporation of information on the structure of household consumption by branches based on HBS. From the trade perspective, Czech ministry of transportation is providing the information related to the cross-regional transported volume of trade, valuable information that should improve the reliability of results by the application of RAS method, related especially to the manufacturing sector. Additionally, we would like to verify the stability of IO coefficients over time.

18 References Bonfiglio, A., & Chelli, F. (2008). Assessing the behaviour of non-survey methods for constructing regional input-output tables through a Monte Carlo simulation. Economic Systems Research, 20, Flegg, A. T., & Tohmo, T. (2013a). Regional input-output tables and the FLQ formula: A case study of Finland. Regional Studies, 47, Flegg A. T. & Webber C. D. (1997) On the appropriate use of location quotients in generating regional input-output tables: reply, Reg. Studies 31, Flegg, A.T. & T. Tohmo (2013b) A Comment on Tobias Kronenberg's Construction of Regional Input Output Tables Using Nonsurvey Methods: The Role of Cross-Hauling. International Regional Science Review, 36, Flegg A.T.; Huang Y.& Tohmo, T "Using Charm to Adjust for Cross-Hauling: The Case of the Province of Hubei, China," Economic Systems Research, Taylor & Francis Journals, vol. 27(3), pages , September. Kowalewksi, J. (2015). Regionalization of national input output tables: Empirical evidence on the use of the FLQ formula. Regional Studies, 49(2), Kronenberg, T "Construction of Regional Input-Output Tables Using Nonsurvey Methods," International Regional Science Review,, vol. 32(1), pages 40-64, January. Kronenberg, T. & Többen, J., "Regional input-output modelling in Germany: The case of North Rhine-Westphalia," MPRA Paper 35494, University Library of Munich, Germany. Lehtonen, O., & Tykkyläinen, M. (2014). Estimating regional input coefficients and multipliers: Is the choice of a non-survey technique a gamble? Regional Studies, 48(2), Miller, R. E. & Blair, P. D. [2009] Input-output analysis: Foundations and extensions [Cambridge, Cambridge University Press] Radvanský, M. (2017): Creating (Multi)Regional input-output table for Carinthia (and other Austrian regions), University of Klagenfurt, working paper - unpublished Sixta, J. & Vltavská, K. Regional Input-output Tables: Practical Aspects of its Compilation for the Regions of the Czech Republic. Ekonomický časopis. 2016, roč. 64, č. 1, s ISSN Vltavská K. & Sixta J.: Input-Output Tables for Regions of the Czech Republic Statistika: Statistics and Economy Journal, Volume 97, Number 2, 2017, pp. 4-14(11) Többen J. & Kronenberg, T "Construction Of Multi-Regional Input-Output Tables Using The Charm Method," Economic Systems Research, Taylor & Francis Journals, vol. 27(4), pages , December. Transport yearbook of the Czech Republic, Ministry of Transport, 2013

19 Trinh B. & Phong N.V. (2013). A Short Note on RAS Method. Advances in Management & Applied Economics, 2013, 3, pp