APPLICATION OF THE DECISION SUPPORT MODEL TO CZECH EXPORTS

: In this paper, we present our own modiﬁ cation of the Export Decision Support Model (EDSM) and its application on evaluation of export opportunities for Czech companies. The main objective of the model is to identify export opportunities which might help to target the export promotion policy of the Czech Republic. The EDSM methodology is based on sequential elimination of less interesting markets. We have amended and calibrated the EDSM for the Czech Republic and created a new variable – our Index of Export Opportunities – which classiﬁ es markets depending on their suitability for export. Calibration was performed on a data sample covering global trade data from 2006 to 2012. The results of the model have been used as a tool for an evidence-based export promotion policy in the Czech Republic. The results were tested by the real foreign demand for Czech products in speciﬁ c industries and importing countries based on information from Czech embassies and trade promotion organisations.


APPLICATION OF THE DECISION SUPPORT MODEL TO CZECH EXPORTS
Matěj Urban, Michal Mejstřík, Jana Gutierrez Chvalkovská*

Introduction
This paper presents modifi cation, calibration and application of the Export Decision Support Model (EDSM) based on Cuyvers et al. (1995), Cuyvers (1996), Cuyvers (2004), Cuyvers et al. (2009) and Pearson et al. (2010) as a tool for identifi cation of export opportunities. The cited papers identify suitable export opportunities for Belgium, Thailand and the Republic of South Africa. Our EDSM is designed specifi cally for the Czech Republic and, compared to earlier studies, it uses different critical values and adds important selection criteria with respect to the specifi cs of the Czech economy, which is small and highly export-oriented with a strong focus on manufacturing. Based on data from the Czech Statistical Offi ce, the gross export accounted for 79.6% and the net export for 8.1% of the Czech GDP in 2012 with a strong focus on EU countries (80.8% of Czech exports was to EU countries). An important improvement to previous studies is that our approach emphasises more the supply side of export opportunities, and also takes into account the import content of exports and enables fl exibility for various industries with respect to country risk, GDP growth and the Ease of Doing Business (World Bank Index). The second stage of our model combines the results of the general EDSM with additional factors to allow an individual sector or regional approach.
The main objective of this study is to demonstrate the modifi cation of the EDSM for the particular features of the Czech Republic, and show how it was applied to an assessment of the Czech Export Strategy. The paper is structured as follows: In the second section, we build the EDSM for the Czech Republic and explain the process of prioritising export markets based on the Index of Export Opportunities. The third section introduces the data used in the model. The fourth section presents the results of the model and compares them to the priority markets from the Czech Export Strategy. Finally, the last section concludes the paper.

The EDSM
The main motivation of this paper is to apply the EDSM to identify the Czech Republic's export opportunities, which might help to better target the Czech Republic's export promotion policy. The model is applied to assess the priority countries selected by the Czech Export Strategy and, besides, it enables us to specify priority markets for various industries. This makes the model a fl exible instrument that can be used for export recommendations on the national and industry levels.
The selection process of suitable export opportunities consists of sequential elimination of less interesting markets, which are gradually fi ltered based on specifi c selection criteria. The selection of the elimination criteria is inspired by Pearson et al. (2010); however, our approach brings three major improvements and calibrates the model specifi cally for the Czech Republic.
Firstly, we do not use the indicator of country risk in the fi rst step of the elimination. That enables us to preserve the possibility of an individual approach towards various industries that differ in their exposure to political and commercial risks. The EDSM also uses a similar approach in the case of expected GDP growth, enabling a selection of its desirable weight for the fi nal ranking of desired markets.
Besides, we have incorporated additional fi lters into our modifi ed EDSM, using the Lafay index (Lafay, 1992) and the index of revealed comparative advantage (Balassa, 1965), which refl ect not only the demand side of export opportunities, but also the supply side, thus refl ecting the ability of Czech exports to succeed on foreign markets.
Another important addition to Pearson et al. (2010) is that we took into account the global value chains, which is especially important for a small and open economy such as the Czech Republic. Only a part of the exported volume is formed by domestic value added, thus the absolute volumes of trade are often misleading indicators. Hence, in order to partially refl ect where the real value added of exports was created, the model adjusts critical values for market size based on the import content of exports.
The thresholds for sequential fi ltration of export opportunities were selected on the basis of a calibration process, during which we tested the model on export opportunities for specifi c markets and product groups along with specifi c geographical locations.
The selection of the calibration method was infl uenced by the following aspects:  the EDSM is a multi-criteria model with a number of parameters;  the EDSM is required to fi nd favourable export opportunities for the whole economy as well as for individual subcategories with different specifi cities;  ranking of parameter signifi cance in determining good export opportunities cannot be set in an unambiguous way;  some of the variables show a high degree of multicollinearity, which makes traditional regression methods diffi cult to use;  emphasis was on the model applicability in common practice. Therefore, one of the assumptions of the calibration process is that successful export in the past is an opportunity in the future 1 . We selected a reference group of exports consisting of export opportunities based on objective criteria 2 from the past and other export opportunities identifi ed as propitious by Czech embassies and trade promotion organisations such as CzechTrade. The reference group consisted of 350 export opportunities. For fi nding optimal sensitivity to various criteria, we tried to build similar reference groups of export opportunities for various subcategories including both country-specifi c markets and product-specifi c markets. The model was also tested on a few specifi c Czech exporters (real-life cases). Finally, the calibration process found the threshold values by minimising the type I error for all the cases while keeping the required number of opportunities and some basic setups of parameters from Pearson et al. (2010).
The structure of the sequential fi ltration is a result of attempts to maintain a balance between the supply and demand sides of export, trends in import, trade barriers, risks and global value chains. We use fi ve fi lters to identify the fi nal sample of export opportunities. The purpose and structure of the fi lters is explained in the text below.
First, we selected export opportunities based on the size of the importing market and its short-term and medium-term growth. As indicated by Havrlant et al. (2011), among the most important factors driving the Czech export are foreign demand (especially demand from EU countries), cost factors and the exchange rate. Cost factors and the exchange rate play a substantial role in the dynamics of the Czech export because they make up the price competitiveness of the country. The price competitiveness refl ects the comparative advantage of a country, which usually (omitting the trade barriers) results in the revealed comparative advantage (RCA). Thus, the specialisation on respective products of the Czech Republic was taken into account as indicated by the RCA modifi ed by the import content of exports in order to encompass where the value added was created. The critical value for export growth was calculated as follows: where gwj is the growth of world export of the product j in the last year and last 3 years respectively; The scaling factor consists of the revealed comparative advantage (RCA) and the import content of exports.
where X cz,j is the Czech export of the product j X cz is the total Czech export X w,j is the world export of the product j X w is the total world export FVA cz,j is the average import content of exports in the industry of the product j (as a share in gross export) FVA cz is the average import content of the Czech export (as a share in gross export) We applied a similar approach in the case of the relative size of the importing market. The critical value was calculated as follows: where M w,j is the world import of the product j.
Then, we sorted the potential export opportunities into categories based on the condition whether the import growth of the product j to the country i and the import volume of the product j to the country i was larger than or equal to the critical value (0 if not fulfi lled, 1 if fulfi lled). The fi rst step of elimination removed export opportunities where Filter 1 equals 0 (see Table 1 for the classifi cation). The following table summarises the classifi cation into 8 categories: Table 1 Classifi cation of export opportunities based on importing markets

Filter 1 Category Short-term growth
Mid-term growth During the second step of the elimination process, we removed opportunities with large trade barriers and diffi cult market penetration. Filter 2(a) uses the Herfi ndahl--Hirschmann Index (HHI) to measure market concentration (Hirschmann, 1964). We assumed that it was usually easier to penetrate less concentrated markets and thus we eliminated export opportunities where the HHI was higher than the critical values. The critical values for Filter 2(a) are the result of the calibration process and depend on Filter 1 as follows: 0.973 if Filter 1=3.
Filter 2(b) uses the assumption of Cuyvers (2004) that if countries with similar export structures and geographical locations are able to penetrate a specifi c market, then also the country of interest can penetrate the market. We used Germany to measure the revealed absence of trade for the Czech Republic, because Germany is a neighbouring country and it has a similar structure of export, measured by the Export Similarity Index (Finger et al., 1979). Germany has the highest value (0.507) of the Export Similarity Index for the Czech Republic. Thus, Germany has a supply structure of products for export similar to that of the Czech Republic, but is more successful in penetrating foreign markets. Filter 2(b) is determined as follows: where: , , , is Germany's export of the product j to the country i, X G,j is Germany's export of the product j, X W,i,j is the world export of the product j to the country i, X W,j is the total world export of the product j.
Another barrier for trade is the transportation costs, which usually increase with the transport distance. Filter 2(c) works with the distance (d i ) of the country's capital to Prague. The critical values for Filter 2(c) are as follows:   2 2,500 2 1 2,500 8,500 0 8,500 Filter 2 then combines Filter 2(b) and Filter 2(c), thus eliminating the export opportunities where both parts were equal to zero.
Both the supply side and the demand side of the export opportunities are incorporated into the model by means of Filter 3. Filter 3 compares the value of the Lafay index (valuing the comparative advantage of the Czech Republic) with that of a potential trade partner and chooses the opportunities where the Czech Republic has a relative comparative advantage. The export opportunities are divided into 4 categories and then eliminated if Filter 3 equals zero.
, , i stands for the country, j stands for the product, t stands for the year, M stands for import, X stands for export.
Filter 4 helps to sort the export opportunities based on market importance. We used Filter 4 for further elimination only in combination with other fi lters (see Equation (9) for the fi nal purge of export opportunities).
To determine whether the Czech Republic has a relatively large or small market share for a specifi c product group and country, we compared the degree of market importance of the Czech Republic with the top 6 exporting countries in the product group j to the country i. The EDSM calibration process brought the following critical values for Filter 4: where , , denotes the market importance of exporting the product group j to the country i for the Czech Republic.
Until now, we have used fi lters for selecting appropriate markets for every product, taking into account the supply side of the export opportunities. However, we have not prioritised among the products with regard to the potential to increase the competitiveness of the Czech economy. We therefore wanted to incorporate the export sophistication in order to favour those opportunities that are worth being supported in export promoting activities. Filter 5 thus uses the PRODY index (Anand et al., 2011) to classify export opportunities according to the perspective of the product for the Czech Republic. We use the Filter 5 for further elimination only in combination with other fi lters. As a result of the calibration, the critical values were selected as follows: 0 10, 000 1 10, 000 18, 000 5 2 18, 000 26, 000 3 26, 000 denotes the GDP per capita of the country i in international dollars.
In the last step of the elimination, we combined Filters 1, 3, 4 and 5 in order to make the fi nal purge of less interesting export opportunities. The specifi c value of the threshold is a result of a calibration process and model testing. We eliminated export opportunities, where: After the elimination process, we needed to incorporate additional factors, allowing an individual approach for different industries. The EDSM does not refl ect the country risk and expected growth when selecting desired export opportunities. We therefore created the Index of Export Opportunities (IEP) to enable combinations of the results of the general EDSM with additional criteria.
The IEP is structured as the weighted average of normalised export opportunities determined by the EDSM (IEDSM), normalised rank of the country in the Ease of Doing Business (IDB), normalised Economic Complexity Index (IEC) and normalised expected economic growth until 2018 (IEG). The IEP takes values from 0 to 1. The country with the highest number of export opportunities reaches the value one, while the IEP of countries with no opportunities equals zero. The following fi gure visualises the procedure of export opportunity selection and combination of the EDSM with other criteria. Procedure of market ranking Source: authors.

Filtering process
Se ng maximum acceptable risk

Data
The model uses annual values of exports and imports among countries classifi ed according to the Harmonised System of tariff nomenclature (to the 6-digit level, HS6) for 2006-2010. We use the Comtrade database as a data source, which provides volumes of imports and exports in USD. Furthermore, the model uses GDP and GDP per capita, including their forecast by the IMF. The import content of exports was estimated based on approximated data of OECD and WTO that assign each sector with an import content of exports (on the basis of the ISIC rev. 3 classifi cation). We developed a model for selecting export opportunities in a wider version, where the users can combine the main results of the EDSM with other factors such as: country risk -OECD (2013), the Ease of Doing Business indicator -World Bank (2013) and the Economic Complexity Index -Observatory of Economic Complexity (2008).
The sample covers 176 countries and 5,920 products according to HS6 classification. All the combinations of products and countries are considered as potential export opportunities (1,041,920 combinations in total). After an elimination process, only 20,041 export opportunities are considered desirable.
The back-testing of the critical values in the calibration process was made on the basis of a comparison between the results and the real demand for Czech products based on information from Czech embassies and trade promotion organisations (for example, the CzechTrade map of sector opportunities).

Results
After the completion of the fi ltering process as described in Section 2, we divided the product groups into categories. The EDSM identifi ed most of the opportunities for the Czech Republic in machinery and electrical equipment (6,544), followed by metal products (2,479), mixed, other (1,799) and plastics/rubber (1,345), as depicted in the following tables showing the frequency of opportunities by product groups.  The following map displays the priority markets for machinery/electrical equipment (in terms of the IEP). The model fi nds the highest number of export opportunities for machinery/electrical equipment in Russia, followed by India, Egypt, China and Germany. Map of export opportunities of machinery/electrical products (higher IEP means higher frequency of export opportunities) The following table simulates 4 how the ranking of suitable markets for machinery and electrical equipment changes with different weights for the IEDSM, IEG, IEC, IDB and maximum country risk (country risk by OECD).

Figure 3
Map of export opportunities of the Czech Republic given by the EDSM

Source: authors.
It can be seen that the priority countries as described in the Strategy correspond to the countries preferred by the EDSM in the majority of cases.
The EDSM assigns the most export opportunities to Germany (1,016), followed by Russia (866), Poland (650), Slovakia (629) and France (612). However, many of these opportunities, especially in Europe 5 , have already been discovered by Czech exporters. Thus, we can fi nd many unused opportunities outside the European continent. The Czech exports are highly dependent on imports from the EU, therefore the Czech Export Strategy 2012-2020 defi nes the key priority of keeping the position on the EU market, but also penetrating markets outside the EU, especially those markets defi ned as the priority and interest markets. From the perspective of the EDSM, the following countries have a very interesting potential for Czech exports: Kazakhstan, Brazil, the United Arab Emirates, China, India, Turkey, Belarus, Russia, USA, Georgia, Egypt, Saudi Arabia, Australia, Canada, and Vietnam.
The robustness of the model was tested by varying cut-off values and comparing the results for the entire portfolio of export opportunities with specifi c subcategories consisting of product-specifi c and country-specifi c markets. The type I error of the entire portfolio of export opportunities is equal to 4.5% when compared to the initial reference group of good export opportunities. Since the model is highly dependent on historical data, it favours the opportunities in traditional industries and disadvantages new dynamic sectors, which implies that the model is not very stable in the dynamic environment of international trade and thus requires recalibration in the future. Further research should also aim at developing a more robust index to account for import restric-tions and a more convenient parameter than distance between markets that takes into account the quality of infrastructure and is a better estimator of the costs of transport.

Conclusion
Our model for prioritising export opportunities is a helpful tool for optimising export activities and competitiveness of the Czech export policy. It identifi es suitable export opportunities based on growth potential, absorption capacity and compatibility in relation to the Czech economy. It comes with substantial improvements compared to previous export models such as Cuyvers et al. (1995), Cuyvers (1996), Cuyvers (2004), Cuyvers et al. (2009) and Pearson et al. (2010), putting emphasis on the supply side of export opportunities, taking into account global value chains and enabling choice of the desired acceptable country risk and weights of indexes specifi c for the market and industry of interest. The model is thus more fl exible and refl ects the needs of individual sectors and countries compared to previous studies. The model is calibrated for the specifi c case of the Czech Republic, but may serve as guidance for building similar models for other countries.