Final Report Summary - MAPCOMPETE (Mapping European Competitiveness) Executive Summary:There is widespread agreement that improving ‘competitiveness’ throughout Europe is at the heart of a recovery from the 2009-2010 crisis. Firms increasingly base their choices on these parameters, and the European Commission increasingly monitors external imbalances using quantitative measures of aggregate competitiveness. For these reasons there is an overall increasing effort to quantify the concept of competitiveness, and even qualitative conditions of countries’ business environments are translated into quantitative indices.However, one of the most important lessons learned during the crisis is that such informational toolbox on which policy makers base their decisions can be outdated in terms of both data sources and data analysis. There is in fact no shared definition of competitiveness, let alone a consensus on how to properly measure it across countries and over time, with a number of aggregate indicators. The toolbox is particularly outdated when it comes to tapping the potential of micro data for the analysis of competitiveness – a serious problem given that firms rather than countries will compete on global markets.The aim of the MAPCOMPETE project was to help fill this gap by providing a review of recent advances in policy oriented theoretical and empirical economic research, so as to guide the investigation of competitiveness indicators and the potential development of new ones. Importantly, for all aspects of competitiveness, a crucial issue of the project has been a thorough assessment of data availability and accessibility, as well as a critical overview of new analytical methods that become possible as new data sources become available to researchers.The research results achieved by Mapcompete rely to some extent on either novel datasets available to researchers, or novel methodologies of analysis, or both. As such, they could bring further the policy debate on competitiveness in Europe, a debate which would greatly benefit from improvements in the quality of the underlying analysis.Nevertheless, the continuous development and improvement in data gathering and accessibility remains key for both the policy-making as well as the research perspective. Official cross-country firm-level data at the European level, although existing, at the time being are practically not accessible to the average researcher, and thus have a very limited use in terms of policy-relevant analysis.In terms of reallocation, the traditional macro stream of literature dealing with the effects of centralisation of wage bargaining institutions on employment and wage outcomes has generally lead to inconclusive results because the variation in the level of bargaining used in these papers is exclusively across countries, with no variation at the industry and firm level. Hence it becomes difficult to detect the impact of these variables separately from industry time trends, time dummies, and country dummies. When data can be brought to the firm-level dimension, instead, an entire new range of policy-relevant results emerge. Still, despite the policy-relevant implications it could generate, working toward a higher availability of data on collective bargaining regimes, and more in general on labour market institutions at the micro-level, does not seem high in the agenda of the European institutions.The results obtained on the impact of Global Value Chains and non-price factors on growth and competitiveness crucially rely on the presence of detailed and comparable trade data that go well beyond the average statistics on imports and exports. The continuous availability of updated and detailed Input/Output tables, as well as reliable information on traded products both in terms of quantity and values remains also central in all future analysis of competitiveness.Project Context and Objectives:“Mapping European Competitiveness” (MAPCOMPETE) is an FP7 project of six European universities and research centres (from Belgium, France, Germany, Hungary, and Italy) to provide an assessment of data opportunities and requirements in order to analyze and compare competitiveness in European countries.Partners of the project are Brussels-based think tank Bruegel, Budapest-based research centre CERS–HAS (coordinator), Milan-based Centro Studi Luca d’Agliano (LdA), Paris School of Economics and Sciences-Po in Paris, and Tübingen-based research institute IAW. Associate partners are the OECD, ISTAT, the ECB, and several European central banks.Competitiveness is at the heart of policy making at the Union level and specifically within the Eurogroup. Therefore, the definition of new country-level competitiveness indicators is an essential task. The aim of this project is to provide a thorough assessment of data opportunities and requirements for the analysis of comparative competitiveness in European countries.The MAPCOMPETE project examines how to interconnect different approaches to research and policy making in this field at the macro, sectoral and micro level as well as to consequently map data availability and needs. The project team comes up with proposals for enhancing standards and consistency of data, with the aim of improving future comparative work on competitiveness. It analyzes data requirements and suggests methods of data collection for selected topics such as global value chains, trade and performance as well as pricing and quality.The project aims at identifying gaps in available datasets and key data requirements for constructing better competitiveness indicators at different levels of aggregation. A key objective is to analyze the combined use of three types of resources: census-type quantitative data (e.g. national tax authority), quantitative survey data (e.g. EFIGE survey), and qualitative (interview-based) information. Integrating these sources may allow for a deeper understanding of a wide range of topics related to competitiveness. Indeed, in terms of competitiveness analysis, the primary focus is on firm performance. However, it is approached from a wider angle. Hence, the data analyzed do not only cover traditional balance sheet figures, but they also include areas such as:- trade statistics, internationalization (outsourcing, direct investment, etc.),- labor statistics (skill composition, remuneration, on the job training, flexibility),- R&D, innovation,- non-tangible assets,- regional and local dimensions,- creation of new firms (entrepreneurship) and attraction of foreign investment (FDI),- stakeholders (entrepreneur/owner characteristics, social capital, state/local government),- customers/suppliers, position in value chains.Based on the framework outlined above, MAPCOMPETE covers the following topics:- Mapping existing datasets: screening of national sector- and micro-level datasets regarding their geographical coverage, time span, and representativeness. A special focus will be on areas where no standard set of variables exists, such as non-tangible assets and innovation. - Consistency issues of different datasets: building on existing research to understand the extent to which some country- and time-specific competitiveness-related indicators can be derived from data contained within available datasets or from cross-reference usage of available datasets at different levels of aggregation.- Conditions and requirements needed to match different datasets: the extent to which datasets relevant for competitiveness can be matched within and across countries and to which data gaps and potentials can be mapped is based on the pilot indicators.- Research directions towards better competitiveness indicators: investigation of how novel data or a combination of datasets can be used to introduce novel research areas. Designing new research directions leading towards better competitiveness indicators.- Benchmarking: identification of steps to enhance quality and availability of existing data and suggestion of new methods and sources for the collection of data.MAPCOMPETE publishes two main reports summarizing the key findings of the group, intended for the widest possible diffusion. One report will present the evidence retrieved from the pilot indicators of competitiveness developed by combining existing datasets, together with their mapping at the EU level as well as the identification of data gaps and potentials.The second report concludes the project by presenting policy options on the possible improvements in data collection and utilization related to the development of existing as well as new competitiveness indicators. An important output of the project is a data map is available on the MAPCOMPETE website. This data map is implemented on the basis of a database that will allow researchers and stakeholders to assess the availability and computability of indicators of competitiveness within and across countries.Project Results:There is widespread agreement that improving competitiveness throughout Europe is at the heart of the structural resolution of past and future crises. Firms increasingly base their choices on parameters related to competitiveness, and the European Commission continuously monitors external imbalances using quantitative measures of aggregate competitiveness. For these reasons, a number of international institutions, such as the European Commission, the European Central Bank, the World Bank and the World Economic Forum, are committed to producing regular comparative reports on competitiveness at the national level, at the regional level.Despite the availability of numerous publications and reports on the issue, there are some serious challenges in terms of the conceptualisation and measurement of competitiveness. First, although many studies and reports measure the competitiveness of firms, regions, or nations, there is no common single definition of competitiveness. Second, there is no consensus on how to properly and consistently measure competitiveness for different countries and/or over time. Even though a number of aggregate indicators (eg real effective exchange rates, unit labour costs, export share and prices) are available and broadly used, they can suffer from measurement errors, and do not necessarily deliver the same ranking for different countries or across time. Third, there is no single and/or harmonised dataset that enables the different facets of competitiveness to be captured in an internationally comparative perspective.EU-wide data collection methodsIndicators that require the use of basic balance sheet data (e.g. labour productivity, TFP) – along with trade indicators – are the most computable among the bottom-up indicators surveyed by Mapcompete, but there are country-specific problems. Also, bottom-up indicators on firm dynamics, which are based on data about company entries and exits, are poorly computable for several member states. In some cases the information needed is available, but only for a subset of enterprises or for a limited time period.Much of this heterogeneity can be explained by the fact that countries report various databases as the best possible source of information on firm dynamics, balance sheet and financial statement items. There are NSIs that report survey data as the best possible source of information, while others indicate that administrative databases are available for statistical use.In most member states, national legislation supports the use of administrative data for statistical purposes – under different confidentiality restrictions – and provides special rights for the NSIs to access these sources. However, there are several factors that hamper the effective use of administrative sources. First, legislation that requires the use of administrative data whenever possible is rare. As a consequence, NSIs are not motivated to make investments in order to fully exploit administrative data. They use such data, but only if it can be used with minor adjustments as part of existing practices.Second, most countries lack a coherent and comprehensive framework for collecting, storing and providing access to collected data. Different production units of NSIs perform admin-data related tasks separately, thus the use of administrative data is based on ad-hoc agreements with limited scope between the NSIs’ production units and the data holders. There are, however, positive examples: Portugal replaced all surveys of Structural Business Statistics with one new data-collection system for administrative and statistical use, while Bulgaria introduced a single entry point for reporting fiscal and statistical information.Third, cooperation between admin-data holders and NSIs is weak or difficult in several countries, partly because of the lack of legislation establishing the corresponding duties of data holders. In most countries, NSIs have no impact on the design of administrative data collection and authorities do not have to consult NSIs when introducing changes to data collection practices. These aspects have been addressed in an amendment to Regulation (EC) No 223/2009 which aims at establishing a legal framework for more extensive use of administrative data sources for the production of European statistics without increasing the burden on respondents, NSIs and other national authorities. NSIs should be involved, to the extent necessary, in decisions about the design, development and discontinuation of administrative records that could be used in the production of statistical data. NSIs should also coordinate relevant standardisation activities and receive metadata on administrative data extracted for statistical purposes. Free and timely access to administrative records should be granted to NSIs, other national authorities and Eurostat, but only within their own respective public administrative system and to the extent necessary for the development, production and dissemination of European statistics.Legal and administrative constraints of access to micro-level dataThe MAPCOMPETE data mapping exercise revealed substantial differences between EU member states in terms of the accessibility of micro-level information needed to compute the surveyed competitiveness indicators. It was observed that there are countries for which many bottom-up indicators have a relatively high level of computability, meaning that the required information exists in some meaningful format at the local statistical authorities, but micro-data access is not allowed for outside users.Legal barriers related to confidentialityWhile the rules of micro-data access are not clearly specified in several countries, it is clear that confidentiality restrictions substantially differ in different member states. The common feature of national laws is that they oblige institutions collecting personal or firm-level data to guarantee the anonymity of respondents. However, various definitions of confidential data and different approaches to data protection are present. Research entities have the option to access personal data in the majority of countries, but there are significant differences in national confidentiality restrictions regarding the transmission of data from the collecting institution to other entities. Some member states do not allow the transmission of certain confidential data, or the implementation is problematic.The new EU statistical law emphasises the importance of the availability of confidential data within the ESS network. It states that the transmission of confidential data between ESS partners may take place “provided that this transmission is necessary for the efficient development, production, and dissemination of European Statistics or for increasing the quality of European statistics”. The access to confidential data for scientific purposes also requires the approval of the national authorities which provide the data. However, our experience suggests that despite the legislative underpinning, there are several factors that hinder the research use of micro-data, and the exact methods, rules and conditions of access are still to be developed in many member states.The mapping of micro-level information also highlights the fact that different types of data are treated differently. In some EU member states, different regulations apply to different databases. In general there are stricter regulations on registry-type data and on databases that have full coverage over the observed population. Survey type data, especially data from harmonised surveys like CIS, is usually easier to access. Experiences with the access to individual-level trade data are mixed, since these databases include information both from administrative sources (ExtraStat) and from a harmonised survey (IntraStat).A distinction in confidentiality restrictions is particularly important when we consider the potential use of bottom-up indicators that are based on information obtained from different sources in different countries. For instance, firm entry and exit information and balance sheet data are obtained from administrative sources in some countries, while others conduct surveys to collect the information. Consequently, the computability and accessibility of bottom-up indicators based on these data is likely to differ in different countries and a harmonised approach to confidentiality protection is hard to achieve.It is worth mentioning that Eurostat provides access for scientific purposes to certain European survey data including the Labour Force Survey and the Community Innovation Survey. Recognised research entities conditional on the approval of their research proposal might access micro-data anonymised by Eurostat on electronic devices or non-anonymised data in Eurostat's ‘safe centre’. Currently, Eurostat negotiates on the possible dissemination of the micro-data on a case-by-case basis and proposes a unique anonymization methodology to all member states. Member states might refuse Eurostat’s proposal if it conflicts with national legislation, and thus micro-data will not be available for all member states.Practical (technical) constraints on accessibilityIt can be observed that in addition to national legislation, the internal regulations of data-collecting institutions and practical constraints also affect the accessibility of micro-data. It occurs that practical barriers hinder the accessibility of the databases compiled by the NSI: a safe environment for data security is not yet in place. Part of the variation in these matters can be explained by the fact that the increased demand for micro-data is a relatively new phenomenon. The resources available to NSIs for disclosure control (output checking), and their prior experience in the field, might influence the speed and direction of adaption. The development of new statistical disclosure methods needed to provide access to micro-data might be hindered by organisational, methodological and software problems.Currently, at the national level, the most commonly used method to provide access to micro-data is the release of scientific use files. In case of research use files, statistical disclosure methods and restrictions on access and use – e.g. license or access agreements – are applied simultaneously. Our data mapping exercise shows that several NSIs provide access to micro-data in data laboratories. Data laboratories allow researchers to use more identifiable data under strict conditions. In most cases, users are legally obliged to keep the data confidential, and are subject to close supervision and output checking. Since setting up a data laboratory takes time and resources, there are countries where this form of micro-data access is not yet available. Remote execution is also possible in a few member states. Note that the cost of operating a data laboratory or remote access services significantly increases with the number of users, mostly because output checking is completely manual in almost all of the member states. Consequently, even in the countries where the NSI already provides access to micro-data, revision of data protection practices will be inevitable in the near future.Non-legal barriersIssues with metadataHaving basic information about datasets in advance is a very important factor that might affect the success of a research project. Researchers need to have detailed information on the available datasets including the identity of the owner of the data, the exact content, the quality of data and the rules of access. These pieces of information are necessary to decide whether the dataset is suitable to their needs and whether they apply for access.International standards already exist for the international exchange of metadata. Statistical Data and Metadata Exchange (SDMX), an initiative sponsored by the Bank for International Settlements, ECB, Eurostat, International Monetary Fund, OECD, United Nations and the World Bank, aims to provide standards for the exchange of statistical information (e.g. formats for data and metadata, content guidelines, IT standards). Particularly for Europe, the European Commission has set up a recommendation ‘on reference metadata for the European Statistical System’ , which refers to the European Statistics Code of Practice and is based on the SDMX framework.While Euro SDMX Metadata Structure (ESMS) Metadata files for all of the statistics published by Eurostat are provided – and other international organisations also provide structured metadata on their statistics – our experience shows that there is still a big hole in the information on data. ESMS metadata files present useful information on methodologies, quality and the statistical production processes in general, but usually provide very little information on the link between the aggregate indicator and micro-data used to compute the given indicator. Also, country-specific information on survey and sampling design is often sketchy. We made use of the information provided in ESMS Metadata files when mapping the readily-available aggregate indicators, but we found that in order to be able to assess the strengths and weaknesses of these indicators to improve their quality or to propose new ones; much more information on the available national micro-data would be needed.Gathering comprehensive information on micro-data available in EU member states proved to be a challenging and time-consuming task. The amount and structure of information available on the websites of NSIs and other national data providers is very different in different countries. It is usually insufficient to fill the MAPCOMPETE MetaDatabase and it is definitely insufficient to plan a research project. In many cases, researchers obtain information on given datasets from scientific publications or through informal channels, which are burdensome and usually result in incomplete information. Also, when conducting cross-country comparative research or research that requires the use of information from more than one source, researchers have to search through several websites and publications, each with different metadata structure and information content.Since in MAPCOMPETE we collected a huge amount of information in a systematic manner, we tried to directly contact staff within the NSIs in all the EU28 countries to gather the relevant information. After a few months of the project, it became apparent that this was highly complicated, so we decided to gather information by exploiting existing contacts built up in another international project (CompNet) and from other personal contacts. In some cases, these contact persons were able to help us fill in the MAPCOMPETE MetaDatabase and in other cases they referred us to people within the NSI. The fact that in most countries economic databases are collected and handled by more than one institution – the NSI and the national central bank (and sometimes other institutions) both collect data in most cases – made it even harder to obtain the required information. Also, smaller countries and newer EU members tend to have less experience in handling requests for micro-data access, and consequently are usually less prepared to provide systematic information on existing data. The experience gained during the data-gathering process shows that the availability of information on the data is at least as important as the availability of data itself. Performing EU-wide research projects on competitiveness or designing new indicators is not feasible without easily available, comprehensive information on national micro-data. This is why the MAPCOMPETE MetaDatabase is especially useful for future research on measures of competitiveness. Furthermore, it serves as a basis for suggestions for possible improvements to data sources, treatment of data, conditions of access etc. It might promote quality research by providing detailed information on the accessibility and availability of data related to the measurement of competitiveness. However, the MAPCOMPETE MetaDatabase is only a snapshot of competitiveness-related data. A regularly updated, structured, easily available and comprehensive meta-database on national micro-data – that might include the experience of other researchers working with the data – might substantially increase the efficiency of international research projects.Issues related to the nationality of the data userAs part of establishing the European research space, conducting research and analysis on the basis of foreign data becomes important. Several specific problems arise in terms of foreign access to datasets located in countries other than the nationality of the researcher. First, in some countries access to micro-data is allowed only to researchers who are citizens of the country of the data provider or affiliated with a national institution. Second, language barriers are obviously a serious burden, since in many countries information is provided only in the national language, but one that can be solved by simply offering data description and variables in English. Several NSIs have made a great deal of progress in this respect, including metadata provision in English. Third, the provision of data on site might not be a burden for locals, but can be very costly for foreign researchers. Hence, setting up secure remote access – such as is available in Finland, France, Germany and Sweden – would be an important step. Finally, making access by foreigners easier by appointing an English-speaking specialist could indeed facilitate European research integration.Unclear rules of accessWhen mapping the accessibility of data, we faced the obstacle that it is often challenging to obtain precise information on the conditions of access to confidential data. Information on the accreditation process, statistical disclosure control methods applied and the practical details of access is usually not clearly specified on the website of the data provider or at any other publicly-available source. We found that one had to contact the data provider directly in order to clear up the details and to find out if access to the data is possible and under what conditions.There are substantial differences between countries in terms of the clarity of rules of access. In many countries there is some settled, formal procedure of applying for access (e.g. Denmark, Finland, France, Netherlands, Slovenia and Sweden) while other countries are less advanced in this respect and handle requests on a case-by-case basis. However, regardless of the sophistication of the application procedure, in most cases, it is required to present a research project which needs to be approved. This approval creates room for discretionary decision-making and informality which might differ from country to country, but is really difficult to assess.The approval procedure might be more problematic when the data provider does not perform output checking itself, but it is the researcher’s responsibility to protect the confidentiality of data. If data protection is delegated to the researchers then the cooperation strongly relies on trust between the data provider and the researcher, and it might be hard to define exact criteria.Truncated dataIn many cases, micro-data is provided in truncated form; that means it is made available with less information than the original source, in order to prevent the risk of disclosure (sensitivity) and for cost reasons. For the purposes of our discussion, this aspect is related to accessibility, but it can affect computability when it prevents the merging of different datasets.Sensitivity truncationSeveral statistical disclosure methods used to protect the confidentiality of data lead to a loss of information and might affect the quality of analysis carried out on the data. Let us first present key obstacles and make suggestions for their treatment.Four issues are identified that matter for practitioners:1. Sensitivity of information on selected firms;2. Recoding data into broader categories;3. Removing or modifying variables;4. Other disclosure measures.The first issue is related to the sensitivity issues of aggregated data. In some sectors, size categories or regions, there are only very few firms. Aggregating data on them would imply that in some categories only one or very few firms would feature and hence, their individual data would not be protected. To avoid this scenario, most statistics institutions and central banks or research outlets protect confidentiality by setting up compulsory aggregation rules. Typical rules include a minimum number of firms per aggregated band (this ranges between 4 and 9, in our experience) and maybe other controls such as market share of the top 5 firms in the aggregate.The second topic is a more general solution to keep identification impossible. This entails aggregating some existing firm categories such as industry or location address to protect the identity of firms. This process is especially useful in smaller countries where some regions or industries might include only a few firms, even if they are not large. Examples include merging four-digit industry codes into two-digit codes, merging municipalities or NUTS3 regions into NUTS2 regions, or replacing employment data with firm size brackets.Third, authorities might remove or replace variables. This might include the deletion of variables that would allow identification – this happens when some activity occurs rarely or is carried out by only a few firms. This might include balance-sheet items, such as subsidies, or some research activities in an innovation survey.Another option to prevent identification in general, and merging of datasets, in particular, is masking. This approach is divided into two categories depending on their effect on the original data: perturbative and non-perturbative masking methods. Perturbation implies the multiplication of all values by a random variable of unit expected value and a small but significant variance. This implies that say, sales values would be altered by a few percent without affecting any statistical relationship (given the unit expected value). Other options include rounding or truncation. In these cases identification or linking of the data to other data sources would be impossible or difficult because of the lack of exact matching.Importantly, researchers can often access sensitive information in, for example, the research lab, but there are strict rules for the information available outside the safe environment. Apart from these more common issues, authorities might apply individual controls or ask for a list of descriptive statistics to control the process. Statistics offices will often ask researchers to submit all relevant documentation – including programme code files, and descriptive tables for output checking before releasing results.Finally, note that in some cases an extreme application of this sensitivity approach is applied: individual data is aggregated right after data collection. In this scenario, firms are clustered by industry, location, size and only aggregate information is released. While this may indeed provide security, it washes out important features of observations that may be important for research.Dataset reduction for cost savingAnother factor that might reduce the scope of available datasets is cost saving. Every aspect of a dataset – number of variables, dimensionality and frequency of observations – will generate additional costs, mainly in terms of attention. Supervisors need to spend time on organisation of dataset management, cleaning and provision, and the costs of these will depend on the size and complexity of the data at hand. Saving resources and reducing administrative burdens are important in an era when NSI budgets are often being cut. As a result, aggregation and truncation of raw data are often carried out not for sensitivity but for cost purposes.One such practice is aggregation of some part of the dataset. Transaction-level data might be aggregated into annual aggregates. For instance, foreign trade is often registered at a very fine transaction level, but data is available mostly at annual aggregate level. Several variables might be deleted in order to avoid spending the time that would be required for consideration of sensitivity issues.Finally, another approach is exclusion of small firms. Dropping firms with fewer than five employees could reduce the size of a dataset by 80-90 percent, while retaining 95 percent of value added. However, such an exercise will limit analysis and understanding of important issues, such as entrepreneurship and firm dynamics.An important aspect of dataset reduction for cost saving reasons is European/international harmonization. Comparing statistics computed on the whole dataset or on firms with more than 10 employees might yield rather different results.Accessibility and matching of data from different countriesData matching opens up rich and novel research opportunities, especially when micro-level datasets are concerned. Existing micro-level data in European countries has significant potential in terms of record linkage and matching, including also commercial data and Big Data. Data matching and issues of matchability have considerably gained in importance in recent years. One reason for this lies in the increased accessibility of micro-level datasets and in the desire of researchers to merge these datasets within and between countries in order to increase the research potential of the data. There has also been significant progress in technical issues, not least driven by the rapid development of computer technology and data storage.The issue of data matching and matchability is of course not confined to the social sciences, but the recent economic crisis has made clear that economists require high-quality data, especially at the micro level, that is comparable across countries, in order to examine cross-country differences in competitiveness. However, comparable micro-data at the firm level in different EU countries is so far only available for some topics, most of which are not directly relevant for competitiveness (notable exceptions are the Community Innovation Survey or the EFIGE dataset). These comparable micro-level datasets are, however, all based on sample surveys.The huge potential of administrative data, which is already leveraged in many countries, is still waiting to be fully realised. There are, however, some serious endeavours in this direction, mainly based on the ESSnet projects and on the Framework Regulation for Integrating Business Statistics (FRIBS). These projects are of special importance because they are concerned with administrative data within the EU, which is of high quality. Any step towards making these data more comparable and accessible is more than welcome by researchers and policymakers. Therefore, ensuring the availability of such data should be a priority for the European Commission because this would ensure vastly improved analysis of cross-country differences in competitiveness, and of labour market issues and related fields.The most serious obstacles to matching micro-level data from different countries are still legal restrictions preventing data from being matched, because privacy and confidentiality are at stake. However, there is some activity in this area, namely within projects to evaluate the potential of analysing micro-level data without directly accessing the data.There are also obstacles to data matching within countries (see the KombiFiD example from Germany). This holds especially true if the datasets to be matched are held by different data providers, e.g. statistical offices, central banks, employment agencies or private data providers. However, progress has been made in this regard in recent years.Important steps to overcome the problem of data comparability between countries, particularly with regard to cross-country analyses of competitiveness, have been taken, for instance by the EFIGE project providing comparable firm-level data for 15,000 firms from seven EU countries. The ECB’s CompNet project is following suit. However, these two projects can only be regarded as first tentative steps towards data that can be used for cross-country analyses in the field of competitiveness that can be highly useful for policymakers.Overall, much has been achieved in the field of data matching within Europe in recent years, but the universe of cross-country and matched datasets is still sparsely populated and quite heterogeneous, with potential for improvement. Because of the ever-increasing need for high-quality datasets that can be used to inform policymakers, much more needs to be done. Cooperation between data providers within and in different countries is key, as is the reduction of red tape. Comparative analysis of competitiveness in different countries is ultimately only possible if comparable (micro) data exists in different countries or if data can be harmonised and made accessible to researchers. Ensuring the availability of such data should be a priority for the European Commission, because it would enable vastly improved analysis of policy-relevant issues.Towards better access, computability and matchability of micro-level data The information currently available to researchers on comparable measures of competitiveness for different countries is insufficient. Aggregate data, which is easily accessible and widely available, does not allow researchers to provide the answers that policymakers need. Micro-data on individual countries is mostly inaccessible to external researchers, and the situation is even worse when one tries to compare figures based on micro-data which are comparable for different countries. Only a few firm-level surveys are available, mostly only for one or a few years; there are few examples of matched data from different countries, and internationally comparable figures can be gathered only from a few micro-distributed data exercises. This is very different from, for example, the United States, where micro-level data from different states has been matchable and comparable since at least the mid-2000s. This implies that Europe lacks proper information to assess of the state of competitiveness at European level, compared to the situation in the United States.The first-best solution to overcome these bottlenecks would be to change the national and EU-level rules of data content, data availability, data matching and data access. The efforts undertaken by the ESS, with programs such as MEETS, FRIBS, FATS, SIMSTAT and ESS.VIP towards greater harmonization of data and the construction of pan-European data sets are useful initial steps in this direction. In particular, these initiatives can contribute to:• The reduction of the burden on enterprises in collecting and providing internal data;• The provision of a common ESS infrastructure framework for the production and compilation of business statistics with an appropriate legal background and new administrative mechanisms allowing for the sharing of information, services and costs among all ESS partners;• The definition of consistent data requirements and a common data quality framework, which will enable the linking and matching of statistics obtained as part of the regular collection of global business statistics.However, the time necessary to complete this process, and for its effects to be felt by researchers, is far too long and in the end might even prove almost useless, since it might well be that when these goals are finally reached, the next generation of researchers might highlight a different set of needs.Therefore, such long-term actions to change regulations need to be complemented with more short-term workarounds.The first workaround is to exploit the availability of improved methods and techniques, such as matching after separate processing (e.g. the Distributed Micro-Data Approach) or imputation. Projects such as CompNet or ESSLait provided important insights into new aspects of competitiveness by producing micro-aggregated statistics going beyond the first moment of the distribution of firms’ competitiveness indicators. However, if not properly supported by policy, these initiatives might remain one-shot exercises, whereas they need to be refined, constantly updated and carried out in a timely way in order to provide the more up-to-date figures for policy decisions. Two examples clearly highlight these risks: the ESSLait exercise provided figures up to 2010, while the more recent CompNet figures refer to the year 2011. Since these initiatives require researchers within data-providing institutions to run the codes prepared by the researchers, proper policy support is needed to enforce the requests to run micro-distributed exercises in as many countries as possible.The second workaround would be to improve techniques for matching and accessing micro-level data, either by improving architectures for data matching (e.g. by involving ‘matching institutions’) or for access to data by researchers (e.g. by improving techniques of data anonymization). Many NSIs have already developed or adopted elaborate methods and organisational arrangements in these areas. For example, in Germany, there is a well-established system of research data centres at several official data providers. Other countries like the Netherlands or France have established techniques of remote-data access. From a theoretical perspective, there are several additional ideas which could be rather easily adopted or, if necessary, adapted to national systems and legislation.It is worth mentioning that – after speaking to officials in NSIs, national central banks and other official data providers in many EU countries – we are quite persuaded that in most countries access to micro-data would be feasible for external researchers, but it is easier for the data providers to restrict access. While the official reason is often linked to legal issues about confidentiality, it seems that other factors might play a role. There are several approaches to allow access to data for researchers while maintaining confidentiality (such various forms of anonymization, or the creation of ‘matching institutions’), but these solutions have costs, and require the data provider to take some responsibility for the release of the data. Restricting access is cost- and responsibility-efficient for the data providers, although very inefficient from the researcher’s perspective. To some extent, it is also a way to protect the monopoly of the data provider in terms of use of the data. But if these are the real issues behind the restrictions on data access, there are readily-available solutions.Data access does not need to be free for all researchers. Instead, researchers can contribute to cover the costs of setting-up the infrastructure for data access using their research funds. Since there are mainly fixed costs, related to setting up the facilities for safe access (including remote connections) and to the anonymization of the data, while the marginal costs for an additional user are relatively low, data providers could use a sort of average incremental cost to establish access. This pricing structure is not new to economists, and it is similar to what happens in network industries. On top of this, since data providers are multi-product monopolies, they would obtain an advantage from allowing access to the greatest number of data sources, in order to increase the number of users .Furthermore, when contacting national statistics institutes and national central banks, we found a general high level of competence. However, in order to foster co-operation and build a truly European infrastructure for accessing micro-data, it is very important that there is also investment in developing capabilities such as language skills and economics knowledge. In this respect, EU support is crucial, especially for smaller member states, which might not be able to afford to bear the fixed costs of setting up new infrastructures and developing the necessary capabilities.The third workaround is to support multiscope, cross-country surveys, which allow researchers to gather information on a wide range of firms’ activities and performance indicators, in order to enable them to assess their contribution to overall competitiveness. The Community Innovation Surveys and the International Sourcing Surveys are interesting examples of this, although they both focus on specific aspects of competitiveness. The EFIGE survey is another example, which takes into consideration more aspects of competitiveness. However, in order for this solution to be effective, there is a need for greater harmonization and coordination. Concentrating resources on fewer surveys could be more effective in covering many aspects of competitiveness and basing results on a larger number of firms followed constantly over time. Thereby, the dynamics of firm competitiveness could also be accurately assessed. Such multiscope cross-country surveys could then be linked to administrative and registry data, and trade and foreign affiliate data, exploiting protocols for micro-data linking, as tested, for example, within the GVC project.In summary, developing national capabilities in order to better service users of micro-level data is the most cost-effective and sustainable way to generate new indicators of competitiveness. Once these permanent structures are in place, access by individual researchers to micro-level data or projects based on the distributed micro-data approach could be more feasible. At the same time, given that setting up these capabilities for all EU28 countries (ideally at sites of NSI) will take time and, in some cases, legislation, we also recommend unification and extension of corporate surveys piloted under various projects funded by the European Commission's Seventh Framework and Horizon 2020 programmes. Carefully crafted annual surveys will allow new measures of competitiveness to be constructed and of greater understanding of its dynamics even in the short term.Assessing competitiveness with European dataThere is widespread agreement that improving ‘competitiveness’ throughout Europe is at the heart of a recovery from the 2009-2010 crisis. Firms increasingly base their choices on these parameters, and the European Commission increasingly monitors external imbalances using quantitative measures of aggregate competitiveness. For these reasons there is an overall increasing effort to quantify the concept of competitiveness, and even qualitative conditions of countries’ business environments are translated into quantitative indices.However, one of the most important lessons learned during the crisis is that such informational toolbox on which policy makers base their decisions can be outdated in terms of both data sources and data analysis. There is in fact no shared definition of competitiveness, let alone a consensus on how to properly measure it across countries and over time, with a number of aggregate indicators. The toolbox is particularly outdated when it comes to tapping the potential of micro data for the analysis of competitiveness – a serious problem given that firms rather than countries will compete on global markets.The aim of the MAPCOMPETE project was to help fill this gap by providing a review of recent advances in policy oriented theoretical and empirical economic research, so as to guide the investigation of competitiveness indicators and the potential development of new ones. Importantly, for all aspects of competitiveness, a crucial issue of the project has been a thorough assessment of data availability and accessibility, as well as a critical overview of new analytical methods that become possible as new data sources become available to researchers.Research results using specific datasetsIn terms of firm-level analysis, Barba Navaretti, Bugamelli, Forlani and Ottaviano show that, due to the variability of firms’ performances (TFP and labour productivity) both within and between countries and industries, these differences may not be detected by country and sector average, which is the parameter on which most policy action is generally based. Even if some countries or sectors might be similar in terms of average productivity, the underlying efficiency distributions could be very dissimilar. The authors thus test the empirical relationship between a competitiveness indicator and different moments of the productivity distributions beyond the simple average. The authors find that asymmetry, the third moment of the distribution, is highly and significantly correlated to the competitiveness indicator, especially for large and international economies, consistently with the evidence of few exceptionally productive firms operating within each industry. The main findings are robust to different specifications, and different type of standard errors. Most importantly, the results are not affected by sample composition, i.e. asymmetry (and mean) is significantly correlated with export competiveness independently of the exclusion of countries from the estimation sample. Dispersion and especially rightwards asymmetries are therefore novel key parameters that any policy aimed at fostering competitiveness should take into account.Békés and Ottaviano explore the relationship between firm-level heterogeneity and regional competitiveness. The authors argue that measuring regional competitiveness should be also based on comparing firm performance across EU regions, rather than simply looking at average regional performance indicators. Given available data, the authors discuss a number of indicators linked to the ability of firms to access and penetrate world markets. The then identify a novel index, export per worker from a region to non-EU destinations relative to the EU average, that can constitute a ‘regional competitiveness’ index, as it captures the capacity of a region’s firms to outperform the firms of the average EU region in terms of exports, and that could be conveniently added to the regional policymaker toolbox.Firm level heterogeneity matters for the reallocation of economic activity and its implications for aggregate productivity, in particular under the lens of the labour market. Fontagné, Santoni and Tomasi show that labour 'gaps', i.e. the extent to which firms depart from an efficient use of the labor input, have been increasing over the 2000s in France, and that institutional features (distortions) have been driving this evolution, notwithstanding some evidence of sectoral shocks. Controlling for firm characteristics, the authors observe that most of the adverse evolution falls on the positive gaps, i.e. the fact that the most productive firms after 2003 have not been able to increase their labor use. The small firms are more affected by distortions but this disadvantage has increased at a slower pace over time. The authors conclude from this that more subtle micro-economic evolutions (the difficult reallocation of resources across firms within sectors) have been contributing to the deterioration of the aggregate performance of the French industry.Altomonte, Colantone and Zaurino look at new developments in competitiveness dynamics under the lens of Global Value Chains. In particular, they look at the causes of the recent trade slowdown, trying to understand whether such a slowdown is a temporary phenomenon related to the economic cycle, or it represents instead a ‘new normal’ resulting from a structural change in global activities. Specifically, they exploit a new dataset on value added trade in order to shed light on an additional cyclical driver of the slowdown, related to global value chains. In particular, they show that those components of trade that are most directly related to GVCs have experienced the largest drop over the ‘great trade collapse’ of 2009. Moreover, these components also display the slowest speed of adjustment after an income shock. Taken together, these two pieces of evidence suggest that at least part of the GVCs-induced trade slowdown is cyclical in nature, and might be re-absorbed in the coming years. In other words, global value chains might not just be playing a structural dampening effect on trade growth, due to their convergence towards a global scale equilibrium.Bas, Fontagné, Martin and Mayer present new evidence on the 'non-price' dimension of competitiveness. The authors show that, in terms of price competitiveness, direct labour costs represent just 23%, on average, of the total value of French exports and 44% when including the cost of labour for domestic intermediate consumption. Hence the non-price dimension is key to the competitiveness of the country. Indeed, they show that the loss of France's world trade share does not seem a result of poor geographic or sectoral specialisation, insufficient exporter support, under-representation of SMEs in exports or credit constraints, but, more fundamentally, is caused by an inadequate “quality/price ratio” for French products on average. Indeed, by relying on a novel indicator of non-price competitiveness, the authors show that when products are of quality, results are exceptional, as demonstrated by the luxury, aeronautical and electrical distribution goods sectors –sectors, with a flagship– and/or by brands, which appear to play a key role in France's exports.These results rely to some extent on either novel datasets available to researchers, or novel methodologies of analysis, or both. As such, they could bring further the policy debate on competitiveness in Europe, a debate which would greatly benefit from improvements in the quality of the underlying analysis.Nevertheless, the continuous development and improvement in data gathering and accessibility remains key for both the policy-making as well as the research perspective. Just to quote some examples retrieved from the previous analysis, the results of the first two paragraphs would soon loose relevance unless firm-level data across countries and regions, complete with the export dimension of firms, are not updated and made available to researchers. Insofar it is possible to rely on recent data collected across-countries within the CompNet project of the ECB, but also these data need maintenance and updating over time. As documented within the first Blueprint produced by the MAPCOMPETE project, however, official cross-country firm-level data at the European level, although existing, at the time being are practically not accessible to the average researcher, and thus have a very limited use in terms of policy-relevant analysis.In terms of reallocation, the traditional macro stream of literature dealing with the effects of centralisation of wage bargaining institutions on employment and wage outcomes has generally lead to inconclusive results because the variation in the level of bargaining used in these papers is exclusively across countries (in general the OECD indicators on Employment Protection Legislation), with no variation at the industry and firm level. Hence it becomes difficult to detect the impact of these variables separately from industry time trends, time dummies, and country dummies. When data can be brought to the firm-level dimension, instead, an entire new range of policy-relevant results emerge, as the above results clearly show. Still, despite the policy-relevant implications it could generate, working toward a higher availability of data on collective bargaining regimes, and more in general on labour market institutions at the micro-level, does not seem to be sitting high in the agenda of National Statistical Institutes.The results obtained on the impact of Global Value Chains and non-price factors on growth and competitiveness crucially rely on the presence of detailed and comparable trade data that go well beyond the average statistics on imports and exports. The continuous availability of updated and detailed Input/Output tables, as well as reliable information on traded products both in terms of quantity and values (so as to be able to infer unit export prices) remains also central in all future analysis of competitiveness.Potential Impact:When contacting national statistical institutes and national central banks, a general high level of competence was found. However, in order to foster co-operation and build a truly European infrastructure for accessing micro-data, it is very important that there is also investment in developing capabilities such as language skills and economics knowledge. In this respect, EU support is crucial, especially for smaller member states, which might not be able to afford to bear the fixed costs of setting up new infrastructures and developing the necessary capabilities. The potential impact of the project depends on whether further steps will be made in these directions. The project has identified the main data issues when addressing competitiveness challenges what EU-countries and the EU are to be faced with. In addition Mapcompete outlined different ways towards a solution which may help coping with the consequences of the recent crisis and being better prepared for the next one.The main dissemination channels were the website, conferences and a launch of a report.MAPCOMPETE website The main objectives of the website are:➢ To act as the main dissemination tool of the project ➢ To provide an information-rich and user-friendly platform to disseminate MAPCOMPETE’s research outcomes➢ To allow for sharing information and data between logged-in users (project partners) ➢ To act as a backbone for all the other communication activities of the project ➢ To build a web-ring with the partners’ websites (inter-linking)➢ To become a point of reference for research and policy-making in the fieldMAPCOMPETE Indicators Availability Map - The webtool is included into the MAPCOMPETE website and it is available freely.Through this webtool, the visitor may search and find information on the most useful indicators of competitiveness for Europe, selected by the MAPCOMPETE team. The information includes:• Description;• Rationale and problems for each indicator;• Different aggregation levels, i.e. country, sector and region;• Country-specific availability of the necessary data, and• For more advanced users, the variables to be used in the construction of the indicator.The Indicators Availability Map addresses mainly to the interested researcher / policy maker, and consists of two components to access information: a map of Europe, that allows exploring the content country by country, and a search tool that allows for a refined research (by competitiveness concept, aggregation level etc.). Through the map and the search tool, the user has the possibility to sail along the MAPCOMPETE database of competitiveness indicators and to easily retrieve all the relevant information, including the original source of the data.The MAPCOMPETE Webtool will boost the dissemination activity of the results of the MAPCOMPETE project. Thanks to its captivating and user-friendly graphic interface, it is easily usable by the general public as well as by researchers and journalists.The Webtool comes complete with a user-friendly Data Manual that explains how to use the webtool.Link to the webtool: http://mapcompete.eu/meta-webtool/ Policy Report for WP3-Blueprint#1, Deliverable 7.2 -This Blueprint picks up some of the main issues of the MAPCOMPETE project and provides an inventory and an assessment of the data related to the measurement of competitiveness in Europe. This Report, and the associated meta-database available can be a key handbook for a researcher interested in measuring competiveness, or for policymakers interested in the feasibility and in the quality of alternative competitiveness measures. This Blueprint also identifies the opportunities emerging from recent progress made in scientific research and facilitated by different data providers who increasingly make their data available to research. Finally, this inventory allows us to identify the main issues that need to be addressed by policy makers in order to improve data accessibility for the economic analysis of competitiveness in Europe.Policy Report for WP6, Blueprint#2, Deliverable 7.3: The second Blueprint presents in different chapters the results of an array of new methods / data to measure competitiveness in the EU and relevant policy implications. It presents policy options on the possible improvements in data collection related to the development of existing as well as new competitiveness indicators.Events• MAPCOMPETE Workshop: Mapping Competitiveness with European Data, on 28 November 2014, in BrusselsIt was discussed how to map competitiveness with European data, specifically the future potential of matching data in Europe within and across countries. The event counted with the important contributions of two discussants: Lauro Panella, DG GROW, European Commission and Jan Hagemejer, Central Bank of Poland.The invitation was sent out to a total number of 3078 recipients. The event was attended by 55 participants, composed of partners, EU officials, representatives of international corporations, members of academia.• MAPCOMPETE Blueprint#1 launch event, on 6 March 2015, in BrusselsPartners presented the research work of the Blueprint#1. Isabel Grilo, Head of Unit B2 “Structural reforms, competitiveness and innovation”, DG ECFIN, European Commission and Ani Todorova, Head of Unit, Unit C2 "National and Regional Accounts Production participated as discussants. The invitation was sent out to a total number of 3078 recipients. The event was attended by 35 participants, composed of partners, EU officials, representatives of international corporations, members of academia.The event was available via webstreaming at http://www.bruegel.org/nc/events/event-detail/event/506-mapping-competitiveness-with-european-data/ • Final Conference Policy session and Blueprint#2 launch event, on 29th May 2015, in BrusselsIt consisted of a full day session. All the research work conducted during the project-life was presented which can be a key handbook for a researcher interested in measuring competiveness, or for policymakers interested in the feasibility and in the quality of alternative competitiveness measures.The invitation was sent out to a total number of 3078 recipients. The event was attended by 60 participants, composed of partners, EU officials, representatives of international corporations, members of academia.During the whole duration of the project, the partners have been actively participating in CompNet of ECB by exchanging information and contributing to the achievement of objectives of both CompNet and Mapcompete. Mapcompete partners received useful contact and data accessibility information from CompNet and contributed to the different workstreams of CompNet.List of Websites:mapcompete.eu
