Combined International Trade and Investment Data “CITID” – A New Innovative and Comprehensive Data Landscape for Germany

1 Introduction

In today’s interconnected world, the increasing global integration of firms holds significant implications for the domestic economy. As nations become increasingly interconnected through trade and investment, understanding these implications on domestic economies becomes paramount. Foreign shocks have the ability to rapidly propagate through trade channels, such as supply chains, investments, and capital flows, swiftly crossing borders and causing potential issues that require informed policy decisions.^[1]

Moreover, as our world becomes progressively more digitalized and intangible products and services assume a pivotal role, the analysis of trade activities might require a simultaneous observation of trade in goods and services. The Sturgeon Report (2013) has raised awareness for the importance of cohesive statistical data sources to effectively monitor the interrelated economies through global value chains.

Therefore, the presence of a comprehensive database that accurately depicts the international integration of firms becomes all the more crucial. However, existing data sources often focus on only one aspect of international connections. In the course of a recent project funded by the German Federal Ministry of Economic Affairs and Climate Action, the collaboration between the Deutsche Bundesbank and the Federal Statistical Office Germany has bridged this gap by combining data from both institutes for research purposes for the very first time.

This allows researchers to jointly examine data pertaining to multiple dimensions of international integration. The data encompasses various facets such as trade in goods, trade in services, inward and outward foreign direct investment (FDI), and international financial and capital transactions, complemented by additional firm-level data. In the context of this article, we will label this new data landscape as the Combined International Trade and Investment Data (CITID).

Researchers will be able to access the anonymized data at the Research Data Centre’s (RDC) guest researcher workstations in a secure environment at selected RDC’s locations.^[2] The availability of CITID opens up new avenues for researchers to understand the patterns, drivers, and consequences of international integration. They can now, for instance, delve into the intricate interplay between FDI, trade flows of goods and services, capital transfers, and other critical aspects of globalization. CITID helps overcoming to rely on fragmented datasets that may only cover one aspect of international exposure, may make it challenging to develop robust theories and models to explain the multifaceted nature of international integration.

The remainder of this paper is organized as follows. Section 2 briefly presents the content of the new data landscape, i.e. the individual datasets that can be combined. Section 3 introduces the underlying methodology, in particular the matching procedure. Section 4 describes the data access and the opportunities for research. Section 5 offers some concluding remarks.

2 Data Methodology

This section provides a concise introduction to the individual datasets that can be integrated within the new data landscape. In the past, these datasets have already proven useful for various research purposes, including the field of international trade.^[3] However, until recently, it was only possible to link the datasets with other data offered by the same institution (either Deutsche Bundesbank or Federal Statistical Office of Germany). For the first time, the new CITID allows researchers to combine data from the Deutsche Bundesbank and Destatis. Currently (as of January 2024) the data that can be integrated within the new data landscape share a timeframe spanning from 2011 to 2020. It is planned to update the data annually. For more detail, please also refer to the individual data or meta data reports (see, e.g. Kruse et al. (2023)).

3 Data Linkage

As mentioned earlier, the individual data mentioned above can be matched. The primary unit of observation is at the year-firm level, although some data is available at smaller time intervals (e.g. monthly frequency for SITS). It is worth noting that there may be multiple observations at the firm-year level when the underlying data allows further differentiation. For example, the AHS-Panel data offers product-level trade information for each firm in a specific month, categorized by destination and other variables. Including an additional destination or product would consequently result in an extra observation at the firm-time level. However, it is possible (but not mandatory) to aggregate all observations from all datasets at the firm-year level to establish a common foundation. Although the individual datasets may cover different time periods, they currently (as of January 2024) all share a timeframe spanning from 2011 to 2020.^[12] It is planned to update the data annually.

3.1 Linking at the Year-Firm-Level

The Record Linkage Unit at the Deutsche Bundesbank’s Research Data and Service Center (RDSC) creates mapping tables for matching purposes using string matching and machine learning techniques. This mapping is necessary because the firm identifier (“ID”) used in the Deutsche Bundesbank’s data, known as the AWMUS-ID, differs from the SBR-ID used in Destatis data. The matching process generally yields good quality results, thanks to the utilization of state-of-the-art matching techniques. However, it is important to acknowledge that not all entities can be matched due to various reasons, such as differences in entity definitions for certain firms.^[13]

The cross table, Table 1, provides an overview of the linkage quotas between the individual data sets provided by the Deutsche Bundesbank and the Federal Statistical Office of Germany. Each cell represents the proportion of firms found in another dataset. The row labels correspond to the baseline data, which serve as a reference for the overlapping shares displayed in the columns. To illustrate, 1.1 % of the firms in the AHS data are also present in SITS, while 56.1 % of the firms in SITS can be found in the AHS data. These figures might indicate that most firms engaged in product trade do not participate in trade in services, and vice versa, most firms engaged in trade in services are also involved in trade in goods.

Table 1:

Linkage quotas BBK Destatis data.

Dataset	SBR compl.	SBR rel.	AFiD-AHS	AFiD-SBS	MiDi	SIFCT	SITS
AFiD-AHS	100.0	95.8	–	16.7	1.1	1.0	2.4
AFiD-SBS	100.0	98.1	35.5	–	1.9	1.5	3.6
MiDi	85.2	71.5	54.4	47.0	–	42.1	42.5
SIFCT	68.8	56.8	37.6	26.3	31.1	–	42.7
SITS	82.8	77.2	55.9	39.4	19.4	27.3	–

1. The SBR considers such units as “analytically relevant” that surpass the current relevance thresholds of the SBR and, as such, are relevant for the calculation of the Gross Domestic Product (GDP).

   [Correction Statement added after online publication 19th March 2024: In table 1, under the first row “Dataset” two captions “SITS” and “MiDi” were interconverted in line 3 and 5.]

Deviations from 100 % suggest that either not all matches were found during the record linkage process (e.g. as firm’s entity definitions might differ) or, more commonly, that firms are absent from a specific dataset due to non-reporting obligations. The SBR dataset encompasses all firms located in Germany with a taxable turnover exceeding 17,500 euros or with at least one employee subject to social security contributions (or at least 12 persons in minor employment).^[14] Consequently, the matching quotas in the first columns can be viewed as an indicator of matching quality, as one would expect every firm present in a particular dataset to also be part of the universe of firms. However, there are other possible reasons for not meeting this threshold.

Considering that each dataset consists of official administrative data with reporting requirements, the resulting data landscape offers a comprehensive representation of international integration in various aspects. Furthermore, it is possible, under specific conditions, to link the data with external firm data, such as data from commercial sources. Past examples of external firm data include data from “Bureau van Dijk” (e.g. ORBIS, AMADEUS), “Hoppenstedt” or “Orbis” patent data.

Figure 1 provides additional information on the number of observations available for analysis when considering different combinations of firm datasets. This visualization, based on data from the year 2020, provides a comprehensive overview of the total number of firms within these datasets. The areas where datasets overlap visually represent firms that are present in multiple datasets simultaneously.

Figure 1:

Overlap of matchable firms in CITID. This figure shows the distribution of firms across the datasets MiDi, SITS, SIFCT, SBS and AHS that can be potentially matched with each other.

The central region of the figure, where all datasets overlap, contains a total of 1976 firms. To determine the total number of observations within a single dataset, you can sum the individual areas corresponding to each dataset’s unique color. For example, the SIFCT Data comprises a total of 13,075 observations, derived by adding up the individual areas as follows: 2309 observations exclusive to SIFCT, 1263 observations found in both SIFCT and SITS but not in any other dataset, and so on.

4 Data Access and Use for Research

Access to CITID data will be granted in secure environments through designated Destatis RDC guest researcher workstations or at the Deutsche Bundesbank’s RDSC workstation.^[16] This access method is necessary to comply with legal requirements regarding the confidentiality of statistical reports while facilitating independent academic research with individual-level data. To obtain access, researchers are required to submit a research data request along with a research proposal. The feasibility of the research project, considering the research data’s suitability for addressing the proposed research questions, is carefully evaluated. Research projects must serve the public interest, and commercial endeavors are not accepted.

The new data landscape brings novelty by offering a comprehensive view of various aspects of international integration simultaneously. Previously, researchers and analysts often had to rely on fragmented datasets that only covered specific facets of international exposure, which might have posed challenges in developing robust theories and models to explain the multifaceted nature of international integration. For instance, when examining solely trade in goods data, one might assume that a shock from a foreign country would not affect domestic firms if there were no import or export relationships with that country. However, firms could still be integrated through means such as FDI, capital transfers, or trade in services. Having a complete understanding of international integration allows for more in-depth analysis of its effects. Kruse et al. (2023) illustrate the potential of the new data landscape in more detail. By connecting the AHS-Panel with MiDi data, the authors provide one such example by exploring the role of foreign direct investment firms and foreign direct investors in the trade of goods.

5 Concluding Remarks

This paper introduces the CITID, a novel and expansive data landscape that integrates firm-level data specifically related to international trade and investment. The CITID provides a comprehensive portrayal of firms’ international integration across various aspects. Researchers have access to this new landscape, offering exciting opportunities to explore and gain deeper insights into the patterns, drivers, and consequences of international integration.