What makes German manufacturing plants move locations?

The German manufacturing industry is one of the most productive industrial sectors in the European Union and the world. However, in times of multiple crises, such as the recent pandemic, climate change, geopolitical tensions, and rising social inequality, German businesses face challenges. Several countries, such as the USA, have created large investment programs (such as the Inflation Reduction Act (IRA), the CHIPS and Science Act, or the Infrastructure Investment and Jobs Act), in order to subsidize the introduction of new firm activities. In this context, firm location decisions and regional economic or location policies have attracted renewed interest.

A recent endeavor of German politics has been to maintain and even raise the competitiveness of the country’s industrial sector, culminating with the release of the so-called National Industrial Strategy 2030. The aim of this strategy is to safeguard the German economy from rising competition from the USA and Asian countries. To achieve that broad goal, it is intended to promote new technologies—such as AI—and further innovation, to regain technological sovereignty by hindering the loss of technological know-how, and, importantly, to improve conditions for firm location, such as infrastructure, taxation and other costs, and labor market conditions (BMWI 2019; Lang 2022). Against this background, investigating where manufacturing firms choose to operate and why may deliver important policy implications for supporting regions’ attractiveness to business.

This paper focuses on the relocation decisions of manufacturing plants within the German economy. Relocation refers to a plant deciding to move its location from one area to another from one time period to another. Understanding what makes plants change their location can constitute a powerful lever to increase the attractiveness of an area in order to foster the establishment of firm activity and, consequently, the provision of employment. A higher number of people being in employment and thus, able to earn an income constitutes a crucial element of societal coherence and regional development.

Aside from the large enterprises, the German economy is known for its so-called Mittelstand, which comprises small and medium-sized businesses and family businesses (BDI 2021). These companies generate more than one-third of the total turnover of companies in Germany. Moreover, almost 60% of employees work in these businesses. They have often been rooted in rural regions and are highly productive in terms of research, development, and innovation. For the analyses of location choices, it will be important to address the heterogeneity of companies (for example in terms of size, i.e., the number of people employed).

The various German regions have different industrial structures and offer different employment opportunities. For example, in 2021, the number of people employed in the German manufacturing industry was more than 7.4 million, with around 107,000 employed in the manufacturing industry in Berlin, 1.4 million in North Rhine-Westphalia, 1.5 million in Bavaria, and 352,000 in Saxony (Destatis 2023). At the NUTS-3 level, the share of people employed in overall industry with regard to the working age population in 2020 was 8.36% in Berlin, 13.75% in Munich, 60.01% in Ingolstadt (where the automobile company Audi has its headquarters and main production facility), 31.03% in Bodenseekreis (in the South–West), and 10.15% in Vorpommern-Greifswald (the large district in the very North-East of the country) (BBSR 2023). As a consequence, the various German regions face different levels of risk in terms of economic decline, international crises, or other external shocks. For that reason, at the regional level, one more or one less plant can make a substantial difference.

Relocation decisions differ from location decisions by a fine degree of nuance. Investigating location decisions—which the bulk of the past literature has done—means disentangling the factors that lead a new plant/an entrant to decide in favor of one location. Investigating relocation decisions, however, involves finding the factors that attract a plant to locate itself in a new area, thus leaving its previous area, or stay in its current location. The relative lack of relocation decision studies compared to location decision studies is most likely due to insufficient data availability.

The aim of this paper is to investigate which factors drive the relocation decisions of manufacturing plants in the German economy. The German case is special due to its historical background: more than 30 years after German reunification, regional disparities, as well as differences in firm performance between regions in the East and in the West of Germany remain. It will be of particular interest to analyze how the relocation of businesses differs between these regions.

Regional economics literature has so far mainly studied the location decisions of new businesses (see, e.g., Armington and Zoltan 2002; Figueiredo et al. 2002; Rosenthal and Strange 2003; van Oort and Atzema 2004; Arauzo-Carod 2005; Feldman et al. 2005; Stam 2007; Glaeser and Kerr 2009; Glaeser et al. 2010; Audretsch et al. 2012; Fritsch and Wyrwich 2014; Glaeser et al. 2015; Artz et al. 2016; Fritsch and Wyrwich 2018; Dong 2020; Arauzo-Carod 2021; Krenz 2023). Further, a wide range of studies on new firm activity has focused on the localization of foreign direct investment (FDI) rather than domestic localization (see, e.g., Guimaraes et al. 2000; Procher 2011). However, there is less evidence available in the literature concerning relocation decisions. The present contribution is complementary to previous analyses that focus on location decisions of new firm activity, as well as studies that primarily analyze multiple alternative choices for location decisions at home or abroad (and FDI). The question here is whether or not (incumbent, i.e., existing and not new) plants decide to relocate, or, in other words, do plants “Stay or Move?”. Which characteristics of a region make it so attractive that plants decide to change their location (which involves leaving their present location and moving to another region)? Which factors of a region make plants want to stay? Finally: which policies are needed to attract plants to move to or stay in a region?

The categories of incumbent, that is, already existing, and new plants differ from one another. Consequently, one might assume that the reasons for relocation decisions made by existing plants will be different from the reasons for new plants’ location choices: Given that a location or relocation decision is undertaken on the grounds of incomplete information about the area or region, a business that already exists might benefit from its previous business experience, which in turn may influence its relocation decision (Pellenbarg et al. 2002).

The present study investigates for the first time the relocation decisions, rather than the location decisions, of manufacturing plants in Germany using the official firm statistics and comprehensive regional statistics held on the German economy. Relocation decisions are investigated across the NUTS-3 regions, that is, the district-free cities and districts (in German kreisfreie Staedte und Kreise), of Germany.¹ Compared to the previous literature, the analysis makes use of a high-quality, large, newly constructed dataset on the German economy, with these data being derived from official regional statistics and data for the entire population of manufacturing firms in Germany. The analyses in this paper consider the relevance of pull (e.g., agglomeration economies and structural funding) and push factors (e.g., labor costs and taxes) for the relocation decisions of plants. The focus is thus on factors that make a plant choose either to move to a new location or to stay in its present location.²

For the analyses, probit regressions, as well as logit, Firth logit, and rare events logit modeling, were conducted. The results reveal that decisions to relocate are, on the one hand, driven by greater accessibility of a given region, as measured by travel times by road, and as such by the quality of road infrastructure. Specifically, a 10% decrease in travel time to the three nearest agglomeration centers leads to an increase in relocation probability of approximately 9.5% on average across all plants and 12.5% on average for small and medium-sized plants. This effect is relevant for small and medium-sized plants but not for large plants. A lower travel time makes small and medium-sized plants more likely to move toward another region. Furthermore, the quality of the workforce, as indicated by worker remuneration, positively affects the relocation choice. Again, this effect is relevant for small and medium-sized plants but not for large plants. Moreover, the potential to sell and find capacity in a market appears to be another driving factor in decisions to relocate, and this effect appears to be important for large plants.

The paper is structured as follows. Section 2 discusses the existing literature. In Sect. 3, the data are described. Section 4 presents the empirical analyses and interpretation of results. The final section offers conclusions.

There are many factors that can motivate a plant to undertake a relocation decision. The theoretical background to this issue emerges from the first writings about the location of economic activity by Marshall (1890) and the analytical framework of the New Economic Geography by Krugman (1991). As Manjon-Antolin and Arauzo-Carod (2011) describe, relocation decisions can be motivated by the same set of determinants as location decisions, the effect strength of the variables, however, can be expected to differ. Neoclassical, cost-related factors can be expected to determine relocation decisions, as a plant will make an optimal decision about a region to locate in and to maximize its profits. These factors comprise labor costs, market size, and agglomeration externalities. In addition to this, institutional factors can be expected to play a role: plants operate within regional systems and in an environment of regulations and laws, such that taxes, regional funding, and regional infrastructure are important factors for consideration (see Arauzo-Carod et al. 2010).

Among the existing studies on the relocation decisions of businesses, for the USA, Lee (2006) uses a probit model to investigate whether plants relocate and which factors drive that relocation. The author finds evidence that plants move out of regions in which industry is heavily concentrated to regions where it is not and describes these new regions as becoming new centers for industry. No evidence is found for the impact of taxes or human capital. Lee (2008) finds only weak effects of state development policies and incentive programs on the relocation of plants in the USA. Conroy et al. (2016) investigate relocations of firms, considering separately the effects of various factors on high-\(R \& D\) and low-\(R \& D\) firms. They find that the former tend to leave states that have a larger share of US manufacturing output in favor of states with a smaller share, are attracted to states that tax property more (which the authors interpret as pointing to the importance of amenities funded by property taxes), and make decisions relatively unaffected by transport infrastructure. The latter firms they find to be attracted to states with lower personal income taxes and states that spend more on higher education and transportation infrastructure. Conroy et al. (2017) find a link between higher wages and positive relocation decisions taken by firms in the US; they explain that this might be due to the decline in low-skilled manufacturing and the rise of high-tech manufacturing and reason that higher wages might indicate that highly-skilled labor is employed. Pan et al. (2020) analyze the relocation choices of firms in the USA and find that taxes and subsidies are important influential factors. Manjon-Antolin and Arauzo-Carod (2011) investigate the determinants of location and relocation decisions made by establishments across the constituent municipalities of Catalunya. Using a count data model, in their preferred specification, they find that relocation is motivated by urbanization and disurbanization economies, as well as human capital but not by infrastructure, location economies, or metropolitan effects. Holl (2004) investigates the relocation patterns of plants in Portugal, finding that greater market accessibility and provision of inter-regional motorways play an important role. Kronenberg (2013) finds that relocating firms in the Netherlands are generally attracted to densely populated municipalities with high wage levels. Weterings and Knoben (2013) find for the Netherlands that relocations within municipalities mainly depend on growth and the need for more space, while in terms of relocations across wider regions, firms stay in locations due to higher concentrations of other firms and greater urbanization and leave locations due to a higher share of innovative firms. The authors explain this as the result of congestion or competition. Knoben (2011) finds that a high level of agglomeration has a negative effect on the relocation of firms. Morkute and Koster (2018) investigate the importance of human capital for firm relocations in the Netherlands and find that human capital influences firms to some degree in deciding to stay but not in deciding to move. Investigating the relocation choices of firms in Poland, Rossi and Dej (2020) find that urbanization economies, availability of highly-skilled workers (indicated by higher pay), and amenities (public goods provision) appear to play an important role. Investigating the relocation choices of Norwegian firms, Nilsen et al. (2020) find that agglomeration, urbanization, and highly-skilled labor are variables that play an important role. Bodenmann and Axhausen (2012) find for Switzerland that relocation choices mainly depend on tax burdens and, to a lesser degree, on the accessibility of regions. Yi (2018) uses a nested logit modeling approach and finds that establishments in South Korea are attracted to relocate due to localization economies in earlier stages of the business life cycle, but not in later stages. Hong (2014) observes that Korean plants base relocation decisions on agglomeration economies. Brouwer et al. (2004) investigate relocation through a survey of 21 countries and focus on large firms with more than 200 employees. They find that larger and older firms are less likely to relocate, while firms serving larger markets are more likely to relocate.

In sum, the literature provides mixed evidence across a range of countries for the relevance of various factors that play a role in the relocation choices of firms. There has been evidence for the influence of urbanization and disurbanization economies (Manjon-Antolin and Arauzo-Carod 2011), (road) infrastructure (Holl 2004; Bodenmann and Axhausen 2012), and market size effects (Brouwer et al. 2004; Lee 2006; Conroy et al. 2016). There is less consensus as to the relevance of localization externalities (see, e.g., Manjon-Antolin and Arauzo-Carod 2011; Knoben 2011; Kronenberg 2013; Hong 2014; Yi 2018), human capital (Lee 2006; Manjon-Antolin and Arauzo-Carod 2011; Kronenberg 2013; Conroy et al. 2017; Morkute and Koster 2018; Nilsen et al. 2020; Rossi and Dej 2020), and taxes (Lee 2006, 2008; Bodenmann and Axhausen 2012; Conroy et al. 2016; Pan et al. 2020) for the relocation choices of firms. The impact of structural funding on firm relocation choices is not yet well explored: the previous literature has focused mainly on the macroeconomic impacts of such funding, such as regions’ GDP or regional employment, rather than firm relocation choices, which is likely to be due to lower data availability.

A range of variables was extracted from the data sources: Table 1 provides a detailed description of these variables. The size of the respective regional population and the number of manufacturing plants per capita in a region were used to model agglomeration economies (similar to Procher 2011). The bulk of the past literature employed population measures in estimating location or relocation choices (see, e.g., Guimaraes et al. 2000; Figueiredo et al. 2002; Manjon-Antolin and Arauzo-Carod 2011; Procher 2011; Kronenberg 2013). Many studies also applied specific measures for localization or urbanization economies. Guimaraes et al. (2000), Figueiredo et al. (2002), and Manjon-Antolin and Arauzo-Carod (2011), for example, used employment shares in either manufacturing or services to model agglomeration effects. Regional GDP per capita and GDP growth served as proxies for market size effects (see, e.g., Procher 2011). Labor costs, measured here as gross wages in manufacturing per employee, reflect, on the one hand, the costs of production that a plant has to pay and, on the other hand, the quality of the regional workforce (because highly-skilled workers are expected to be paid higher wages; see, e.g., Smith and Florida 1994). Tax revenues, travel times by road, and structural funding are institutional factors that model the provision of public goods, regional infrastructure, and regional policy effects.

Taxes were measured as regional business tax revenue per capita. Some studies found a positive effect of taxes, potentially indicating that more public investment is undertaken from the higher tax income; other studies found no significant effect, at all (see, e.g., Carlton 1983; Artz et al. 2016). Taxation of firms in Germany involves several elements. One of these is a business tax on firms that varies between communities (the German Gemeinden). Information on this was available from the INKAR database and was used for the regression analyses. In INKAR, tax revenue is summed up at the level of the NUTS-3 region and expressed as per capita figure.³

Average travel time in minutes by road to the three nearest agglomeration centers⁴ is a variable that is directly recorded in the INKAR database. It measures the accessibility of a region and the quality of its road infrastructure. According to the precise BBSR definition, this variable captures the average travel time in minutes by car to the closest 3 of 36 agglomeration centers in Germany and the neighboring countries. The BBSR conducts regular accessibility analyses (through its so-called Erreichbarkeitsmodell, see BBSR 2019), showing, for example, that the most north-easterly German regions are peripheral in terms of accessibility, whereas the West German regions of the \(\textit{Ruhrgebiet}\) are highly accessible. For its analyses, the BBSR uses georeferenced information on a set of 662,000 routes and 518,000 nodes for Germany and calculates travel times and distances using ArcGIS. For road traffic, the BBSR uses information on types of streets (Autobahn (highway), Landstrasse (country road), etc.), length of streets, velocity, and travel time. The shortest travel times and/or distances are computed and stored in time or distance matrices.

Regional structural funding is measured by the long-term spending on regional infrastructure over the past ten years by the German government’s ”Joint Task on Regional Economic Development” (Gemeinschaftsaufgabe Verbesserung der Regionalen Wirtschaftsstruktur, abbr. GRW) complemented by co-funding from the European Regional Development Fund (ERDF). This variable is directly available within the INKAR data. A zero amount of funding in the dataset was replaced with a value of one euro in order not to lose many observations within the regressions.⁵ GRW funding is a means to support investment in business and infrastructure for economic purposes in the German economy and to help even out regional disparities. This funding comes in addition to basic funding undertaken by the firm itself.

Data from the official firm statistics maintained by the German Federal Statistical Office and the Statistical Offices of the Federal States were merged. Because German law mandates that all enterprises and plants must report to the official statistics, this is the most comprehensive firm dataset that exists for Germany. Data were accessed for the years 2013 and 2012. Specifically, data for manufacturing plants (Amtliche Firmendaten fuer Deutschland (AFiD), Industriebetriebe, see RDC 2023) were used, covering plants that employ at least 20 employees. This cut-off is used in the AFiD data itself; its aim is to protect the sensitive data of micro-sized firms. Importantly, however, above that cut-off, the data cover the whole population of plants, rather than a random draw from it. Moreover, focusing on firms of this size means that any decision regarding relocation is less likely to be based on an individual decision by the owner or manager. This is important given that the dataset does not contain any manager- or owner-related micro-level variables. This could be a potential avenue for future research addressing individual-level influences on relocation decisions.

In the following section, separate analyses are presented for the groups of all plants, small and medium-sized plants, and large plants. Small and medium-sized plants are defined for the purpose of these analyses to be those with fewer than 250 employees and total annual revenues of up to 50 million euros, while large plants are those with at least 250 employees and total annual revenues of more than 50 million euros. This follows the definition of the European Commission. The data include information on each plant’s location. It is this information that allows us to model the relocation decisions of manufacturing plants and investigate how relocation choices are affected by a location’s characteristics. From the year 2012 to the year 2013, among the class of plants with at least 20 employees, 73 plants relocated (\(Y=1\)) and 43,205 plants stayed in their existing location (i.e., a NUTS-3 region; \(Y=0\)); of the sample of small and medium-sized plants, 68 relocated and 37,571 stayed in their existing location, and of the sample of large plants, 5 relocated and 5634 stayed.⁶ Regional reclassifications (merging or division of regions or changes in regional identifiers) were taken care of for the dataset.

The present work investigates the relocation decisions of manufacturing plants across the NUTS-3 regions of Germany. It complements previous analyses that focus on new firm activity as well as studies that focus on multiple alternative choices in location decisions by asking the question, “Do plants decide to relocate: yes or no?”, or more simply “Stay or Move?”. It exploits for the first time the rich information provided by the official German regional statistics as well as the official German firm statistics that are maintained by the German Federal Statistical Office and the Statistical Offices of the Federal States. Based on these data, a novel and comprehensive dataset was constructed to investigate the relocation decisions made by German manufacturing plants.

The results reveal that relocation decisions are driven by greater accessibility of regions, as measured by travel times by road and, as such, by the quality of regional road infrastructure. An important relationship is found between the improvement of the accessibility of a region as represented by the regional road infrastructure and travel time and the relocation choice of a plant. Specifically, a 10% decrease in travel time to the three nearest agglomeration centers leads to an increase in relocation probability of approximately 9.5% on average across all plants and 12.5% on average for small and medium-sized plants. This effect is relevant for small and medium-sized plants but not for large plants. A lower travel time makes small and medium-sized plants more likely to move toward another region. Further, the quality of the workforce, as indicated by worker remuneration, positively affects the relocation choice. Again, this effect is relevant for small and medium-sized plants but not for large plants. Moreover, the potential to sell and find capacity in a market appears to be another driving factor for the relocation decision, and this effect appears to be important for large plants.

Through the use of the present regression framework, the computation and interpretation of the effects of these variables in terms of relocation probabilities were possible. The effects appear to be economically and statistically significant. However, a limitation of this research is that the dataset does not provide any manager- or owner-related individual-level variables. Addressing the role of those factors in relocation decisions would be an interesting avenue for future research.

In terms of policy implications, the results of this study provide evidence that policies that endeavor to support firm activities at a regional level need to address regional road infrastructure. Improving road infrastructure and making regions attractive to highly skilled workers appear to be crucial elements that lead to plants deciding to leave the region in which they are located and move to a new region.