Regional manufacturing in the Czech Republic

ABSTRACT

The New Economic Geography (NEG)’s agglomerative forces strongly determined the location of firms, in spite of policy intervention blatantly activist intended to reverberate the geography of production (during the communist regime). Recent models of geography and trade recognize a bijective interaction between NEG forces and comparative advantages, they fully explain productive specialization. Our interpretation of the Czech manufacturing is in accordance with the NEG and its recent integration with the comparative advantages stream. In the panel data float up some invariant characteristics enhancing the attractiveness of traditional industrialized territories. The estimates embody the population density and gravitational and agglomeration forces. The time-invariant characteristics are present in the most industrial regions, influencing the localization of manufacturing activity in these spots. In spite of the forceful influence of path dependence, the very manufacturing cores underwent a process of restructuration, adapting the old-fashioned specialization pattern toward a more coherent production tailored to local assets, endowments and technologies.

KEYWORDS:

• New Economic Geography
• Comparative advantage
• Industrial agglomeration
• Production specialization
• Czech Republic

1. Introduction

Amid the New Economic Geography (NEG) framework, the space, the proximity and the geography came along to explain the manufacturing location and the agglomeration of firms and employment. Mainly in the Krugman’s work, the trade Theory and the Geography were reconciled for explaining the stubborn fact of the spatial concentration of the production and the economic activity (Royal Swedish Academy of Sciences, 2008).

Accordingly, Krugman (1993) and Krugman (1999) showed a key insight commonly overlooked in the traditional Trade Theory: the joint research approach in the analysis of the inter-regional and international trade. The set of tenets in both fields can be the same, but the spatial level of analysis can be zoomed in or zoomed out. The requirement for a such research strategy is to adjust the assumptions and the incorporation of footloose factors (mainly capital), and to allow the intervention of models based on increasing returns, with firms performing in a microeconomic context of monopolistic competition (Krugman, 1991).

The purpose of this article is to interpret the process of regional localization and specialization in the Czech manufacturing industry, applying a theoretical and empirical approach. We explore the traits in the Czech manufacturing, emphasizing the interactions set in motion by the forces that stem from the geographical proximity, combined with the typical variables pointed out by the NEG authors. However, it also accounts for historic events and by political turmoil and crises.

We purport to identify the determinants of regional specialization (NUTS 3), in an analysis relying on a clear theoretical background able to account for the formation of regional productive profiles, and the definition of patterns of production. The theoretical paradigm proposed by the NEG excels as a relevant interpretation, due to the relevance of transport costs in a well-connected country, the influence irradiated from the urban cores and the traditional profile of the Czech economy. In this vein, we highlight the role of core spots of agglomeration whose economic dynamic is transmitted around, being able to pull other neighbor geographical units. The combination of proximity and manufacturing relevance, forge geographical common areas of agglomeration and spatial continuities, where the employment and the investment coincide.

With such purpose, we tackled an empirical analysis of possible determinants of manufacturing specialization, and of the forceful economic interactions stemming from the expansion of the manufacturing activity. In doing so, the structure of the article is arranged as follows. Firstly, we overhaul the previous theoretical developments, then we set forth the empirical and econometric strategies, and finally provide the conclusions.

2. Overview of Literature: Bridging the gap between the Classical world and the Geography

From the Ricardo times, the departure in the research paths between international and intra-national trade pertains to a stake for theoretical coherence. In fact, the classical assumption of immobility of capital cannot be applied within the limits of any country, but fits well in an international context (Blaug, 1990). From then, the development of the traditional trade theory has been underpinned by strong and almost incontrovertible assumptions: null transports cost, immobility of factors and perfect competition. And in such context, there are no serious analytical implications, provided that the free circulation of commodities can be understood as a substitute for the factor mobility (Mundell, 1957).

In contrast, since the appearance of the NEG and the New Trade Theory (NTT), the analysis in both fields has moved forward dramatically and unleashed a torrent of research. The point of departure was the blatant empirical fact, amid the trade exchanges, of the intra-industry trade, claiming a new theoretical framework based on monopolistic competition and under the assumption of increasing returns (Grubel, 1967). Krugman (2008) and Krugman (1991) showcase a compelling overview of a long path starting from the exclusive focus on international trade, to a final destination ending up in the realm of Economic Geography.

The previous discussion set a double approach for interpreting a myriad of compelling issues: trade, geography, specialization and patterns of production. Thus, a dual alternative is defined between the Comparative Advantage and cost differentials (including the Heckscher–Ohlin’s factor intensity) approaches, and the NEG–NTT.

However, the most recent research in Geography and Trade wisely concluded that there is no dominance of any of the two approaches. Ricci (1997), Ricci (1999), Pflüger and Tabuchi (2016) and Brakmann et al. (2009), inter alia, deployed a set of models of trade and specialization blending the powerful influences of Comparative Advantages and NEG forces.

This theoretical convergence has been pointed out since the Bertil Ohlin times. In this vein, some economists have pleaded for an integration in the determinants of regional and international specialization, and occasionally, Krugman assumed that this joint research approach would be a crucial point of departure for the research path (Krugman, 1993, 1999).

In contrast, the classical tradition was concerned in diverting the regional analysis from the international specialization, a disruption which relied on strong theoretical reasons. Under the assumption of international immobility of capital, a different theory must be developed in the realm of the regional specialization, in which the factorial mobility is the rule (Blaug, 1990).

In the Ricardian world, the international exchanges make sense only if the partners exploit their comparative advantages based on costs and productivity; therefore, the trade is the appropriate scenario to demonstrate the differences in terms of productive conditions. Subsequently, exchanges between similarly endowed countries never take place, and to the extent that each country took advantages of the differences in costs and technologies, a full specialization landscape emerges, rendering an array of countries producing commodities resulting from the more cost-efficient process, with exchanges based on inter-industry trade (Ricci, 1997).

The standard Heckscher–Ohlin model relies on the factor endowments and the intensive exploitation of them, assuming some mobility restrictions in the factors, but conferring unrestricted free mobility to commodities exchanged. Nonetheless, an imperceptible implicit trade of factors is assumed when the trade of commodities is blatantly eased (Ottaviano & Puga, 1997). In this kind of assumptions, the factor mobility restrictions become irrelevant and the openness in trade can lead to an equalization in the price of factors, without a noticeable damage to the welfare of workers. In such terms, the Heckscher–Ohlin’s celebrated theorem of factor prize equalization emerges (Krugman, 1993). This paradigm can be better understood in light of the interaction between factor abundance and factor intensity (Krugman, 1999).

In the Ricardian models of comparative advantages or in the Heckscher–Ohlin’s model based on endowments differences, the sources of trade are the prominent differences between trade partners, and in consequence, in the absence of such dissimilarities the trade will never take place.

Extrapolating this kind of paradigm of constant returns into the spatial dimension, the result is clear: a final landscape showing a situation of ‘Backyard capitalism’ with no room for agglomeration (Krugman, 1993; Ottaviano & Puga, 1997). Thus, in the absence of the increasing returns, with null transport cost and with dismissible differences between regions and countries, each space will tend to be completely autarkic and engaged in the production of a set of commodities intended for auto-consumption.

However, after the World War II, the predominance of intra-industry trade posed stunning challenges to the Ricardian view of international trade. The upswing of models of the NTT and the NEG aimed to provide a theoretical structure for explaining these new empirical facts. In the new models, the dominance of monopolistic competition assumptions enables the firms to exploit enhanced markets along with the economic integration, based on the production of a set of varieties of similar goods with diverse degrees of substitutability.

Conversely, the irruption of the microeconomic context of monopolistic competition fully incorporated the assumption of increasing returns, in which each firm takes over the production of a proper productive variety, differentiating it from the rivals and driving the specialization pattern toward the production of different varieties in the region or the country. To the extent that everywhere the commodities are imperfect substitutes, the countries can draw benefits from trade with no need to reveal any comparative advantage (Ricci, 1997).

The NTT and NEG models skipped the Ricardian world in which the comparative advantage emerged as the ultimate tenet explaining the trade and the specialization, and embraced a new paradigm based on a different view of trade. As Krugman (1999) points out, countries have the ability to concentrate industries through the exploitation of increasing returns and the home market effect, which is defined by Davis and Weinstein (1996), and provides the sufficiently large scales for defining the specialization pattern of trade.

This kind of models is encompassed in a market context of monopolistic competition with firms producing specific varieties of the same good, in the presence of transport costs with the inclusion of other additional parameters, namely: the love for variety and the elasticity of substitution across industrial goods. The theoretical convenience of this microeconomic structure is outright. Even when there are several firms producing direct substitutes in the market, the individual monopolist can locally exert its market power, because the distance contributes to differentiate its product, and the market power is reinforced by its preponderance on the space (Krugman, 1993).

The disruptive contribution of Krugman and Fujita, inter alia, and the Balassa’s empirical observation of the predominance of intra-industry, cleared the way for a torrent of subsequent research in the theoretical streams of the NTT and the NEG. The first generation of models relied forcefully on the basic parameters determining the localization of manufacturing, namely: transport Costs (τ), economies of scale (F/S) and footloose workers (μ). Therefore, the agglomeration forces can reinvigorate the dominant role of the cores of agglomeration: if τ becomes lower, F/S is reinforced and industrial workers grasp a higher share in the labor force. In such terms, according to a very light model outlined in Krugman (2008), the conditions driving to a concentration process are given by:

(1) $F/S>\tau \ast \left(1-\mu \right)/2$(1)

In general terms, the NEG models predict a pervasive process of concentration resulting in an unambiguous core–periphery scheme, with the main bulk of manufacturing activity set up in one region. A powerful cumulative causation mechanism drives the process into a strong differentiation in a core manufacturing region, even when both regions had originally the same conditions,

Upon enforcing the tenets of a variety of models focused on explaining the location in terms of F/S, τ and μ, a set of new models strived to bridge the gap between this approach and the traditional Ricardian explanation of the patterns of production and specialization. Some theoretical endeavors introduced the interaction between such forces (Brakmann et al., 2009; Pflüger & Tabuchi, 2016; Ricci, 1997, 1999). The recognition of a double influence coming from each theoretical source, enables a proper frame for interpreting the spatial distribution of production. In the end, the recognition of the biunivocal influence of the endowments and the increasing returns on the trade patterns, allow to solve the challenges posed by the dispersion of productive factors (Krugman, 1999).

Ricci (1997) proposed a model integrating Ricardian differences in cost and productivity with typical parameters of geography: increasing returns, transport cost and monopolistic competition. As a consequence of these location forces, agglomeration could predominate amid firms relying on comparative advantages or geographic agglomerative forces in response to the values that parameters can assume. Obviously, the trade that emerge from the model is a thread of intra-industry and inter-industry commercial flows, depending on whether the increasing returns or constant returns sectors predominate.

Ricci (1997) assumes two locations, each enjoying a particular advantage for producing one commodity. In the presence of increasing returns, the monopolist is induced to produce one variety in one single spot.

Once the trade costs are included in the analysis, the two markets are segmented to the extent that exports to the other market involve an increase in the price in the destination market. In this condition, the parameter of substitution is activated, and the goods produced locally have an advantage to local consumers; such effect will be augmented with the substitution elasticity.

In such conditions, an unexpected result can be achieved under the effects of comparative advantages, increasing returns and trade costs. Any country can host production of a commodity in which it has a disadvantage, if the trade costs are sufficiently high, thus affecting the price of the commodity originated in the other country. This scenario is possible only if the local gap in productivity is not excessive, as to reinforce the genuine advantage of the alternative location. Conversely, a more Ricardian full specialization will be propitiated in a scenario with abysmal differences in the genuine comparative advantage, with decreasing trade costs, and if the scale economies are sufficiently strong.

Such a sequence of causalities can make up combined models incorporating Ricardian tenets, but driving the local production of the commodity affected by a comparative disadvantage.

In Pflüger and Tabuchi (2016), a mix of Ricardian and increasing returns appears with a definitive intervention of the trade costs, but the model assumes a turn in the sense that final goods are produced under constant returns, and the intermediates are submitted to increasing returns, operating under monopolistic competition. The model assumes that only one productive factor is used, that the production can take place in two locations, and that the output is a continuum of consumer goods and services.

In the first case, the model tackles a problem of international trade setting up the existence of two partners with fair differences in size, and assuming factorial immobility. The outcome leads to a differential in wages in both economies shaped by agglomeration economies and the comparative advantage, also intervening in the transport cost across intermediate and final goods. If the dynamic of the model leads to a reduction in the trade cost of final goods, the comparative advantage is bolstered. Conversely, if the reduction in transport cost affects the intermediate sectors, the increasing returns forces lead the agglomeration. For Pflüger and Tabuchi (2016), the second case involves one turn into geography and assumes footloose factors. The conclusion establishes that the model leads to a partial agglomeration, inasmuch as the trade costs undergo a drop in the sector subject to increasing returns.

The definition of this kind of market structure as a frame for encompassing the location models has startling implication also for economic policy. One of them is the suggestion that economic integration between either countries or regions has no ostensible effects on employment, and participants in exchanges can benefit by the expansion of markets, under the condition of a stringent differentiation of goods. Inasmuch as each partner exploits the advantages of a specialization relying on economies of scale, each one can trade a manufacturing good clearly identifiable, which is slightly differentiated regarding the incoming production coming from the trade partner.

3. Analysis of the Problem

Comparatively, the reliance on Ricardian tenets drives to a wage reduction in the regions with a large endowment of labor, while in the opposite stream, the assumption of increasing returns leads to an upward pressure on wages, inasmuch as the flow of firms increases the demand of labor (Pflüger & Tabuchi, 2016). It makes sense because the Ricardian category assumes factorial immobility and conversely, the New Trade inspiration allows footloose workers.

On the other hand, the assumption of increasing returns conveys forceful implications for the outcomes of the models. Unlike neoclassical models that predict a trend to convergence in the regional GDP per capita, the NEG core–periphery models predict that the trends to agglomeration and spatial concentration drive to a divergent process in which the firms are enticed by traditional spots, and finally the consolidated manufacturing regions swallow the industrial activity. In this condition, a process of catching up is farther from being accomplished (Krugman, 1993)

Assuming the influence of the economies of scale and mobility of factors, the result will be the emergence of units of production taking advantage of the specialization, by increasing returns and shipping the production to neighboring places. This exploitation of economies of scale will unleash a process of concentration of production, that will attract the footloose factors of production (mainly μ), that contribute to enhance the market size and the accumulative process of agglomeration.

Conversely, in the presence of constant returns and an even spatial distribution of resources, the final result is a perfect dispersion of production, in which all agents end up producing all the set of disposable goods domestically, with no necessity to trading with surrounding unities, in a typical landscape of ‘backyard capitalism’ (Krugman, 1991, Krugman, 1993; Ottaviano & Puga, 1997).

The irruption of transport cost and factor mobility introduces the geography into the analysis of location and specialization. The classical models of international trade had assumed factorial immobility and free trade of commodities. Mundell (1957) proposed to embed this assumption in a more global hypothesis, namely: the commodity transactions are substitute of factor movements. In such terms, with either mobility of commodities or mobility of factors, the outcome is equalization of price of factors or equalization in the price of commodities, respectively.

Any interregional trade originated either in the factor intensity or in the economies of scale can be interpreted as substitute of factor mobility. The implicit exchange of factors compressed in trade involves the exchanges of abundant resources by scarce, and the exploitation of greater scales of production (Krugman, 1991).

Unlike those assertions, the original NEG models lead to the emphasis that the commodity and factor mobility complement each other, but drive an opposite conclusion: the mobility of factors can end up in divergences in factor prices and even, in commodity prices (Krugman, 1999).

The NEG models based on scale economies under monopolistic competition conditions, can blatantly predict an economic integration without relevant harmful consequences in terms of employment, because each partner can specialize in a manufacturing variety, enjoying the widening of the market, thanks to the economic integration (Krugman, 1993, 2008), with prefigurated ideas present in Grubel (1967) as well. In another work, Krugman pointed out that under the influence of the increasing returns paradigm and the monopolistic competition framework, the observable dispersion of manufacturing activity all over the world, does not seemingly involve a reduction in the trade flows (Krugman, 1999) (a trend in the direction of shaping more even economic structures).

In the Krugman’s exposition of Ohlin’s work there is an interesting argument (Krugman, 1999). Already in 1933 Ohlin recognized the combination of comparative advantages and economies of scale as sources of trade; therefore, the trade performs a mitigating role in the inconveniences carried out either by the uneven distribution of endowments, or the proper indivisibilities of the economies of scale. In both scenarios, the trade substitutes the factor mobility.

According to Krugman (1999), the work of Ohlin exposed the idea that international trade and regional location can be treated as unified fields of analysis, already recognizing the importance of increasing returns in defining the specialization patterns, and enthroning them as a subsidiary explanation for international trade (Krugman, 1999). Several decades later, Krugman pleaded for an integration of the fields of International Trade and Location trade (Krugman, 1991, Krugman, 1993, 2008). Perhaps, the key elements of this substantive integration rely on the degree of factor mobility and the readiness of commodities during the exchange.

As Krugman (1991) asserts, the economic process and the influx of market forces are not constrained by arbitrary borders, which define administrative territories, in spite of the potential influence that can be conveyed by the national policies.

The underlying theoretical task is to explain the pattern of trade of nations and the production profile in the regions. The answer needed an explanation of productive patterns in which the production required the exploitation of scale economies that reinvigorated the productive specialization, in a process of circularity reinforced by the market size. In the modern economic theory, any sort of indivisibilities and scale economies lay in the basis of some efficient processes that require minimal scales of production. Due to the presence of those forces, the spatial cluster of firms launches a continuous process of agglomeration.

4. Applied Methods and Econometric Strategy

A basic exercise of sorting the number of regional employees hired in each sector, can provide insights into the importance of particular industries at the territorial level (Kim, 1995). At a glance, the regional breakdown of manufacturing in the Czech Republic exhibits a strongly specialized pattern. Table 1 depicts the sectoral share in the regional manufacturing employment using data at a NUTS 3 level. In several cases, the setup of factories along the enduring consolidation of the manufacturing relied on the access to natural resources, according to the purely localized Marshallian forces (Pflüger & Tabuchi, 2016). The steel plants in Ostrava and the wood industry in the southwest and southeast represent clear examples of it.

Table 1. Czech Republic. Regional Specialization of Manufacturing 2021. (Sectorial Distribution of Employment).

	Prague		Central Bohemian Region		South Bohemian Region	Plzeň Region	Karlovy Vary Region	Ústí nad Labem Region	Liberec Region
Total	64.426		110.502		46.155	55.278	14.096	50.574	41.141
Manufacture of food products	13.88%		6.60%		10.35%	1.90%	-	4.12%	-
Manufacture of beverages	2.48%		0.61%		-	4.97%	-	-	-
Manufacture of textiles	-		-		-	-	6.82%	2.00%	2.19%
Manufacture of wearing apparel	0.86%		-		1.09%	-	-	-	-
Manufacture of leather and related products	-		-		-	-	-	-	-
Manufacture of wood and of products of wood and cork, except furniture; manufacture of articles of straw and plaiting materials	-		0.90%		2.58%	3.65%	-	-	-
Manufacture of paper and paper products	3.28%		1.41%		1.47%	-	-	3.95%	4.79%
Printing and reproduction of recorded media	2,26%		1,95%		-	-	-	-	-
Manufacture of chemicals and chemical products	2.33%		2.82%		2.27%	-	-	10.27%	-
Manufacture of basic pharmaceutical products and pharmaceutical preparations	3.12%		0.90%		-	-	-	-	-
Manufacture of rubber and plastic products	5.75%		3.96%		5.57%	8.62%	10.38%	7.19%	9.31%
Manufacture of other non-metallic mineral products	10.27%		3.37%		4.16%	5.02%	15.91%	10.64%	6.75%
Manufacture of basic metals	2.04%		2.42%		-	-	-	4.35%	2.32%
Manufacture of fabricated metal products, except machinery and equipment	3.94%		4.18%		12.67%	11.71%	19.33%	8.54%	5.83%
Manufacture of computer, electronic and optical products	3.81%		-		3.03%	-	-	1.54%	-
Manufacture of electrical equipment	20.23%		2.29%		7.80%	14.69%	9.90%	10.37%	4.69%
Manufacture of machinery and equipment n.e.c.	8.65%		7.45%		13.28%	11.94%	-	9.79%	14,94%
Manufacture of motor vehicles, trailers and semi-trailers	5.20%		49.99%		24.26%	21.76%	12.94%	17.08%	41.91%
Manufacture of other transport equipment	2.16%		-		-	6.06%	-	-	-
Manufacture of furniture	-		-		-	-!	-	-	-
Other manufacturing	1.10%		4.39%		6.37%	3.29%	-	2.11%	1.19%
Repair and installation of machinery and equipment	7.59%		1.95%		-	-	-	5.41%	-
Czech Republic. Regional Specialization of Manufacturing 2021. (Sectorial Distribution of Employment)
		Hradec Králové Region	Pardubice Region	Vysočina Region		South Moravian Region	Olomouc Region	Zlín Region	Moravian-Silesian Region
Total		45.866	47.655	46.395		69.761	44.168	51.141	98.213
Manufacture of food products		5.19%	5.41%	9.16%		9.55%	8.86%	7.86%	4.26%
Manufacture of beverages		-	-	-		0.42%	1.45%	-	-
Manufacture of textiles		8.64%	3.30%	-		3.56%	-	-	0.98%
Manufacture of wearing apparel		-	-	2.32%		-	2.27%	-	-
Manufacture of leather and related products		-	-	-		-	-	0.85%	-
Manufacture of wood and of products of wood and cork, except furniture; manufacture of articles of straw and plaiting materials		1.06%	-	5.77%		1.16%	1.73%	1.17%	-
Manufacture of paper and paper products		-	-	1.71%		-	2.18%	-	2.59%
Printing and reproduction of recorded media		-	-	-		1.38%	-	-	-
Manufacture of chemicals and chemical products		1.48%	5.63%	-		1.82%	-	6.99%	1.08%
Manufacture of basic pharmaceutical products and pharmaceutical preparations		-	-	-		5.92%	-	-	3.16%
Manufacture of rubber and plastic products		9.43%	6.50%	7.97%		6.63%	7.91%	23.78%	3.52%
Manufacture of other non-metallic mineral products		3.31%	-	-		5.74%	2.85%	-	1.14%
Manufacture of basic metals		-	-	-		3.20%	3.33%	1.76%	19.83%
Manufacture of fabricated metal products, except machinery and equipment		7.35%	7.64%	12.99%		8.84%	11.54%	21.65%	7.79%
Manufacture of computer, electronic and optical products		-	12.74%	-		5.47%	6.59%	4.76%	0.74%
Manufacture of electrical equipment		20.62%	10.28%	10.13%		7.90%	19.16%	7.82%	7.55%
Manufacture of machinery and equipment n.e.c.		7.22%	13.54%	15.97%		19.61%	10.12%	-	9.85%
Manufacture of motor vehicles, trailers and semi-trailers		11.78%	18.54%	15.63%		8.25%	5.15%	5.89%	25.57%
Manufacture of other transport equipment		-	-	-		1.86%	4.09%	-	5.08%
Manufacture of furniture		-	-	1.31%		2.90%	-	1.80%	-
Other manufacturing		6.13%	-	3.91%		3.05%	2.71%	1.22%	-
Repair and installation of machinery and equipment		1.27%	-	3.50%		2.06%	5.28%	1.96%	2.38%

*Due to statistical confidentiality, there are no published data for tobacco products and coke and refined petroleum products. Summing up all sector participations does not equal 100%. Source: CSO.

When zooming in the analysis of specialization, some sectorial and geographical facts stem. Some evidences exist about the connection between the basic metal production and the auto industry in the Moravian-Silesian Region (Sadilek, 2017). On the other hand, the printing and reproducing media sector discloses its typically urban character. More basic and simple activities such as food production and beverages, demonstrate a rather ubiquitous spatial pattern (Kim, 1998).

The location of the emblematic automotive industry is forthrightly related to the specialized inputs and vertical integration as trumpeted in Pflüger and Tabuchi (2016), with the presence of Škoda in Mladá Boleslav, Vrchlabí and Kvasiny. The other carmakers and transport equipment manufacturers settled down in Kolín, Nošovice, Děčín-Křešice and Kutná Hora. The other sectors relying on factor abundance are textile industry in northeast, and the food production in Southeast and Central Moravia.

We will perform a panel data model bearing in mind the conviction that the most traditional industrial Czech regions NUTS 3 lodge the higher density of manufacturing firms and workers, although in general, each spot can exploit its endowments and grasp the benefit offered by its resources. The reliance on natural endowments is assumed as a dispersing force as the resources tend to be randomly scattered across the space, and in case of an excessive strength in the factor endowments, a full specialization is achieved.

Theoretically, for some data arranged in the way of panel data, the assumption of uncorrelation of error with the exogenous variables seems very strong. Moreover, the unobserved effects must be treated as random variables extracted from the data, similar to the observed endogenous and exogenous variables (Wooldridge, 2002).

Some empirical works pursue to model the patterns of trade at a country level, incorporating a set of variables embodying the abundance of resources and the idiosyncratic endowments. Lederman and Lixing (2007) relying on a classical vision of trade, estimate the determinants of the trade structure using some country-specific characteristics as exogenous variables, assuming that the endowments determine the national comparative advantage. In making up the empirical strategy, they include the conventional variables of labor and capital as proxies for endowments, but also ancillary aspects as the domestic infrastructure, domestic institutions, knowledge and schooling, and macroeconomic volatility, revealing some traits in such idiosyncratic conditions, which count forcefully as factors of international differentiation. Arias and Antosová (2020) used a similar approach when analyzing the manufacturing specialization, applying a panel data for the set of Colombian departments.

Our interest is to model the determinants of the regional distribution of firms, assuming the number of workers as a proxy of the spatial attraction force that induces a stubborn process of spatial concentration. and in consequence, a consolidation of the regional divergences. The manufacturing employment has also been used as an endogenous variable in Daniele et al. (2016), when estimating the determinants of the regional agglomeration in Italy.

We need to bear out the seemingly existence of fixed effects of time-invariant variables, that are embodied in some favorable conditions at the advanced manufacturing regions. Pursuing the empirical objectives, we applied a panel data technique and constructed a balanced panel with 14 NUTS 3 Czech regions, and arranged the data on a regional scale covering recent years (2016–2021). This strategy based on panel data was also applied in Lederman and Lixing (2007), and in Arias and Antosová (2020).

We presume that regional industry can be located in specific regions following local profitable dynamics, and that the action of the accumulative causation enhances the original advantages exhibited by primal centers of industrial agglomeration. Theoretically, such unobserved effects frequently appear in the panel data analysis, and are interpreted according to the unit of analysis. Regarding the features of an individual, the effects can be interpreted as abilities, the personal motivation, or the particular background influencing the individual performance. In the case of the firms, such effects can encompass the managerial quality or the corporative structure (Wooldridge, 2002).

It is also assumed that such time-invariant characteristics are specific for each individual, and should be independent regarding other individual features. For our regional panel data, we presume particular regional specificities in terms of their productive traditions, organization and economic development. Those unobserved effects are related to the most favorable conditions which more advanced regions can offer for the location of firms, which bestow upon them a stronger advantage for attracting manufacturing activities. The source of information is the manufacturing data drawn from EUROSTAT and the Czech Statistical Office¹ (CSO), and in such conditions, a balanced panel was elaborated with the 14 NUTS 3 Czech regions (Kraje).

This assertion is reinforced by the most diverse points of view. Encompassed in a microregional level analysis, Ženka et al. (2015) echoed the idea that the evolution of small manufacturing regions is clearly affected by the previous history of specialization, that tends to be pervasive in the long run, in particular for mature medium technological branches, having accumulated a set of local assets surrounding the local productive evolution.

Another powerful force steams from the economic influence coming from the conterminous regions at the neighboring countries. The detailed description of the analyzed regions at the NUTS 3 level appears in Figure 1.

Figure 1. GDP per capita in the Czech Conterminous Regions 2018. NUTS 3.

Source: Own elaboration based on EUROSTAT.

The run of a fixed effects model assumes that the time-invariant characteristics can affect the X or Y variables, and a procedure of control must be implemented, in such way that the estimation process removes the potential influence of the individual time-invariant characteristics, and the net effect emerging from the exogenous variables can be established.

For our purposes, the number of manufacturing workers in each region acts as the endogenous variable, as a proxy for the agglomerative forces that attract the industrial activity that is prone to be established locally, and along with this corporative flow of firms, workers are enticed to settle up in the same region as well.

We assume that agglomeration forces intervene in the spatial distribution of plants, and that cumulative causation leads to the concentration of industry, in the locations where scale economies and linkages could be effectively exploited. As mentioned early, we also expect that several natural conditions gave rise to a spatial process of localization, or some kind of Marshallian shared inputs can predominate for agglomerating firms and workers, and we presume a strong influence of the abundance of specific natural resources.

4.1. The Model

The source of information for all variables is the Czech Statistical Office, and the focus was zoomed in at the NUTS 3 level of geographical breakdown.

We alternated diverse manufacturing exogenous variables calibrating which one casts the best results. As said above, the final model of the panel data ended up using the regional industrial employment as the endogenous variable. The panel data specification has the following form:

(2) $\mathrm{M}\mathrm{i}\mathrm{s}\mathrm{s}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{o}\mathrm{p}\mathrm{e}\mathrm{n}\mathrm{b}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{e}\mathrm{f}\mathrm{o}\mathrm{r}\mathrm{s}\mathrm{u}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{s}\mathrm{c}\mathrm{r}\mathrm{i}\mathrm{p}\mathrm{t}{Y}_{it}=X_{it}\beta +\left({\alpha }_i+{\epsilon }_{it}\right)$(2)

where:

${\alpha }_i$ collects the unobservable factors that do not change on time and

${\epsilon }_{it}$ is the idiosyncratic error collecting all unobservable factors but that can change in time.

Performing this estimation, the technique of panel data enables us to determine unbiased estimators even in the presence of omitted variables (Wooldridge, 2002). Upon arranging the data as panel data, we need to identify the optimal procedure for estimating the fixed effects. In doing so, we will determine whether the most suitable econometric structure must be estimated by fixed or random effects, depending on the presence or absence of correlation between individual effects and the exogenous variables.

In the first case, we assume that ${\alpha }_i$ can be correlated with${\mathrm{X}}_{it}$, so this regressor can be endogenous. In such a situation, ordinary least squares (OLS) estimations of $\beta$ are inconsistent; however, the estimators regressed by fixed effects are consistent.

In the case of random effects, we assume that ${\alpha }_i$ is a randomly generated process unrelated to${\mathrm{X}}_{it}$; consequently, such a regressor is exogenous, and all estimations render consistent parameters.

As mentioned earlier, the inclusion of persons employed in the industry performed fairly well, and it can be assumed that this variable adequately embodies the agglomerative capacity of the regions for attracting workers in search for industrial opportunities. Theoretically, the abundance of workers in an area can unleash different effects according to the conceptual structure. In the realm of the comparative advantages, the presence of workers in a region drives the salaries down, while in the increasing return paradigm, the agglomeration process and the advent of workers will push the wages up (Ricci, 1997).

The selected exogenous variables performed relatively well in the estimation. The number of enterprises (lavgenterprise) indicates the appropriate economic conditions offered by regions, and the preexistent agglomerative forces. We include the logarithm of population density (lnpopdensity), pertaining to the size of regional markets that spur the confluence of undertakings around the mass of human agglomerations.

The summary of descriptive statistics for used variables appears in Table 2 as follows:

Table 2. Descriptive Statistics. The period of data is 2016–2021.

		Mean	Std. Dev.	Min	Max	Observations
lavindemploym	Overall	10.92024	.4249819	9.68	11.62	N = 84
	Between		.4368466	9.79	11.585	n = 14
	Within		.0356466	10.81024	11.00024	T = 6
lavgenterprise	Overall	5.059618	.3479247	4.060443	5.549076	N = 84
	Between		.3577484	4.158507	5.522442	n = 14
	Within					T = 6
NMA Gravitational	Overall	4.474625	.3160271	4.073274	5.254302	N = 84
	Between		.324717	4.096573	5.199224	n = 14
	Within		.0278843	4.961554	4.961554	T = 6
lnpopdensity	Overall	4.98076	.8720676	4.14847	7.89751	N = 84
	Between		.8995321	4.154593	7.873097	n = 14
	Within		.0093968	4.946608	5.005173	T = 6

Source: Own elaboration.

One interesting variable was adapted from a previous work developed by Italian authors analyzing the regional industrialization. The variable foreign market access (FMA) represents any sort of gravitation forces affecting the regional localization of industrial employment, perceiving the neighbor economies of the same country as foreign markets. As shown in Daniele et al. (2016), the construction of the variable is as follows:

$FM{A}_{j\hspace{0.17em}}=\sum _f^N{Y}_f{d}_{fj}^{-1}$

This spatial determinant is highly suitable in analyzing the Czech Republic industrialization, because the geographical surround, recently spurred the regional industrialization. The Czech regions are nested amid and international confluence of economic interests, with the intervenient participation of other European players, mainly Germans and Austrians. The movements of commuters, and the flows of foreign capital are forceful propulsors propelling the regional industrialization within the Czech Republic.

The FMA of each Czech region j is calculated by the sum of the GDP of all-other regions (f), weighted by the inverse relative distance measured between j and f. This variable assumes higher values as long as for each case, the neighbor regions are more important economically and are closer. In this exercise, we include the international conterminous German, Austrian, Polish and Slovakian regions, defined according the EUROSTAT NUTS 3 classification.

This gravitational intuition can be of utmost importance due to the influence of geographical and economic positions of Czech regions, along the East–West gradient of innovation and strength (Blažek & Csank, 2005). Such exogenous variable can model the regional macrogeographic position, assumed as a spatial distance of any Czech territory to the borders with foreign cores of manufacturing activity, in particular the German Bavaria and Austrian Ober-östereich. In the national realm, the role of Prague and the central Bohemian Region stands out.

However, the influence of an international and national hierarchy in the context of the West–East gradient (Blažek & Csank, 2005) must be critically assessed. The Czech involvement in an international production process is concentrated in low value-added sectors, generating a territorial disconnection and propitiating the development of hollow clusters. Under such terms, the Czech function can be reduced, in some cases, to second- and third-tier supplier roles (Ženka et al., 2015).

We realize that variables drawn from the industrial information perform very well with the regional number of workers, revealing a sort of internal coherence in data, having the CSO as a source including a broad range of indicators as attributes of the Czech regions, with a breakdown at the NUTS 3 level.

The population density and the gravitational variable are closely related to the powerful influence of the Home Market Effect. so ubiquitous in the NEG literature (Davis & Weinstein, 1996; Krugman, 2008). But also, the population density can be incorporated as a proxy for the local availability of labor, and of factor endowments as is explained in Leamer (1995), and Lederman and Lixing (2007). Furthermore, we include variables in natural logarithms to reduce the risk of heteroscedasticity (Lederman & Lixing, 2007).

5. Discussion of results

Firstly, we performed an OLS regression just for the sake of comparison, and for demonstrating the suitability of the panel data technique for our purposes. The output showed in Table 3 provides an insight about the existence of time-invariant characteristics, suggesting an intrinsic heterogeneity across regions based on a proper path dependence, and a divergent manufacturing history.

Table 3. Panel data estimations. Endogenous variable: Manufacturing employment by region.

	Fixed-effects (within) regression	Random effects GLS regression	OLS
	lavindemploym	lavindemploym	lavindemploym
lavgenterprise	.7091533***	.9518508***	1.095676***
Robust Std. Err.	.1458998	.0905864	.0508987
NMA Gravitational	.1645892 **	.1716428 *	.2701995***
Robust Std. Err.	(.0722545)	.0936711	.0608087
lnpopdensity	1.067498*	.0100001	−.0529257**
Robust Std. Err	.5282345	.0571594	.0224155
_cons	1.278767	5.286391 ***	4.431107***
Robust Std. Err.	(2.302815)	.5498557	.2806715
N	84	84	84
R2			0.8928
sigma_u	.94810958	.16010647
sigma_e	.02719234	.02719234
rho	.9991781	.97196341	(fraction of variance due to u_i)

Source: Own elaboration. Fixed effects regression in STATA robust to heteroscedasticity and autocorrelation: (fe vce(robust)).

In the presence of such individual time-invariant characteristics, the decision criteria impart the elements for choosing the fixed effects or the GLS random effects, as the most suitable econometric framework. Facing the flaws that loom when estimating the fixed effects model, we implemented the STATA VCE(robust) command to rectify the potential heteroscedasticity and autocorrelation presence. This command arranges a covariance consistent matrix, and the estimation process is proceeded with a variance of idiosyncratic errors robust to heteroscedasticity and serial autocorrelation.

We have the intuition about the influence of strong economic interactions exerted by the proximity to dynamic regions, whose manufacturing activity blossoms. This gravitational force exerts a strong attraction to firms that rely on the advantages of consolidated urban markets, and on agglomeration economies. It can be observed that the gravitational variable is significant either in the fixed effects and GLS random effects models, reinvigorating a powerful economic influence across geographical entities, as long as they are close to each other. It also helps to bear out the existence of the implicit time-invariant determinants previously explained. Later in the fixed effects models, the influence of regional individual features is confirmed by the significance of rho parameter.

Table 3 shows the results of panel data and OLS techniques. The comparison of the three models reveals that in OLS some exogenous variables strongly lose their significance, hinting that the OLS estimations are biased, and warn effectively about the existence of time-invariant characteristics.

In our fixed effects model, rho indicates that the 99% of variance is attributed to the differences across panels. All exogenous variables exhibit interesting interactions with the industrial employment at NUTS 3 level regions.

The exogenous variable representing the market is highly significant, excelling the local population density as a support for the development of economies of scale, and as determinant for the agglomeration. The NMA variable is significant, also hinting that some gravitational principles can guide the decision to mobilize industrial factors into the regions, and the geographic location near the powerful neighbor regions also reverberates in the agglomeration. The variable average number of enterprises is consistent with the endogenous variable, and they are fully compatible because both are drawn from the same statistical source.

The Brausch and Pagan test provides information confirming the existence of individual time-invariant effects characteristic for the Czech regions, that reverberate in the number of workers present regionally, and in the firm’s decision for location, bolstering or weakening the trends to spatial concentration. All exogenous variables are strongly significant and, in the majority, the mathematical signs are consistent with economic intuition.

If unobservable factors α are not correlated with the exogenous variables, we must assume that α performs as an additional factor reverberating in the endogenous variable. However, a serious drawback emerges if Cov ($X_i$, $\alpha$) ≠ 0 for some i, and the process continues embedding $\alpha$ into the error term (Wooldridge, 2002). With such caveat, the Hausmann test serves as the decision criterium about what model to choose, in case of independence between the non-observed factors and the exogenous variables, or in the case of Cov ($X_i$, $\alpha$) ≠ 0.

The Hausman test suggests that the fixed effects model estimates robustly the interaction between the regional industrial employment, and the exogenous variables. Trustfully, we are guided by the Hausman test for choosing the fixed effects model as the optimal structure of modeling for our data panel structure.

The fixed effects model seems to reinforce the relevance of spatial interactions set in motion by the geographical proximity. In addition, it excels the pervasive position of the most developed areas to perpetuate a continuous agglomeration process, in spite of the strong restructuring process undergone in particular branches. It is commonly stated that the city of Prague arranges a productive functional relationship with the Central Bohemian Region, shaping a clear division of labor over there. However, the persistence of some manufacturing branches in the very capital city, in spite of the taxation and the urban congestions, reflects the forceful influence of processes inspired in the NEG insights.

Ženka et al. (2015) identify a persistent reinforcement of the manufacturing activities in the Czech regions with the highest levels of per capita value added. This analysis is valid regardless of the disruptive transformation required during the transition era and the reconfiguration of productive forces once the international capital flowed into the Czech manufacturing structure.

5.1. Economic Policy Implications

One result of decades of centrally planned economy was an artificial regional specialization imposed by bureaucratic decision, which ignored the idiosyncratic endowments and the genuine vocation of regions. This policy also responded to external coactions stemming from other socialist powers. Accordingly, the legacy of socialist industrialization was a territorial deployment of large enterprises dominating the local economy lacking any basic diversification. The result of this strategy was the local overspecialization of the territories, which increased the local dependence on particular branches, and increased the exposure of the local economy to high risks of instability, during the restructuring cycles of the manufacturing sector. Afterward, when the Czech Republic (then Czechoslovakia) embraced the market economy, the market forces mobilized huge flows of capital into the manufacturing, defining the role of the country in an international system, and the full integration in Europe determined the destination of exports. It also involved the exposure to the risk of international crises and disruptive global downturns.

There is an unambiguous conclusion about the proved efficiency of the market forces for mobilizing technologies and resources toward the regions, shaping the production profiles. Facing the technological evolution and the development of new products and lines, the regional economy requires to adjust the production profile towards more modern and dynamic sectors. This process is typically driven by market forces, which has been triggered by the international mobility of factors and technologies. However, the initial conditions for the development of technological projects can be triggered by ad hoc stimulus, stemming from a multilevel action (local, regional, national and European). Based on the promotion of ad hoc systems of innovation, Prague recently has received more technological oriented branches (Blazek et al., 2011). Alongside, the growth of other urban activities and technological-intensive branches, can explain a recent blossom of specific manufacturing projects in the capital city.

Some recent theories have exceled the role of an activist policy in promoting specific sectors, namely: in technological lines of production and innovative sectors. This overarching strategy is commonly justified in the framework of strategic industrial policies, and evokes the Stuart Mill’s argument about the infant industry. It purports the implementation of a blatantly interventionist model of development with several goals and sectorial strategies. The external sector is the focus of transversal policies spanning the promotion of specific branches and the expansion of expenditure in R&D activities. There appears an interest for providing public funding, intended to endow the private production scheme with cutting-edge technologies in specific industries. There also appear goals in terms of spurring a digital industry as well.

In this article we offered an illuminating theoretical discussion which points out the forces driving the industrialization of the regions. A useful clue for promoting the industrial expansion, is the recent effort to reconcile the Ricardian theory and the NEG forces, combining the comparative advantages, the transport costs and the economies of scale (Puga, 1999). As demonstrated by recent NEG models, there is latent opportunity for industrializing the peripheries, to the extent that the comparative advantages are powerful, and the centrifugal forces expel manufacturing activities from the cores.

The key point is to purport an initial production development able to unfetter a self-sustaining process of growth and of spatial agglomeration, ready to be communicated spatially to adjacent areas.

6. Conclusions

We emphasize the role of gravitation principles in the configuration of manufacturing links and arrangements that stem from close exchanges between neighbor regions. These spatial connections alongside the typical NEG variables, as market size and agglomeration of firms, are tested using the econometric technique of panel data. A robust fixed effects model enables us to confirm the existence of a set of time-invariant features in each individual region, which contributes to explaining the production specialization, and the configuration of broader economic spaces.

In the Czech Republic, the take-off of traditional spots of manufacturing was strongly associated with the access to natural resources as coal, water sources, and ore mines. Many firms located their activities in the neighborhoods of important Czech cities such as Ostrava and Brno, inter alia, and this accidental location subsequently prompted a dynamic process of cumulative causation. The initial localized advantages tilted to physical and natural advantages, unleashed subsequently a pervasive process of industrialization, jet operating actually.

The initial process of exploitation of common resources for the production of metallurgic and related industries, can be interpreted by the Marshallian categories of natural shared inputs as the basis for subsequent manufacturing development. However, gradually, in those very territories, more advanced regional structures were developed, able to underpin ulterior productive processes with a higher complexity. The clustering of metallurgy, automotive and transport equipment branches in Silesia-Moravia, demonstrates this phenomenon.

This historical detail in the Czech manufacturing process is relevant for understanding the national economy (Dvořáčková, 2016). Being a relatively small country, transport costs and the internal distances can lose relevance for the spatial concentration of production, but the hurl to natural resources in the original industrialization can play a more influential role. Besides, the limitation of the internal market forced the local economy to tilt fully to the external customers, pursuing the expansion of the economies of scale.

In the econometric strategy, we strived to build a panel data made up by exogenous variables representing some regional traits (NUTS 3) associated with market size and gravitational forces. The exogenous variables demonstrate to be significant, indicating a strong economic interaction in the industrial location, when the neighbor regions are especially powerful in terms of GDP. It means that larger and closer neighbors are better for bolstering the local manufacturing agglomeration. Eventually, this gravitational influence can be very powerful when the foreign neighboring spaces are included in the analysis. The agglomerative forces and the self-reinforcing process of a firm’s setup, can relate to the local presence of plants, a variable that comes up as a significant influence.

The process of contagion of economic activity toward neighbor spaces turns out evident in the econometric exercise. The degree of detail at the NUTS 3 level contributed to zoom in the process of regional specialization, and the connection between adjacent spaces.

According to the empirical literature, the regional population density variable is significant, but also associated with one important endowment based on labor availability, giving rise to the influence of Ricardian forces.

The analysis of the recent Czech manufacturing drift can be encompassed in a staggered scenario of the national evolution, and the international mobility of resources and factors. The geographical position of the Czech Republic and the results of our econometric strategy, focus the attention into two simultaneous expansive process in Czech manufacturing. The two processes were quoted properly as the ‘West–East gradient’, and the vertical arrangement of regional Czech manufacturing (Blažek & Csank, 2005), and summarize the national regional adjustment originated in the reconnection of the regional specialization pattern with the local assets and endowments, but also the role of the Czech manufacturing amid a further international strategy of productive expansion (Ministry of Industry and Trade, 2018).

In terms of economic policy, the more recent developments dealing with geography and localization, give similar relevance to determinants born with a comparative advantage as a scattering force, and the increasing returns paradigm as the intrinsic force running the agglomeration process. In the end, the reasons for the location of production rely on the overlap of natural conditions which are very idiosyncratic in some regions, and one specialization pattern that springs up from the increasing returns. The presence of transport cost and segmented markets can reverse the divergent tendencies of the core periphery model. This argument, together with the assumption of minimal natural endowments in the peripheral regions, can conduce to a theoretical possibility of industrializing the peripheries.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. The Excel files consulted are ‘Position of regions in the Czech Republic by selected indicator’ and ’Selected data on industry by region’.