Abstract
We show that key functions are spatially clustered with, or dispersed from, each other even within manufacturing industries in West Germany, and that these clustering or dispersion patterns have changed significantly during recent decades. Estimating levels and changes (1992–2007) of localizations and colocalizations of selected functions (production, headquarter services, R&D) within 27 West German industries by means of \(K\) densities, we identify two broad groups of industries. In “fragmenting” industries, which account for half of manufacturing employment, functions were more clustered with each other than the industry as a whole after the fall of the Iron Curtain but have, in accordance with regional theories of spatial fragmentation, been unbundled spatially from each other subsequently. In “integrating” industries, by contrast, which account for one-third of manufacturing employment, functions were initially dispersed from each other but have subsequently been rebundled spatially with each other. This spatial rebundling may be a consequence of offshoring, i.e., international fragmentation.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
The term “integrating” is actually meant to be the antonym to “fragmenting.”
While these models focus on offshoring of production, similar arguments could be made for the outsourcing of production or other activities. Bargaining over the legal details of a contract, fixing and adjusting the technical details of a product, and coordinating production plans will be more costly across larger distances as well.
We will discuss the related literature in the international economics domain below.
See Combes and Gobillon (2015) for a recent critical review of this strand of the literature.
Somewhat complementary to the evidence on headquarter locations, Atalay et al. (2014) find that the main ownership advantage of multiplant firms in the USA results from intra-firm transfers of intangible inputs such as managerial oversight, while there is very little intra-firm shipment of goods between vertically related plants. This does not imply that firms do not outsource domestically, of course. On the contrary, several studies, including Fally (2012), Fort (2013) and Schwörer (2013), show that domestic sourcing of intermediate inputs is still quantitatively much more important than foreign sourcing even though foreign sourcing has expanded disproportionately during recent decades. Mayer et al. (2010) find a similar “home bias” for the location choices for investments in production affiliates by French manufacturing firms. None of these studies addresses the spatial dimension of domestic sourcing or investments, though.
Since the establishments do not have a unique identifier in our dataset, we actually cannot trace the individual establishments over time. We can neither identify ownership relationships between the establishments from the data.
An alternative, similarly implausible explanation would be that these industries have switched to new products or processes that are associated with substantially higher costs of managing remote production plants, in spite of advances in ICT.
There have been a few changes in the administrative structure of West German municipalities during the period under study. Most of these changes involve mergers of small municipalities with larger cities. We have not eliminated these changes over time from our data because we do not expect them to affect our results.
Distance-based measures like the \(K\) density have been used frequently for assessing the localization or colocalization of industries. See, among others, Duranton and Overman (2005, 2008), Klier and McMillen (2008), Arbia et al. (2010), Ellison et al. (2010), Marcon and Puech (2010), and homas-Agnon and Bonneu (2014).
To characterize this reference for our measures, we additionally estimate (employment-weighted) \(K\) densities of localization for each industry to assess if the industry as a whole is more or less localized than the manufacturing sector on aggregate.
For the analysis of the localization of a single function (Eq. 1), it is sufficient to limit the set of observations to the \(N_{it}(N_{it}\)-1)/2 unique establishment pairs (see Duranton and Overman 2005), while the analysis of the colocalization of two functions (Eq. 2) requires evaluating all \(N_{it}^2 \) establishment pairs.
Following Duranton and Overman (2005), we use kernel smoothing to account for measurement errors as far as possible. We use a fixed bandwidth of 20 km for all \(K\) densities. Columns (3)–(6) of Table 9 indicate that the classification of industries into fragmenting, localizing and other industries is fairly robust to variations of the bandwidth.
We prefer this method over the dynamic bivariate (space-time) \(K\) functions discussed in Arbia et al. (2010) mainly because we have only two observations in time. Dynamic \(K\) functions require observations in continuous time or at least longer time series in order to mitigate edge effects.
Estimated \(K\) densities for population or aggregate employment typically show a local peak at around 150 km. Examples of city pairs at a distance of roughly 150 km from each other are Munich–Nuremberg, Frankfurt–Stuttgart, Frankfurt–Cologne, Frankfurt–Kassel or Hamburg–Hannover.
The estimated \(K\) density for 2007, which is not reported here, shows that the two functions were still significantly colocalized up to a distance of 47.5 km at the end of our study period, though.
We use “codispersion” as the antonym to “colocalization.” We will say that two functions are codispersed if the actual \(K\) density is below the lower bound of its confidence interval at short distances. Notice that codispersion does not necessarily mean that both functions are individually dispersed. Both may be individually clustered, though at distinctly different places.
The industries that exhibit no significant changes in the colocalizations of function pairs at all are grouped into a third category, called “other” industries.
Some additional, practical aspects are worth being noted. First, following (Duranton and Overman 2005), we use the reflection method (Silverman 1986) to prevent the \(K\) density estimates at distances close to zero from being biased downward. Second, we aggregate the observed distances between establishments to intervals of 500 m to speed up the \(K\) density estimations. Each distance interval is represented by its upper bound (i.e., 500 m, 1 km, 1.5, ..., 891.5 km). This implies that all \(K\) densities reported in this paper are estimated from 1,783 unique distance points, respectively, 3,566 points after applying the reflection method. Column (1) of Table 8 (“Appendix 3”), which reports the results for distance intervals of 100 m, indicates that this aggregation of distances does not affect the classification of industries significantly.
No entry indicates that there was no significant (co-)localization or (co-)dispersion, or no significant change over time. Table 7 in “Appendix 3” reports the corresponding point estimates for the threshold distances where the estimated \(K\) densities dip into their confidence intervals. Table 8 reports the results for 2007.
This share of localized industries is somewhat higher than that of 50 % reported for the UK by Duranton and Overman (2005).
A third group, “other” industries (lower part of Table 1), comprises those six industries that exhibit no significant changes in the colocalizations of functions at all. Like the fragmenting industries, these industries are characterized by significant colocalizations of the three functions with each other in the early 1990s. We will not discuss the results for these industries in more detail below.
If, for example, Siemens, a German consumer electronics company headquartered in the southern Germany city of Munich, offshores a production line from one of its northern German plants to China, the aggregate distance between Siemens’ remaining production plants in Germany will decrease, and more mass will move to shorter distances in our estimated univariate \(K\) density for the localization of production in the consumer electronics industry. In addition to this, the aggregate distances between Siemens’ remaining production plants and its HQ and R&D centers will decrease, and more mass will move to shorter distances in our estimated bivariate \(K\) densities for the colocalization of production with HQ or R&D in the consumer electronics industry.
Important exceptions from these general patterns are “Measuring, Testing, etc.” and “Other Manufacturing,” whose shares in manufacturing employment decreased considerably. As we will argue in the next subsection, these two industries may actually match the characteristics of domestically localizing industries better than those of fragmenting industries.
These industries are “Metal Forming Machinery and Machine Tools,” “Consumer Electronics,” “Leather and Apparel” and “Printing.” In “Consumer Electronics,” non-production employment expanded considerably, though.
The data on outward FDI stocks is available from the Deutsche Bundesbank (Bestandserhebung über Direktinvestitionen, Statistische Sonderveröffentlichung 10, Table I.3).
These indicators are developed by Geishecker (2006), where they are labeled “material offshoring.” See also Baumgarten et al. (2013) and Schwörer (2013). These indicators assign material imports to domestic industries by means of the industries’ input coefficients for these goods. We thank Tillman Schwörer for sharing his most recent estimates with us.
The last column of Table 3 additionally depicts location coefficients for East Germany, which will be discussed in the next subsection.
The absolute amounts of FDI and imported intermediates show a similar picture. The FDI stocks (in Euro) increased by 129 % (from 42 to 96 bill.€) for the domestically fragmenting industries and by 279 % (from 27 to 101 bill.€) for the domestically integrating industries. And the shares of imported intermediates (wide) increased by 117 % for the domestically fragmenting and by 145 % for the domestically integrating industries.
Indeed, “Other Manufacturing” is one of the three industries Feenstra and Hanson (1996: 242) mention explicitly as being particularly prone to offshoring. As for the “Measuring, Testing and Electromedical equipment” industry, it is worth noting that production of traditional surgical instruments has been offshored to a considerable extent, mainly to Pakistan but also to Poland, Hungary and Malaysia (Nadvi and Halder 2005).
This radius is calculated as \(\tau _r =\sqrt{A_r /\pi }\), where \(A_{r}\) is the municipality \(r\)’s area (in sqkm). The radii range from 0.35 to 15.5 km for West German municipalities, with a mean of 2.7 km and a 90th percentile of 4.7 km.
We actually use the same counterfactual distributions as those constructed for the references for the univariate measures of localization (see above).
This approach is conceptually very similar to that used to evaluate changes in inequality measures over time (see, e.g., Mills and Zandvakili 1997).
References
Aarland K, Davis JC, Henderson JV, Ono Y (2007) Spatial organization of firms: the decision to split production and administration. Rand J Econ 38(2):480–494
Alecke B, Alsleben C, Scharr F, Untiedt G (2006) Are there really high-tech clusters? The geographic concentration of German manufacturing industries and its determinants. Ann Reg Sci 40(1):19–42
Antras P, Yeaple SR (2014) Multinational firms and the structure of international trade. In: Helpman E, Rogoff K, Gopinath G (eds) Handbook of international economics 4. Elsevier, Amsterdam, pp 55–130
Arbia G, Espa G, Giuliani D, Mazzitelli A (2010) Detecting the existence of space-time clustering of firms. Reg Sci Urban Econ 40(5):311–323
Atalay E, Hortacsu A, Syverson C (2014) Vertical integration and input flows. Am Econ Rev 104(4):1120–1148
Bade F-J, Laaser C-F, Soltwedel R (2004) Urban specialization in the internet age: empirical findings for Germany. Kiel working paper 1215, Kiel Institute for World Economics
Baldwin RE (2006) Globalisation: the great unbundling(s). Paper prepared for Finnish Prime Minister’s Office for EU Presidency. www.hei.unige.ch/baldwin/
Baumgarten D, Geishecker I, Görg H (2013) Offshoring, tasks, and the skill-wage pattern. Eur Econ Rev 61:132–152
Bickenbach F, Bode E (2013) New economic geography and reunified Germany at twenty: a fruitful match? Spat Econ Anal 8(2):120–153
Bode E (2004) The spatial pattern of localized R&D spillovers: an empirical investigation for Germany. J Econ Geogr 4(1):43–64
Carlino G, Kerr WR (2015) Agglomeration and innovation. In: Duranton G, Henderson JV, Strange WC (eds) Handbook of regional and urban economics 5. North Holland, Amsterdam. http://real.wharton.upenn.edu/~duranton/Duranton_Papers/Handbook/Agglomeration_and_innovation.pdf
Combes P-P, Gobillon L (2015) The empirics of agglomeration economies. In Duranton G, Henderson JV, Strange WC (eds) Handbook of regional and urban economics 5. North Holland, Amsterdam. http://real.wharton.upenn.edu/~duranton/Duranton_Papers/Handbook/The_empirics_of_agglomeration_economics.pdf
Combes P-P, Lafourcade M (2005) Transport costs: measures, determinants, and regional policy implications for France. J Econ Geogr 5(3):319–349
Daudin G, Rifflart C, Schweisguth D (2011) Who produces for whom in the world economy? Can J Econ 44(4):1403–1437
Davis JC, Henderson JV (2008) The agglomeration of headquarters. Reg Sci Urban Econ 38(5):445–460
Desmet K, Henderson JV (2015) The geography of development within countries. In: Duranton G, Henderson JV, Strange WC (eds) Handbook of regional and urban economics 5. North Holland, Amsterdam. http://real.wharton.upenn.edu/~duranton/Duranton_Papers/Handbook/The_geography_of_development_within_countries.pdf
Duranton G, Overman HG (2005) Testing for localisation using micro-geographic data. Rev Econ Stud 72(4):1077–1106
Duranton G, Overman HG (2008) Exploring the detailed location patterns of UK manufacturing industries using microgeographic data. J Reg Sci 48(1):213–243
Duranton G, Puga D (2005) From sectoral to functional urban specialisation. J Urban Econ 57(2):343–370
Ellison G, Glaeser EL, Kerr WR (2010) What causes industry agglomeration? Evidence from coagglomeration patterns. Am Econ Rev 100(3):1195–1213
Fally T (2012) Production staging: measurement and facts. Mimeo, University of Colorado-Boulder. http://are.berkeley.edu/~fally/Papers/Fragmentation_US_Aug_2012.pdf
Feenstra RC, Hanson GH (1996) Globalization, outsourcing and wage inequality. Am Econ Rev 86(2):240–245
Fort T (2013) Breaking up is hard to do: why firms fragment production across locations. Mimeo, Tuck School of Business at Dartmouth. http://faculty.tuck.dartmouth.edu/images/uploads/faculty/teresa-fort/FORT_Fragmentation.pdf
Gabe TM, Abel JR (2013) Shared knowledge and the coagglomeration of occupations. Staff Report 612. Federal Reserve Bank of New York, New York
Geishecker I (2006) Does outsourcing to central and eastern Europe really threaten manual workers’ jobs in Germany? World Econ 29(5):559–583
Görlich D, Bode E, Schwörer T (2014) Ursachen der deutschen Exportstärke: Zur Bedeutung von Vorleistungsimporten und nicht-preislicher Wettbewerbsfähigkeit. Monatsbericht des BMF, Oktober, German Ministry of Finance, Berlin, pp 19–27. http://www.bundesfinanzministerium.de/Content/DE/Monatsberichte/2014/10/Downloads/monatsbericht_2014_10_deutsch.pdf?__blob=publicationFile&v=3
Helpman E (2006) Trade, FDI, and the organization of firms. J Econ Lit 44(3):589–630
Henderson JV, Ono Y (2008) Where do manufacturing firms locate their headquarters? J Urban Econ 63(2):431–450
Holloway SR, Wheeler JO (1991) Corporate headquarters relocation and changes in metropolitan corporate dominance, 1980–1987. Econ Geogr 67(1):54–74
Hummels D, Rapoport D, Yi K-M (1998) Vertical specialization and the changing nature of world trade. Econ Policy Rev 4(2):79–99
Hummels D, Ishii J, Yi K-M (2001) The nature and growth of vertical specialization in world trade. J Int Econ 54(1):75–96
Jaffe AB, Trajtenberg M, Henderson R (1993) Geographic localization of knowledge spillovers as evidenced by patent citations. Q J Econ 108(3):577–598
Jofre-Monseny J, Marín-López R, Viladecans-Marsal E (2011) The mechanisms of agglomeration: evidence from the effect of inter-industry relations on the location of new firms. J Urban Econ 70(2–3):61–74
Klier T, McMillen DP (2008) Evolving agglomeration in the US auto supplier industry. J Reg Sci 48(1):245–267
Kolko J (2010) Urbanization, agglomeration, and coagglomeration of service industries. In: Glaeser EL (ed) The economics of agglomeration. National Bureau of Economic Research, Cambridge, pp 151–180
Lovely ME, Rosenthal SS, Sharma S (2005) Information, agglomeration, and the head-quarters of US exporters. Reg Sci Urban Econ 35(2):167–191
Marcon E, Puech F (2010) Measures of the geographic concentration of industries: improving distance-based methods. J Econ Geogr 10(5):745–762
Marin D (2010) The opening up of Eastern Europe at 20: jobs, skills, and ‘Reverse Maquiladoras’ in Austria and Germany. Munich Discussion Paper 2010–14, Munich
Mayer T, Mejean I, Nefussi B (2010) The location of domestic and foreign production affiliates by French multinational firms. J Urban Econ 68(2):115–128
Mills JA, Zandvakili S (1997) Statistical inference via bootstrapping for measures of inequality. J Appl Econ 12(2):133–150
Nadvi K, Halder G (2005) Local clusters in global value chains: exploring dynamic linkages between Germany and Pakistan. Entrep Reg Dev 17(5):339–363
Ota M, Fujita M (1993) Communication technologies and spatial organization of multi-unit firms in metropolitan areas. Reg Sci Urban Econ 23(6):695–729
Rosenthal SS, Strange WC (2001) The determinants of agglomeration. J Urban Econ 50(2):191–229
Rossi-Hansberg E, Sarte P-D, Owens R III (2009) Firm fragmentation and urban patterns. Int Econ Rev 50(1):143–186
Schott PK (2004) Across-product versus within-product specialization in international trade. Q J Econ 119(2):647–678
Schwörer T (2013) Offshoring, domestic outsourcing and productivity: evidence for a number of European countries. Rev World Econ 149(1):131–149
Silverman BW (1986) Density estimation for statistics and data analysis. Chapman and Hall, New York
Strauss-Kahn V, Vives X (2009) Why and where do headquarters move? Reg Sci Urban Econ 39(2):168–186
Thomas-Agnon C, Bonneu F (2014) Measuring and testing spatial mass concentration with micro-geographic data. Working Paper 474, Toulouse School of Economics, Toulouse
Acknowledgments
The authors thank Frank Bickenbach, Sebastian Braun, Dirk Dohse, Ursula Fritsch, Holger Görg, Tillmann Schwörer, an anonymous referee and Janet Kohlhase for helpful comments, discussions and suggestions, and Michaela Rank for valuably research assistance. Older versions of this paper circulated as “Does Domestic Offshoring Precede International Offshoring?” or “The Spatial Fragmentation of Production Activities: Patterns of Industrial and Functional Localization”.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix 1: Bandwidth choice and measurement errors for distances
The choice of the bandwidth is important for several reasons. One reason is that it affects the bias of the kernel density estimator (Silverman 1986). Several methods for selecting an optimal bandwidth are discussed in the literature and are available in standard statistical software packages. These bandwidths usually take the statistical properties of the sample dataset into account. A frequently used selection method, Silverman’s rule of thumb, suggests, for example, that the bandwidth should increase with the interquartile range and decrease with the size of the sample. Such a bandwidth, which would range between 50 and 90 km for our data, depending on the functions under study, would oversmooth our Kernel densities greatly because our aggregation of establishment pairs across distance intervals of 500 m (see Sect. 3) reduces our sample size significantly.
Another reason is that the bandwidth should account for the measurement errors in our distance data. Our approximation of interregional distances is subject to measurement errors from three sources. First, Euclidean distances do not take into account the curvature of the earth. This bias can be expected to be negligible for a small country like Germany where the maximum distance is below 1,000 km. Second, Euclidean distances do not take into account the density and quality of the available infrastructure. The actual traveling time per km may differ between high- and low density areas. On the one hand, the denser road networks in high-density areas will offer more direct connections. On the other hand, congestion may reduce the speed. While Combes and Lafourcade (2005) show that Euclidean distances and economic distances are correlated very highly with each other (0.97), we have no reliable information on the magnitudes of the errors that result from approximating economic by Euclidean distances.
And finally, not all establishments are located at their municipalities’ centroids. The corresponding error may be positive or negative, depending on where exactly the establishments are actually situated relative to the centroids. If the municipalities’ areas were perfectly circular, the magnitude of this error ranged from zero to the sum of the radii of the respective two municipalities. The variance of this error thus tends to be higher for larger municipalities, ceteris paribus. To give an idea of the possible magnitudes of the errors in interregional distances, Fig. 2 depicts the distribution across municipalities of their lower and upper bounds, calculated under the assumption that all municipalities are circular. For any two municipalities \(r\) and \(s\) these errors are bounded between \(-(\tau _{r}+\tau _{s})\) and (\(\tau _{r}+\tau _{s})\) where \(\tau \) denotes the radius of a municipality.Footnote 30 For comparison, Fig. 2 also depicts a normal distribution whose tails roughly encompass the highest possible approximation errors. The standard deviation of this distribution is 5 km. A bandwidth of 5 km will thus be sufficient to account for the highest possible measurement errors that result from approximating firms’ locations within municipalities by the municipalities’ centroids. Adding to these 5 km another 15 km to account for the mismatch between economic by Euclidean distances, we choose a bandwidth of 20 km for all pairs of functions as a baseline. Columns (3)–(6) of Table 9 (“Appendix 3”) report the results of robustness checks for bandwidths of 10, 15, 25 and 30 km. They indicate that the classification of industries is fairly robust to choice of the bandwidth.
Distribution of approximate maximum approximation errors for distances between West German municipalities and normal distribution with standard deviation of 5 km. Solid line density of the maxima of the approximation errors for distances between establishments from different municipalities under the assumption that all municipalities’ areas are circular. Dashed line density of \(N\)(0, 25)
Appendix 2: Counterfactual references
This appendix discusses the methods of constructing counterfactual references for the uni- and bivariate \(K\) densities as well as for their changes over time.
1.1 Significance of localization or colocalization
A question that arises naturally from inspecting the extent of localization or colocalization at a given point in time is whether this is the result of systematic, purposeful location decisions by firms, motivated by the wish to locate functions close to each other, or just the consequence of a series of independent location decisions that happened to generate some accidental spatial clustering of functions. We follow Duranton and Puga (2005) in using Monte Carlo methods to construct a counterfactual reference for each \(K\) density. We construct this reference under the null hypothesis that the location patterns of the functions under study are the results of random, industry-specific location decisions.
For the measure of localization of a single function in (1), this counterfactual reference indicates how the density distribution for the localization of this function may have looked like in 1992 (or 2007), if firms from the respective industry had had no incentives or disincentives for colocating establishments performing this function with each other. We construct this reference for function \(j\) in industry \(i\) by repeatedly resampling (without replacement) all establishments that perform function \(j\) (including their function-\(j\) workers) randomly among the population of all industrial sites occupied by establishments from industry \(i\) in West Germany in the same year, irrespective of whether or not function \(j\) was actually performed at this site. By resampling only among the sites occupied by the industry rather than among those occupied by any manufacturing industry, we focus on the motives for clustering establishments within this industry. We do not want to call a function localized just because its industry is more localized than manufacturing as a whole. By resampling without replacement, we make sure that each feasible industrial site is occupied by at most one establishment in the counterfactual distribution. And by resampling the existing establishments together with their actual number of function-\(j\) workers, we retain not only the number but also the size distribution of this function across establishments. We thereby exclude the effects of managerial decisions on optimal lot sizes from our analysis. We do not want to call a function localized just because its large minimal optimal lot size requires concentrating employment in only a few sites.
Repeating this random resampling 1,000 times, we obtain 1,000 counterfactual spatial distributions of the actual establishments for function \(j\) in industry \(i\), from which we estimate 1,000 counterfactual weighted univariate \(K\) densities in the same way as we estimate the actual \(K\) density (see Eq. 1). We use these 1,000 counterfactual \(K\) densities, in turn, to construct a two-sided 90 % confidence interval, which we take to cover, for each distance, \(d\), the range of densities consistent with no localization of the function in question. The robustness check reported in Column (2) of Table 9 indicates that 1,000 repetitions are enough. Increasing the number of repetitions to 2,000 does not change the classification of industries into domestically fragmenting, domestically localizing or other industries notably.
For the measure of colocalization of two functions in (2), we construct a similar counterfactual reference that indicates how the density distribution for the colocalization of the two functions may have looked like in 1992 (or 2007), if firms had had no incentives or disincentives for colocating them, given the size distributions of the two functions across establishments. We randomly resample each of the two functions independently of each other 1,000 times in the same way as described aboveFootnote 31 and estimate from these 2 \(\times \) 1,000 random distributions 1,000 counterfactual weighted bivariate \(K\) densities in the same way as we estimate the actual bivariate \(K\) density (see Eq. 2). From this, we construct a two-sided 90 % confidence interval, which we take to cover, for each distance, \(d\), the range of densities consistent with no colocalization of the two functions in question.
1.2 Significance of changes in localization or colocalization over time
To assess the significance of the changes of localization or colocalization over time, we construct a counterfactual reference that indicates how the density distribution for the localization (or colocalization) may have changed between 1992 and 2007, if there had been no incentives or disincentives for colocating establishments performing this function (these functions) at both points in time. This counterfactual reference should account not only for the changes in the location patterns of the industry as a whole. It should also account for the changes in the locational patterns of each function that are due to changes in their size distributions or optimal lot sizes. We do, for example, not want to conclude that a function became more localized just because its industry as a whole became more localized, or because the optimal lot sizes for individual functions increased over time. We construct the confidence interval for the change of an estimated \(K\) density over time from the differences between corresponding counterfactual \(K\) densities for 2007 and 1992. These counterfactual \(K\) densities are constructed independently of each other for each point in time in the same way as those for the levels of localization or colocalization (see above).Footnote 32 If the difference between estimated \(K\) densities for 2007 and 1992 lies above the upper bound of the 90 % confidence interval of the distribution of these 1,000 counterfactual differences at short distances, we will say that the respective function became more localized (or the two functions became more colocalized) over time. And if it lies below the lower bound of this confidence interval, we will say that the function became more dispersed (or the two functions became more codispersed) over time.
