VoxEU Column Industrial organisation

How firms organise production: Headquarter separation and multi-plant operation

The core-periphery gap raises important questions for economic geography. Using Japanese data, this column examines firms’ decision to separate non-production activities from production plant facilities. Large plants, plants which intensively purchase materials, and plants located further from the core are more likely to have separate corporate headquarters, though the magnitude of this effect is small. Small-sized plants appear to be especially vulnerable to remoteness from urban cores.

Geographical disparity across regions, especially the core-periphery gap, has become an important policy issue in many countries. Firms fragment various functions and locate them in different regions, as examined by Duranton and Puga (2005). Production activities are often divided within a firm across multiple plants in remote locations to save production and/or transport costs. Among non-production activities, corporate headquarter (HQ) functions are increasingly concentrated in urban core regions. The constraint on internal resources, such as management capability, could make small-sized plants especially susceptible to the remoteness from agglomeration. Accordingly, it will be informative to examine how plants operated by firms with HQs in different locations (possibly in urban core) differ from plants collocated with HQs in terms of plant size or other plant characteristics. The comparison of plants operated by multi-plant firms with those by single-plant firms will also be useful to discuss how a firm splits production across multiple plants.

Corporate HQs control production plants of various sizes. Large-sized plants tend to need a wide variety of corporate services often available in agglomerated cores, and thus require separated HQs. From US firm-level data, Atalay et al. (2014) found that transactions of goods between upstream and downstream plants within the firm boundary are extremely rare, and that plant/firm size is the strongest determinant for vertical ownership. They indicate that the provision of corporate intangible inputs from the HQ, rather than intra-firm trade in goods, determines which plants are owned by vertically linked firms.

A plant with a large number of workers, however, might require more management attention and the local presence of plant managers, necessitating the collocation of HQs with these large-sized plants. Recent studies based on US plant-level data (Giroud 2013, Kalnins and Lafontaine 2013) found that HQ-plant communications affect plants’ performance, although they do not examine the HQ separation decision itself. Decreasing returns to scale in production or in management are likely to induce the separation of HQ or the splitting of production across multiple plants. Among US firms, Aarland et al. (2007) reported that firms with separated HQs are substantially bigger than those without. Consequently, corporate organisation decisions involve serious trade-offs. This implies the importance of empirical investigation of this issue.

Findings from plant-level data

To investigate these issues, we derived plant-level panel data from Japan’s Census of Manufacture covering 1992-2008 (Okubo and Tomiura 2016). The survey asks each plant whether its HQ is physically collocated with or separated from the plant,[1] and whether each plant is a part of a multi-plant firm. In addition to these valuable data, the survey contains basic plant information, such as output (sales), employment (number of regular workers), and expenditure on materials for all plants with four or more employees in all manufacturing industries in Japan. We estimated plant-level probit regressions with these two corporate organisation decisions as the dependent variables. The principal results are as follows.

The most notable finding is that a large-sized plant (in terms of employment) tends to be separated from HQs and one of the multiple plants of the firm. Plants that intensively purchase materials are also likely to be operated by firms with separated HQs or by multi-plant firms. These patterns suggest decreasing returns to scale in plant operations, such as rising management burdens owing to an increase in employment or materials. HQs are also likely to be separated from production plants to give more autonomy to the managers of large plants. Although the impact of plant size is statistically significant at any conventional significance level, we must note that the impact is not large in magnitude. A 10% expansion of the number of workers results in the increase in probability of HQ separation merely by 0.4% to 0.5% evaluated at the mean. We also find that plants with high productivity (total factor productivity or per-worker value-added) or with high wage (richer human capital) tend to be separated from HQ or to be operated as a part of multiple plants of the firm.

Factors of economic geography, such as distance from core, also have significant impacts on corporate organisation of production. Plants located far from agglomerated cores (Tokyo or Osaka in the case of Japan) are likely to have separated HQ and to be a part of a multi-plant operation. Firms should have incentives to separate their corporate HQ from production plants if plants are located in peripheral regions distanced from the urban core. The locations of plants operated by multi-plant firms (possibly with differentiated product lines or large-scale production of commodities) should be geographically dispersed, with at least some of them in the periphery, to serve the respective local markets. However, this effect of distance appears somewhat attenuated in larger plants, suggesting the role of internal resources.

To check the robustness, we also examine dynamic switching patterns of plant status, such as separating HQ, transforming to multi-plant operation, or entry of new plants. Our panel data results confirm that our main findings from static regressions are unaltered.

Concluding remarks

We show that plant size is significantly related to firms’ decisions to separate HQ and to divide production into multiple plants. We have also examined how corporate organisation decisions are interacted with the spatial organisation (location) of production. Our results indicate that small-sized plants are especially vulnerable to the remoteness from urban cores.

Corporate organisation inside the firm and location decisions of firms are among the vital issues for regional economic policy of many developed countries, as intensified global competition puts pressure on many firms in high-wage countries to reorganise and relocate. Our findings suggest that corporate reorganisations are significantly affected not only by internal factors but also by external factors such as the distance from agglomeration. By integrating plant-level data with other data sources, it will be informative to expand our research scope to include overseas locations, multinational corporations and/or outsourcing across firm boundaries in future studies.


Aarland, K, J Davis, J V Henderson and Y Ono (2007) “Spatial organization of firms: The decision to split production and administration”, RAND Journal of Economics, 38: 480–494.

Atalay, E, A Hortaçsu and C Syverson (2014) “Vertical integration and input flows”, American Economic Review, 104: 1120-48.

Duranton, G and D Puga (2005) “From sectoral to functional specialization”, Journal of Urban Economics, 57: 343–370.

Giroud, X (2013) “Proximity and investment: Evidence from plant-level data”, Quarterly Journal of Economics, 128(2): 861-915.

Kalnins, A and F Lafontaine (2013) “Too far away? The effect of distance to headquarters on business establishment performance”, American Economic Journal: Microeconomics, 5(3): 157-179.

Okubo, T and E Tomiura (2016) “Multi-plant operation and corporate headquarters separation: Evidence from Japanese plant-level panel data”, Japan and the World Economy, 39: 12-22.


[1] The survey asks this question on HQ separation only to a plant operated by a single-plant firm, not to a plant operated by a multi-plant firm. The location of HQ is not captured by this survey of manufacturing plants.

1,050 Reads