Communication Costs and the Internal Organization of Multi-Plant Businesses: Evidence from the Impact of the French High-Speed Rail

NON-TECHNICAL SUMMARY
Large corporations operating multiple affiliates located in different areas are prevalent and account for a disproportionate fraction of output and employment. The Figure below shows in particular that a very large fraction of employment throughout the entire French metropolitan territory is managed at arm’s length by headquarters located in Paris. The splitting and implantation decision of such corporations is well understood: they involve a trade-off between the gains to access remote markets and the managerial costs to operate distant affiliates. Yet, the precise nature of these managerial costs remains largely a black box. In this paper, we contribute to fill this gap and provide detailed evidence about the nature of the higher managerial costs implied by geographic dispersion.

We take advantage of large scale administrative and survey data allowing us to describe the structure of the workforce of all French corporate groups over 19 years, from 1993 to 2011. We also rely on the expansion of the French high-speed rail (HSR) network over that period to assess the organizational impact of decreases in communication costs between headquarters and affiliates of groups benefiting from the new infrastructure. Our empirical investigation is guided by theoretical predictions. The literature in corporate finance (Giroud, 2013; Giroud and Mueller, 2015) suggests that the geographical dispersion of affiliates is likely to hamper information gathering and monitoring by managers of the group’s headquarters. This amplifies moral hazard problems and decreases the incentives to invest in remote affiliates, thus negatively affecting their size. The literature in economic geography (Duranton and Puga, 2005) suggests in turn that these information flows are to be understood as transfers of HQ services from headquarters to affiliates. Rendering these transfers more costly increases the need to locate the corresponding support activities at remote affiliates, which lowers the functional specialization of affiliates on their production activities.
Our regressions show that the impact of reduced travel times is positive on both affiliate size (production capacity) and functional specialization, as predicted by theory. Furthermore, we obtain that this impact is highest in the service industries, where information to be transmitted is arguably softer (Petersen and Rajan, 2002). Results are also significant in the trade and manufacturing industries, but point estimates are lower. To gauge the orders of magnitude, we propose to quantify the impact of the HSR network as a whole on the management structure and size of remote affiliates benefiting from it as of 2011. We obtain that for an affiliate benefiting from the average decrease in travel time to its headquarters, the availability of the HSR infrastructure induced the shift of roughly one job from administrative to production activities in service industries against 20% of a job in other industries (retail, trade or manufacturing). Affiliates operating in the manufacturing and business services industries also experienced decreases in production labor costs, of around half the cost of a production job on average. At the group level, our regressions suggest that the impact on the operational profit margin ranges from 0.5 to 1.5 percentage points depending on the industry. Our empirical analysis can also be used to provide estimates of the overall cost of geographical dispersion for multi-site businesses, although in a partial equilibrium setting. For example, our estimates imply that remote affiliates in the personnel service industries would operate with roughly 1.5 additional production jobs on average if geographical distance could be fully abolished by a perfect communication technology. The figure would rise to 4 production jobs on average in business services industries, and to 2 production jobs in manufacturing industries. The latter quantifications are to be interpreted as upper bounds for the productivity effect that can be expected from internal communication technologies.