Spatial organization of firms: the decision to split production and administration.

A firm's production activities are often supported by nonproduction activities, such as administrative units including headquarters, which process information both within and between firms. Firms may physically separate such administrative units from their production activities and create stand-alone central administrative offices (CAOs). However, activities in multiple locations may cause internal communication costs. What types of firms are more likely to separate such functions? If firms separate administration and production, where do they locate CAOs? This paper examines firms' spatial organization using microlevel data from the US Census Bureau.

1. Introduction

* Firms' production activities are supported by various nonproduction activities such as management, marketing and administration. While these administrative functions can, in principle, be performed at the same site as production, many firms favor stand-alone central administrative offices (CAOs), for a variety of reasons. First, management may want more pleasant office facilities than offered in typical production plants, separating administrators from noise and pollution. However, for sectors such as finance and service industries, there may be less need for a physically separate office for management.

Second, plants in manufacturing, for example, may be located in smaller cities near resource inputs where land and labor costs are also low. Yet the firm's management and administration may outsource a variety of business services such as advertising, specialized legal services, and the like (Ono, 2003). A firm may find it advantageous to locate headquarters and administrative functions in a large metro area with better availability, quality, and diversity of business and financial services, away from its manufacturing production facilities. "Functional specialization" across cities is modeled in Duranton and Puga (2005).

Third, management may require highly educated white-collar workers, who may have thin labor markets in small cities where production is concentrated. Differences in input requirements between management and production and the difference in input availability across space may result in the physical separation of management and production.

CAOs may also benefit more from agglomeration effects in large cities than production facilities, gathering information from other CAOs and from service firms (Davis and Henderson, forthcoming; Lovely, Rosenthal, and Sharma, 2005) about market conditions. Moreover, a firm may want a CAO "representative" in large markets, to market the firm.

However, a stand-alone management office requires fixed costs. Small firms may not have sufficient scale to justify having a separate management and administrative office. A separate CAO can also incur communication and monitoring costs (Holmstrom, 1979). Advances in transportation and communication technology have made it easier to operate in multiple locations, but it is more costly to monitor and communicate from a distance.

While the first-order effect of separation may be higher communication costs within the firm, it may be advantageous to separate management from production. Cremer (1995) shows that a firm may choose a lower level of monitoring because more accurate monitoring reduces agents' incentives to work hard to signal high ability. See also Aghion and Tirole (1997). A firm's choice of a lower level of monitoring could be reflected in its decision to physically separate administration from production. According to Puga and Trefler (2002), Boeing believes that separation encourages local managers to take initiative. Eccles (1985) discusses the related idea that suspicion among plants about unfair treatment can corrupt a firm's incentive schemes. Physical separation of headquarters from all production plants could signal impartiality. (1)

Our goal is to document some facts as a baseline for future theoretical and empirical work. We present a variety of evidence on the nature and roles of CAOs. CAOs account for about 2.5 million workers, but fewer than 5% of US firms have CAOs, and they tend to be large firms. Among the set of large firms, we explore what firm characteristics and organization determine whether a firm has a CAO.

In Section 2, we describe the data. In Section 3, we present various facts about CAOs' roles and firms that have CAOs. In Section 4, we examine the relationship between the likelihood of having CAOs and firm characteristics. In Sections 5 and 6, we describe where CAOs are located. Section 7 concludes.

2. Data

Our main data set is microlevel data from the Auxiliary Establishment Surveys (AES) compiled and organized by the US Census Bureau. The AES is a census performed every 5 years that captures the activity of all auxiliary establishments. Auxiliary establishments do not perform production or transaction activities; they manage, service or support the activities of, and are physically separate from, other establishments in the firm.

Auxiliary establishments are classified by type, such as central administration, R&D, computer data processing, communications, central warehouses and trucking. We want to identify establishments engaged primarily in central administration and that are thus CAOs. The survey asks questions about an establishment's primary function, as well as a breakdown of employment by function. Before 1997, the function question had many missing values, while the employment function questions almost universally recorded responses. Many auxiliaries self-classify, but in any census, about 1/3 are classified by surveyors. Before 1997, this classification (as opposed to self-reported) was applied to the employment function questions, where a high proportion of responses are bunched at a single number, such as 100% of employees all in management (probably due to imputation). For AES census years before 1997, we treat the surveyors' classification based on "inside knowledge" as accurate and define a CAO as an establishment for which the joint category of management, administrative and clerical employees dominates each of the other employment categories. In contrast, the 1997 data set has missing values for employment by function but complete data for the auxiliary's function, indicating that surveyors then used function to classify auxiliaries. (2) Thus, for 1997, we base our definition on the question regarding the main function reported by each auxiliary establishment. (3)

CAOs in 1997 are auxiliaries identified as a "corporate, subsidiary, or regional managing office or office of a holding company, providing a range of services to other establishments of the enterprise such as long term strategic and organizational planning, financial management, payroll and personnel management, centralized billing, advertising, and public relations" (US Census Bureau, 2001b), which corresponds with the definition of NAICS industry 551114. Establishments with managerial responsibility are distinguished from what we call "back offices" in the sense that "the establishment primarily engaged in providing a range of day-to-day office administrative services, such as financial planning, billing and recordkeeping, personnel, and physical distribution of logistics are classified in NAICS 56111 (office administrative services)" (US Census Bureau, 2001b). While CAOs have a specific definition, the question is whether CAOs perform the same functions that people have in mind in using the nontechnical term headquarters (HQs). Before 1997, the AES survey asked whether an auxiliary was an HQ. We report some information about these self-identified HQs, but they are problematic. The definition of HQs was not given in the survey, and in 1992, about 30% of auxiliaries surveyed did not answer this question. (4)

The other important data set for our study is the Standard Statistical Establishments List (SSEL), the master register of all private establishments in the United States used by the Census Bureau to conduct the economic censuses as well as to draw samples for various firm and establishment surveys. The SSEL includes basic information such as location, industry and total employment. We restrict attention to multiestablishment firms in the SSEL that have at least one production plant plus at least one other establishment of any kind (plants, sales offices, auxiliaries, retail outlets, etc.). (5) Establishments under common ownership or control, as self-reported in the Company Organization Survey, share a common firm identifier in the SSEL, which can be used to construct firm-level characteristics. (6) This information is then linked to CAOs in the AES file using this firm identifier. We record by firm the main production activity (industry), degree of industrial diversification, the geographical relationship among production facilities and CADs, the location where the main production activity is performed and so forth. (7)

3. What are CAOs and what do they do?