Industrial and Corporate Change, Volume 17, Number 1, pp. 155–195doi:10.1093/icc/dtm036Advance Access published December 19, 2007Where do transactions come from?Modularity, transactions, and theboundaries of firmsCarliss Y. BaldwinThis article constructs a theory of the location of transactions and the boundariesof firms in a productive system. It proposes that systems of production can beviewed as networks, in which tasks-cum-agents are the nodes and transfers—ofmaterial, energy and information—between tasks and agents are the links.Transactions are defined as mutually agreed-upon transfers with compensationand are located within the task network. Placing a transaction in a particularlocation in turn requires work to define, count (or measure), and pay for thetransacted objects. The costs of this work (labeled mundane transaction costs) aregenerally low at the thin crossing points of the task network, which correspond tomodule boundaries. Therefore, transactions are more likely to be located atmodule boundaries than in their interiors. Several implications arise from thistheory. Among these: Modularizations create new module boundaries, hence newtransaction locations. Areas in the task network where transfers are dense andcomplex should be located in transaction-free zones, so that the cost oftransacting does not overburden the system. The thin crossing points betweentransaction-free zones constitute breakpoints, where firms and industries maysplit apart.1. IntroductionFor the last 30 years, economists have used the concepts of “transaction,” “transac-tion cost,” and “contract,” to illuminate a wide range of phenomena, includingvertical integration, contract design, the allocation of property rights, and thestructure of industries. These concepts are now deeply embedded in the fields ofeconomics, sociology, business, and law. But although economists and managementscholars have explored the design of transactions in a wide variety of settings, in mostof this literature, it is assumed that a pre-existing division of knowledge and effortmakes a transaction possible at a particular place in the larger productive system. Thetheories explain how to choose between different forms of transactional governance,ß The Author 2007. Published by Oxford University Press on behalf of Associazione ICC. All rights reserved.but they almost never ask why the opportunity to have a transaction occurs where itdoes. As a result, the forces driving the location of transactions in a system ofproduction remain largely unexplored.The goal of this article is to explain the location of transactions (and contracts)in a system of production. To do this, the article brings together three strandsof literature—transaction cost economics and contract theory, knowledge-basedtheories of the firm, and modularity theory. Building on this work, I construct atheory of the location of transactions and the boundaries of firms by observingsystems of production at a deeper level. Not surprisingly, the results obtained at thisnew level of observation are wholly consistent with prior theories of the firm.However, as is common in science, there are also new constructs and unifyingprinciples that can only be seen by taking a deeper view of the phenomenon(Wilson, 1998).First, drawing on modularity theory, I look at systems of production as networks,in which tasks-cum-agents are the nodes and transfers—of material, energy andinformation—between tasks and agents are the links. At this new level ofobservation, transactions are not primitive units of analysis as they were forCommons (1934) and Williamson (1985). Transactions, defined as mutually agreed-upon transfers with compensation, are located within the task network and serve todivide one set of tasks from another. Local properties of this network, specifically itsmodularity, make transactions more or less costly in different locations. In particular,thin crossing points at the boundaries of modules have low transaction costs, whilethick crossing points in the interior of modules have high-transaction costs.Transactions are designed to match their locations. In a given location, the objectsbeing transacted must be defined and counted (or otherwise measured), and thepurchaser must compensate the supplier. Thus, work goes into making the tran-saction. I call the costs of this work mundane transaction costs to distinguish themfrom the opportunistic transaction costs that are the focus of analysis by Williamson(1985, 1991) and the contract theorists (Grossman and Hart, 1986; Hart and Moore,1990; Holmstrom and Milgrom, 1994). Expenditures on mundane transaction costs,I will argue, can reduce opportunistic transaction costs. Thus, to arrive at the optimalform of a transaction in a particular location, designers must make tradeoffs betweenthe two types of cost, taking into account the local modular structure of the tasknetwork, i.e., the thickness of the crossing point in question.The modular structure of the task network is not completely fixed, however.Modularizations create new thin crossing points where transaction costs are low.These new module boundaries provide points of entry for competitors and break-points, where vertically integrated firms and industries may split apart. Therefore,transaction locations are not technologically determined, but arise through theinterplay of firms’ strategies and knowledge and the requirements of specifictechnologies. Strategies, knowledge, and technologies all change over time, hence thelocation of transactions changes as well.156 C. Y. BaldwinThere are areas in the task network where transfers are dense and complex and notsubject to modularization. In general, it is not cost-effective to place transactions inthese locations: in terms of Coase (1937), the cost of “using the market” is too high.These areas, I will argue, should be located in transaction-free zones so that the cost oftransacting does not overburden the productive system. Transactions can be used tomove valuable goods into and out of transaction-free zones. Some transaction-freezones may be completely surrounded by transactions and thus encapsulated, creatingthe legal form of a modern corporation. Others zones may be unencapsulated, as isthe case for many online and open source communities.The theoretical arguments in this article support the following empiricalpredictions:1. Transactions are more likely to be found at module boundaries than in