these enterprises and the need to maintain them as social protection units, local governments first resist any attempts to declare bankruptcy, then seek subsidies for the plants (which are actually subsidies for the plants' social infrastructure) through their own channels. To illustrate, the manager of a large tank plant in Siberia that had not been receiving any defense orders had been begging the federal government to close the plant. He obtained some personal rents from its assets and wanted a safe retirement. It was the local government which threatened that if the manager stopped "fulfilling his social obligations" (i.e., maintaining the social infrastructure), it would engineer a comprehensive audit of the plant, with the objective of revealing abuses of managerial authority. This is an example of forced managerial entrenchment that illustrates the rapidly forming rent-seeking alliance between local government and enterprise management that is unable to adjust.
At this point it is worth carrying a step further the argument of Krueger (1974), who emphasizes that rent-seeking activity does not emerge out of nowhere: in addition to the distortions imposed by rent seeking, there is associated deadweight loss.12 Rent seeking is just one option in an enterprise's portfolio of adjustment responses. When performed outside the lobbying group/rent-seeking network, it is not even particularly efficient. Our case studies show that more learning-intensive options are often preferred. Once chosen, each adjustment often goes through a process of being perfected and improved and becomes self-reinforcing. As the formation of rent-seeking networks advances, the switch to restructuring options becomes increasingly unattractive: rent seeking crowds out restructuring. On the other hand, once restructuring options have been mastered, rent seeking becomes unattractive: learning-intensive restructuring crowds out rent seeking. Because of the cumulative nature of learning, early choices determine long-term outcomes.
Two policy implications follow from this analysis. First, until substantial progress in learning-intensive restructuring is made,13 the federal government should abstain from discretionary sectoral policies that are particularly prone to rent seeking. Such policies would encourage an early choice of restructuring options (which is likely to perpetuate itself) from a portfolio of adjustment responses. Second, the federal government should allot more specific and transparent subsidies to social infrastructure, thereby discouraging alliances between antireform enterprise managers and local authorities.
The distinction between rent-seeking and restructuring-oriented networks once again underlines the differences between rudimentary learning with little