AUTONOMY IN CHINESE STATE ENTERPRISES in the reform period of the 1980s. Before the reforms, industrial productivity had been almost stagnant. Total factor productivity grew at an annual rate of only 0. 4 percent between 1957 and 1978 (according to Chen, Hongchang, Wang, Zheng, Yuxin, Jefferson and Rawski [1988]), but this changed after the reforms began Between 1978 and 1985 industrial productivity grew at an annual rate of 4.8 percent [Chen et al. ] and it continued to grow strongly after 1985. For the firms in our sample, between 1980 and 1989 total factor productivity rose at an average annual rate of 4.5 Not all of this improvement in productivity is attributable to the particular reforms we investigate here. A large number of reforms were introduced in gradual and piecemeal forms. Changes in behavior were the result of the total impact of these incremental reforms. An important source of gains is the extra discipline resulting from the increased product-market competition that these firms have faced, both from other state firms and from new nonstate firms [McMillan and Naughton 1992]. Gains also came from better methods of selecting managers and from linking managers' pay and career prospects to their firms'performance [Groves, Hong, McMillan, and Naughton 1993a]. The reforms we analyze here are only part, but an important part, of a broad I. THE COSTS OF HIERARCHY In order to learn about costs so as to decide on appropriate output quantities, and in order to learn how to organize production o as to minimize costs the ultimate decision-maker (the central planner in a planned economy, or the firm s manager in a decentralized economy) must rely in part on information that comes from below. Information about how high costs are whether workers could be reassigned to increase productivity, whether ng held, what improvements tion techniques could feasibly be introduced how well newly introduced techniques are working, and so on must be gathere from workers and foremen. Information inevitably becomes dis- es from other data sets show similar increases le of twenty state enterprises, that productiv 978-1982