Sequential Procurement Auctions

Procurement by governments accounts for 10-15 percent of national product in most Western nations [McAfee and McMillan (1985a)]. Consequently, analyses of procurement policies apply to a significant portion of economic activity. This paper considers sequential procurement: a buyer will purchase items in sequence, with costs correlated across time. Thus, if auctions are held, bids in the first auction reveal something about costs in the second. This paper, then, contributes to the optimal auction literature, with the major application in government procurement. It is arguable that nonsequential procurement is the exception; once a government buys an item, it continues buying it for the rest of time. Governments fix roads, buy military items, plant trees, etc. every year. These all represent sequentially procured items. Often the same firms bid each year, and this is presumed here for simplicity. Research and development projects are often followed by production projects. Generally, the winner of the R&D project (the incumbent) has an advantage in production, for he may have learned something useful during the research that he keeps secret, or because the mere fact that he won the R&D bidding is itself a signal of an advantage in this area. Such a situation means that the government possesses useful information: in the second project, expectations concerning the incumbent do not coincide with expect-