Theoretical analysis characterizes optimal mechanisms for a variety of problems and environments such as the construction of optimal incentive schemes, auctions, regulation, optimal taxation, public good provision and more. Optimal mechanisms are, in general, quite different from the mechanisms that are relied on to solve these problems in practice. This gives rise to two related questions: (1) What accounts for the prevalence of the mechanisms that are observed in practice? and (2) Why aren't the optimal mechanisms that are identified by the theory used in practice? The project consists of two parts corresponding to these two questions. Both rely on a single methodological strategy: A general argument is presented and is investigated in a specific context. What accounts for the prevalence of the mechanisms that are observed in practice? The general answer is that existing mechanisms are simple, and thus easy to apply, and, while perhaps sub optimal in many environments, are highly effective in all environments. (The effectiveness of a mechanism is defined as the ratio between what is achieved by the mechanism and what could be achieved by an ideal mechanism. Simplicity is defined as independence of the mechanism from the details of the specific environment.) This argument is elaborated in the context of auction design. The research focuses on single-object, private-values auctions, and identifies worst case bounds on the effectiveness of the most commonly used auction form the English auction. This lower bound is surprisingly high, rendering the English auction universally effective. For example, when there are 10 bidders whose valuations for the object are positively correlated and when their expected valuations of the object are at least as high as 60% of the maximal possible valuation an English auction sells the object at a price that is at least 83% of its value to the bidder with the highest valuation (total surplus). If bidders' valuations are independent of each other, worst case effectiveness increases to 96%. It is conjectured that similar arguments can be given to justify the prevalence of other commonly used mechanisms as well as to the use of `rules of thumb` and other modes of boundedly rational behavior. Why aren't the optimal mechanisms that are identified by the theory used in practice? The general answer is that the so-called optimality of optimal mechanisms rests on a number of subtle assumptions that are unlikely to be satisfied in practical situations. Arguably the most important such assumption is that there exists a one to one relationship between agents' valuations or payoffs, and agents' beliefs. When this assumption is relaxed, the effectiveness of optimal mechanisms is dramatically affected. In contrast, the effectiveness of the simpler mechanisms that are used in practice is hardly influenced. The consequences of relaxing this assumption are investigated in the context of designing an optimal auction and arranging for the voluntary provision of public goods. It is shown that, when this assumption is relaxed, even in highly correlated environments, full information rent extraction (as in Cremer and McLean (1985, 1988) and McAfee and Reny (1992)) becomes impossible.
