I present a model of optimal learning for creativity generation in a field. The field is defined as a knowledge structure. A creative contribution is based on combining two previously unconnected elements, generating a new element. Individuals may possess private information/intuition about new combinations. Individuals contribute the maximum creative contribution they are able to produce based on the set of elements they have learned. In equilibrium individuals learn different elements, with some degree of overlap, effectively pursuing distinct creative interests. The model illustrates a general approach for analyzing the creative development of fields and, in a broader sense, cultural innovation.

Data from more than 1,000 inspections to the Nuclear Regulatory Commission form the basis for an investigation into the nature of safety regulation at U.S.commercial nuclear reactors. Poisson (and binary choice) models of the rate of occurrence of violations during each inspection period are specified and are extended to control for nondetection and for the possibility that violations persist from one inspection to the next. These models are used to study the factors associated with noncompliance, relative rankings of plants according to propensity to violate, the variation in detection rates among NRC inspectors, and the relationship between undetected violations and abnormal occurrences.

Data from more than 1,000 inspections to the Nuclear Regulatory Commission form the basis for an investigation into the nature of safety regulation at U.S.commercial nuclear reactors. Poisson (and binary choice) models of the rate of occurrence of violations during each inspection period are specified and are extended to control for nondetection and for the possibility that violations persist from one inspection to the next. These models are used to study the factors associated with noncompliance, relative rankings of plants according to propensity to violate, the variation in detection rates among NRC inspectors, and the relationship between undetected violations and abnormal occurrences.

Historical evidence indicates that hyperinflations can disrupt individuals' normal trading patterns and impede the orderly functioning of markets. To explore these issues, we construct a theoretical model of hyperinflation that focuses on individuals and their process of economic exchange. In our model buyers must carry cash while shopping, and some transactions take place in a decentralized setting in which buyer and seller negotiate over the terms of trade of an indivisible good. Since buyers face the constant threat of incoming younger (hence richer) customers, their bargaining position is weakened by inflation, allowing sellers to extract a higher real price. However, we show that higher inflation also reduces buyers' search, increasing sellers' wait for customers. As a result, the volume of transactions concluded in the decentralized sector falls. At high enough rates of inflation, all agents suffer a welfare loss.

Historical evidence indicates that hyperinflations can disrupt individuals' normal trading patterns and impede the orderly functioning of markets. To explore these issues, we construct a theoretical model of hyperinflation that focuses on individuals and their process of economic exchange. In our model buyers must carry cash while shopping, and some transactions take place in a decentralized setting in which buyer and seller negotiate over the terms of trade of an indivisible good. Since buyers face the constant threat of incoming younger (hence richer) customers, their bargaining position is weakened by inflation, allowing sellers to extract a higher real price. However, we show that higher inflation also reduces buyers' search, increasing sellers' wait for customers. As a result, the volume of transactions concluded in the decentralized sector falls. At high enough rates of inflation, all agents suffer a welfare loss.

We address the question of how the threat of antitrust enforcement spills over from one industrial submarket to another. We build a formal model describing the technology through which an initial antitrust indictment can reveal evidence of additional infractions in the same industry. Our approach is organized around the belief that the flow of information on antitrust violations is closely linked to commodity flows. As it turns out, our method proves able to explain the differential response of highway construction cartels in various states to recent changes in the severity of antitrust sanctions.

We present a theoretical and empirical analysis of the behavior of a bidder's cartel in a multiperiod auction market in which the purchaser is relatively uninformed. Our work establishes a tentative connection between the economics of information and collusion by concentrating on the cartel's informational monopoly, and its ability to both increase profits and mask its presence by passing misinformation to the purchaser. In the models we develop, costs are assumed to be stochastic, and projects awarded in adjacent periods to be substitutes for one another. As a result, the purchaser will have incentive to intertemporally reallocate his demand, buying more or less of the current period's project, depending upon whether the current market price is lower or higher than the price which he expects to prevail in the future. A purchaser who needs to form such price expectations, yet is ignorant about production costs and market structure, will be particularly attracted to the auction mechanism, because the data contained in past auctions' bids can apparently be incorporated into an explicit structural model of expected future price. In fact, such a purchaser may hold auctions frequently merely to acquire information. Such pleasant appearances are deceiving. Once the auction's bidders become aware of the informational use to which their bids are being put, they will have incentive to form a cartel, thereby extracting a premium for their knowledge by misinforming the purchaser and skewing his intertemporal decision making to their advantage. In some respects the mechanics of our model resemble the case of a multiproduct monopolist selling goods that are substitutes for one another. What is unique to our analysis is the dependence of one of the goods, future demand, on price expectations, which are an information resource. We are suggesting that information that is distributed asymmetrically among market participants is susceptible to the inefficiencies of collusion. In some ways the implications of our work are considerably stronger than this, because our cartel never reveals its true knowledge, and thus our market may be more inefficient than more standard monopolies, in which the good is always traded, albeit at noncompetitive prices. Furthermore, by explicitly modeling the purchaser's rational structural approach to price forecasting, we point out that complete rational expectations models may be more susceptible to exploitation by informational cartels than simpler forecasting techniques. We have tested the implications of our model in a case study of North Carolina highway construction cartels. The highway construction industry is an auction market characterized by government procurement agencies who are uninformed relative to suppliers, and comes complete with a substantial record of known collusion in recent years. Thus it seems an ideal hunting ground for verification of our theoretical findings. Our paper proceeds as follows. In Section I, we model a multiperiod competitive auction market, and describe a Bayesian mechanism through which the purchaser uses past data to improve the accuracy of his price expectations. Section II establishes theoretical results for the bidders' cartel which has incentive to form in this market: we describe *Massachusetts Institute of Technology, Cambridge, MA 02109 and University of Arizona, Tucson, AZ 85721, respectively. Nold was formerly of Rhodes Associates and is now deceased. We thank Richard Schmalensee and the participants of the Economic and Legal Organization Workshop at the University of Chicago for valuable suggestions. Part of this research was supported under research grant 81-IJ-CR0062 from the National Institute of Justice.