Over the past several decades, market-based approaches to natural resource management have received increased attention as a means to cost-effectively achieve environmental quality goals. Following on what has been hailed a success for reducing air pollution, water quality trading (WQT) has more recently been seen as the next great opportunity for reducing water pollution, especially for nutrient loading. Numerous trading programs have been pilot tested and/or adopted in states throughout the nation, with more than 70 programs now in operation (Breetz et al., 2004). WQT would allow multiple contributors to surface water degradation to determine how best to meet an overarching collective goal related to pollution reduction. WQT takes advantage of differences in pollution abatement costs. In the case of point/nonpoint source trading, such as between wastewater treatment plants (WWTPs) and agricultural producers, it is often the agricultural producers who can achieve a given level of nutrient reduction at less cost through their adoption of various best management practices that reduce sedimentation and nutrient loading to surface waters. Trading would allow WWTPs to purchase “credits” generated by producers who reduced their pollution loading to achieve an equivalent level of reduction as might be required by a regulatory discharge permit at a lower overall cost. While there is substantial evidence that nonpoint sources have lower nutrient reduction costs than point sources, experience with WQT reveals a common theme: little or no trading activity. The success of WQT seems, in part, to depend on the structure of the market created to bring buyers and sellers together to transact exchanges. These outcomes suggest the presence of obstacles to trading that were not recognized in the design of existing programs. To examine the ways that various market imperfections may impact the performance of a WQT market, an agent-based model was constructed which simulated a hypothetical point-nonpoint market. In particular, the market was modeled using a variant of the sequential, bilateral trading algorithm proposed by Atkinson and Tietenberg (1991). Our proposed paper first presents an overview of the simulation modeling technique and then analyzes the effects of two prominent market impediments identified in the WQT literature: information levels and trading ratios. Information levels refer to buyers’ and sellers’ knowledge of each others’ bid prices. A frictionless WQT market would be one where all of the potential buyers (i.e., point sources) would know all of the sellers’ (i.e., nonpoint sources) offer prices and vice versa. In this full information environment, we can expect that trades would be consummated in the order of their gains. That is, first buyers and sellers to be paired together for trading would be the buyers with the highest offer prices and the sellers with the lowest bid prices. Successive trades will have successively smaller gains until the gap between bid and offer prices reaches zero. This is the textbook Walrasian market and would closely approximate a double auction institution, where all buyers and sellers submit their offers and bids, which are then sorted and matched by a centralized market manager. While the full information scenario serves as a useful benchmark, most existing WQT markets are decentralized in nature, so that limited information causes traders to be matched in a less efficient sequence. A variety of information levels are possible. One side of the market may have more information than the other (limited information) or neither side having any knowledge of the other side’s bid or offer prices (low information). Each of these scenarios leads to a different sequencing of trades. This paper analyzes the effect of different information levels on market performance. Market performance is measured in terms of cost savings, the number of credits traded, and the average reduction costs under different information scenarios. Trading ratios are a common component of many existing WQT programs. A typical trading ratio of 2:1 requires a nonpoint source to reduce two pounds of expected nutrient loading in order to receive one pound of trading credit. These ratios serve as a “safety factor” and are incorporated to account for the uncertainty in the measurement and monitoring of nonpoint source loading. Because nonpoint traders must reduce loading by 2 pounds for every 1 pound emitted by point source traders, there will be a net reduction of 1 pound of expected loading for each trade. So, while inhibiting some trades from ever occurring, trading ratios also have the potential to improve water quality beyond trading with a 1:1 trading ratio. This paper examines these tradeoffs in terms of effects on market performance and then describes procedures that can be used to characterize an optimal trading ratio if one exists. Because WQT programs, by nature, involve complex interactions between economics and the biophysical world, accurately simulating a real-world WQT market requires at minimum a basic understanding of the types of data that watershed models can provide. This paper concludes by briefly discussing data requirements, points of consideration, and integrative techniques used in the simulation of WQT in real-world watersheds.