This dissertation studies the market microstructure of over-the-counter markets. All three chapters are based on the municipal (tax-exempt) bond market. However, the implications on price formation, price discovery and dealer's market power can be generalized into the other OTC markets such as corporate bond and derivatives markets as well.

In the first chapter "Price Clustering and Dealer's Market Power'', I find that prices, yields and markups in the municipal bond market are very clustered at rounded numbers. The level of rounding is higher in trades between dealers and customers than in inter dealer trades. Rounding is also more prominent in trades of retail sizes. I attribute the heterogeneous rounding observations to the dealers' market power. Such discriminative practice is described in a sequential bargaining model; and the prediction of the model is statistically verified in the data. Dealer's profit is indeed monotonically increasing with the pricing grid, and dealers are indeed rounding prices asymmetrically. They round price up when they are selling, round price down when they are buying. Thus, unlike the findings in more competitive financial markets; where the benefits of rounding is shared by both sides of the trades, rounding solely benefits the dealers and constitutes additional transaction cost to investors in the tax-exempt market.

In the second chapter "Agreeing on a Higher Price—Investor Rationality Test in Tax-Exempt Bond Market". I find that buyers in the tax-exempt bond market are bounded rational. They do not fully consider that tax consequences of their trades when purchasing bonds in the secondary market at prices at or just below a de minimis threshold; which are associated with higher future taxes. Retail investors again, are more likely to make sub-optimal decisions. Buyers are losing by not agreeing on a higher price that is more tax-favorable. The result is robust to the specification of tax rates and investment horizons. The paper calls for better investor education and simplification of tax law.

In the third chapter "Price Discovery in Illiquid Markets", coauthored with Rick Green and Norman Schurhoff; we find that municipal bond prices react to information sluggishly. Price discovery following macro-economic information releases lags that of the treasury market. Volume does not change significantly on or after news release days. Granger causality tests suggests one-way causality from the treasury to the municipal market, not vice versa. Municipal bond prices also adjust asymmetrically to changes in fundamental values. Dealers can exercise local market power by holding prices unchanged when bond values change unfavorably. They adjust prices promptly when bond values change favorably. Thus, price discovery in this illiquid OTC market is delayed, inefficient and biased.

All of the research would not be possible without the recently available high frequency transaction data collected by the MSRB (Municipal Securities Rule-making Board). The post-trade transparency has lifted the veil off a large and under-researched market, and enabled economists to investigate it at the "atomic" level. I hope the three chapters in this dissertation contribute to the ongoing effort of many in seeking better understanding of the over-the-counter markets, and help us form a more holistic view of various forms of financial markets overall.