This paper develops a structural model of intraday price formation that incorporates both public information shocks and microstructure effects. The model aims to explain observed patterns in bid-ask spreads, price volatility, transaction costs, and autocorrelations of transaction returns and quote revisions. Key findings include: 1. **Information Asymmetry and Volatility**: Information asymmetry decreases throughout the day, while dealer costs increase, leading to a U-shaped pattern in bid-ask spreads. 2. **Transaction Costs**: The cost of trading is significantly smaller than the bid-ask spread, but it increases over the day, reflecting the costs of carrying inventory overnight. 3. **Autocorrelations**: The autocorrelation of quote returns can be positive or negative, depending on the model's parameters, and the model's implied autocorrelations closely match empirical data. 4. **Effective Trading Costs**: The model provides an estimator of execution costs that accounts for the possibility of executing within the bid-ask spread, which is smaller than the bid-ask spread and increases over the day. The paper uses transaction-level data from NYSE stocks to estimate the model, demonstrating that the parameters are economically reasonable and provide insights into price discovery and effective trading costs.This paper develops a structural model of intraday price formation that incorporates both public information shocks and microstructure effects. The model aims to explain observed patterns in bid-ask spreads, price volatility, transaction costs, and autocorrelations of transaction returns and quote revisions. Key findings include: 1. **Information Asymmetry and Volatility**: Information asymmetry decreases throughout the day, while dealer costs increase, leading to a U-shaped pattern in bid-ask spreads. 2. **Transaction Costs**: The cost of trading is significantly smaller than the bid-ask spread, but it increases over the day, reflecting the costs of carrying inventory overnight. 3. **Autocorrelations**: The autocorrelation of quote returns can be positive or negative, depending on the model's parameters, and the model's implied autocorrelations closely match empirical data. 4. **Effective Trading Costs**: The model provides an estimator of execution costs that accounts for the possibility of executing within the bid-ask spread, which is smaller than the bid-ask spread and increases over the day. The paper uses transaction-level data from NYSE stocks to estimate the model, demonstrating that the parameters are economically reasonable and provide insights into price discovery and effective trading costs.