We extend the classic ”martingale-plus-noise” model for high-frequency returns to accommodate an error correction mechanism and endogenous pricing errors. It is motivated by (i) novel empirical evidence documenting that microstructure noise exhibits frequently changing patterns of serial dependence which are interwoven with innovations to the efficient price; (ii) building a bridge between high-frequency econometrics and market microstructure models. We identify temporal pricing error correction and noise endogeneity as complementary components driving high-frequency dynamics and inducing two separate regimes, characterized by the sign of the return serial correlation and an implied bias in realized variance estimates. We document frequent fluctuations between these regimes, which can be associated with price discovery in a setting with incomplete information and learning. The model links critical concepts from high-frequency statistics and market microstructure theory, suggesting new avenues for volatility estimation.

我们扩展了高频收益的经典“鞅加噪声”模型，以适应误差校正机制和内生定价误差。它的动机是 (i) 新的经验证据表明，微观结构噪声表现出频繁变化的序列依赖模式，这些模式与有效价格的创新交织在一起；(ii) 在高频计量经济学和市场微观结构模型之间架起一座桥梁。我们将时间定价误差校正和噪声内生性确定为驱动高频动态的互补成分，并引发两个独立的制度，其特征是回报序列相关的符号和已实现方差估计中的隐含偏差。我们记录了这些制度之间的频繁波动，这可能与信息和学习不完整的环境中的价格发现相关联。该模型将高频统计和市场微观结构理论中的关键概念联系起来，为波动率估计提供了新的途径。