A Random Attention Model: Identification, Estimation and Testing

显示性偏好理论是现代经济学和其他社会与行为科学的基石之一。然而，它假设家庭、政治家或公司等决策者充分考虑所有选择，这在许多情况下是不可能的。因此，本项目旨在通过允许决策者有限地关注他们的选择来改进显示偏好理论。研究者将联合收割机结合理论和计量经济学，在有限的关注下发展可检验的决策理论和实证实施。该项目的研究结果将有助于政策制定者和社会科学家更好地理解决策过程，并进一步制定更好的政策或干预措施，以适应劳动力市场，选举或行业等不同环境。研究人员将首先开发一个随机有限注意力模型，兼容一大类不同的决策过程。该模型将全面推广经典的显示性偏好理论，以及先前介绍的确定性有限注意框架。本计画将在此模型基础上，获得具体可验证的意涵，并运用现代计量经济学与统计技术，建构具有良好有限样本性质的推论程序。非参数识别，估计和推理方法将进行详细研究。作为这个项目的一部分，方法引发未观察到的，异质性偏好的随机有限注意力模型的背景下，将提出和实施使用真实的经验数据。

Revealed preference theory is one of the cornerstones of modern economics and other social and behavioral sciences. However, it assumes that decision makers such as households, politicians or firms take all choices into full consideration, which is unlikely in many situations. This project thus aims to improve upon revealed preference theory by allowing decision makers to pay limited attention to their choices. The investigators will combine both theory and econometrics to develop testable theory and empirical implementations of decision-making under limited attention. The results from this project will help policy makers and social scientists to better understand the decision-making process and further develop better policies or interventions in different settings such as labor markets, elections, or industries.The investigators will first develop a random limited attention model, compatible with a large class of different decision-making process. This model will generalize classical revealed preference theory under full attention as well as previously introduced deterministic limited attention frameworks. Based on this model, this project will obtain concrete testable implications, and will employ modern econometrics and statistical techniques to construct inference procedures with good finite samples properties. Nonparametric identification, estimation and inference methods will be studied in detail. As part of this project, methods for eliciting unobserved, heterogeneous preferences in the context of the random limited attention model will be proposed and implemented using real empirical data.