Nonparametric Identification and Inference in Duration Analysis

计量经济持续时间分析侧重于随机变量的风险率，这些随机变量捕捉了单位在某种状态下花费的时间，比如个人的失业持续时间。作为解释变量和持续时间的函数的风险率通常可以用经济理论来解释。通常情况下，一些解释变量可能未被观察到。这就产生了众所周知的动态选择问题，如果忽略它就会导致错误的推断。现有的识别和估计方法依赖于特殊的混合比例风险假设。在这个项目中，我们的目标是超越这些假设。我们将为模型开发识别结果，这些模型允许危险率决定因素之间的相互作用。我们特别关注治疗效果对持续时间结果的识别和推断。

Econometric duration analysis focuses on the hazard rate of random variables that capture the time that a unit spends in a certain state, like the unemployment duration of an individual. The hazard rate as a function of explanatory variables and the elapsed duration can usually be interpreted in terms of economic theory. Often it is likely that some explanatory variables are unobserved. This creates well-known dynamic selection problems which lead to incorrect inference if ignored. Existing identification and estimation approaches that deal with this rely on ad-hoc Mixed Proportional Hazard assumptions. In this project we aim to move beyond these assumptions. We will develop identification results for models that allow for interactions between determinants of the hazard rate. We pay particular attention to identification and inference of treatment effects on duration outcomes.