Efficient Inference with Time-Varying Identification Strength

In the last two decades, there has been a lot of empirical evidence suggesting that many macroeconometric and financial models (e.g. for inflation, interest rates, or exchange rates) are subject to both parameter instability and identification problems. In this paper, we address both issues in a unified framework, and provide a comprehensive treatment of the link between them. Changes in identification strength provide an additional source of information that is used to improve estimation. More generally, we show that detecting and locating changes in instrument strength is essential for efficient asymptotic inference, and we provide a step-by-step guide for practitioners. In our simulation studies, our global inference procedures show very good size and power properties.