Estimating the full effect of a partially anticipated event: A market-based approach applied to the case of TLTROIII

This paper presents an event-study methodology that combines market data and survey-based probabilities to infer the full effect of a policy decision, as seen through the lens of financial markets. The market reaction to an event's outcome reflects its surprise or announcement effect, and generally not its full effect. However, under certain conditions, the unobserved full effect can be derived from the observed surprise effect. Most importantly, the ex-ante probabilities of different outcomes must be known. We apply this methodology to a real-world example: the European Central Bank's announcement of its third series of targeted longer-term refinancing operations (TLTROIII). The introduction of TLTROIII was highly anticipated, and therefore partially priced in, as market participants feared a "cliff effect" with the preceding operations under TLTROII coming due. We estimate the announcement's full effect, focusing on its impact on a set of asset prices, as compared to a baseline wherein TLTROIII would not have been introduced. The full market impact surpasses the surprise effect by a factor of fifteen. We also find that the announcement had a highly heterogeneous impact on euro area sovereign bond yields.