Essays in macroeconometrics under noninvertibility

This dissertation investigates a fundamental challenge in empirical macroeconomic research: the informational asymmetry between the researcher and the economy under study. This issue, known as noninvertibility, arises from the common observation that economic agents form expectations about the future path of economy and make decisions accordingly, well before those expectations are realized. As a result, while the economy responds in real time, the researcher only gains access to relevant information with a delay. This dissertation focuses on understanding and addressing this phenomenon, with particular emphasis on how it affects the estimation of monetary policy transmission to the economy. It consists of an introductory chapter that outlines the context and motivation, followed by three self-contained chapters.

The second chapter evaluates, through simulations, commonly used estimation methodologies for identifying the dynamic effects of conventional monetary policy in an economy where the central bank signals future interest rate changes. The primary motivation for this research stems from the use of forward guidance policies during the 2010s, when traditional interest rate policy was constrained by the zero lower bound. The study aims to assess the size and direction of the estimation bias, as well as the degree of estimation uncertainty associated with the researcher's failure to account for forward guidance. The main findings indicate that while popular instrumental variable methods accurately capture the direction and shape of the economy's response to a conventional monetary policy shock, they tend to overstate the magnitude of these responses. Moreover, the widely used approach of employing a composite instrument for the monetary policy shock, identified from high-frequency data, delivers reliable estimates of monetary transmission.

The third chapter examines the dynamic effects of two distinct monetary policy tools: the conventional short-term interest rate policy and the unconventional forward guidance policy. The objective is to identify these effects within a framework that enables the joint identification of shocks to both policies. The chapter introduces a novel identification strategy that accommodates the noninvertibility of shocks -- a theoretically grounded characteristic of forward guidance. This identification approach relies on distributional assumptions about the data, thereby extracting more information than standard methodologies. Applying this method to U.S. monetary policy, I find that forward guidance has a strong and persistent effect on prices, while conventional monetary policy accounts for a significant contraction in output. A key takeaway is that the two policy tools influence output over different time horizons: forward guidance exerts its maximum impact in the long run, whereas the effects of conventional policy are relatively short-lived.

Finally, the fourth chapter offers a methodological contribution to the identification of reduced-form dynamic factor models. These models have become increasingly popular in empirical macroeconomic research due to their ability to incorporate a much larger number of observed variables than standard approaches, thereby helping to address issues related to noninvertibility. The novelty of the proposed method lies in its ability to reduce the number of parameters in the model without compromising its capacity to capture the key features of high-dimensional macroeconomic data. Building on existing strategies for parameter space reduction in time series analysis, the chapter shows how these techniques can be effectively applied to dynamic factor models. A model selection procedure is proposed, and the performance of the parameter reduction strategy is evaluated using simulated data.

Contact Juho Koistinen

Email: juho.koistinen@helsinki.fi
Home page: https://labore.fi/author/juho-koistinen/