Open Access

Johannes Huber

• Going back to the seminal work of Kydland and Prescott (1982) and Long and Plosser (1983), over the past four decades, the class of dynamic stochastic general equilibrium (DSGE) models has become one of the dominant analytic frameworks in modern macroeconomics. For instance, Glandon et al. (2022) find that about 42 percent of the “theory-centered” articles published in five leading macroeconomic field journals – the Journal of Monetary Economics, the Journal of Money, Credit and Banking, the American Economic Journal: Macroeconomics, the Journal of Economic Dynamics and Control, and the Review of Economic Dynamics – from 2016 to 2018, employ DSGE models. Part of the reason for the growing interest in DSGE models is that they provide a consistent framework for analyzing the impact of macroeconomic policy changes. Thereby, as Fernández-Villaverde and Guerrón-Quintana (2021) emphasize, “[t]he class of DSGE economies is not defined by a particular set of assumptions, but rather by anGoing back to the seminal work of Kydland and Prescott (1982) and Long and Plosser (1983), over the past four decades, the class of dynamic stochastic general equilibrium (DSGE) models has become one of the dominant analytic frameworks in modern macroeconomics. For instance, Glandon et al. (2022) find that about 42 percent of the “theory-centered” articles published in five leading macroeconomic field journals – the Journal of Monetary Economics, the Journal of Money, Credit and Banking, the American Economic Journal: Macroeconomics, the Journal of Economic Dynamics and Control, and the Review of Economic Dynamics – from 2016 to 2018, employ DSGE models. Part of the reason for the growing interest in DSGE models is that they provide a consistent framework for analyzing the impact of macroeconomic policy changes. Thereby, as Fernández-Villaverde and Guerrón-Quintana (2021) emphasize, “[t]he class of DSGE economies is not defined by a particular set of assumptions, but rather by an approach to the construction of macroeconomic models.” Within this approach, the researcher is forced to formulate the assumptions underlying his analysis in a clear and transparent manner. On the one hand, this transparency marks DSGE models as a frequent target for critics (see, e.g., Stiglitz, 2018). On the other hand, this transparency also prepares the ground to address their shortcomings by adding new features to the DSGE framework. Another reason for the growing interest in DSGE models is the progress in developing new algorithms to evaluate and estimate these models. Most notable here is the tremendous evolution of Bayesian estimation methods in the wake of the seminal articles by Smets and Wouters (2003, 2007), which established DSGE models as an essential part of the monetary policy analysis in central banks around the world. However, since DSGE models typically cannot be assessed analytically, these models’ accurate evaluation and estimation can become a time-consuming computational task. This thesis addresses how to estimate and evaluate DSGE models efficiently. The challenging aspect of an efficient evaluation and estimation of DSGE models is the trade-off between ensuring that the model’s evaluation and estimation are sufficiently accurate, on the one hand, and that it is computationally feasible, on the other. The details of this trade-off depend on the research question at hand. The three essays in this thesis consider different approaches and procedures for a more efficient evaluation and estimation of DSGE models. Thereby, the essays cover linear and non-linear solution techniques, as well as likelihood-based and limited-information estimation methods. Chapters 2 and 3 focus on the likelihood-based estimation of DSGE models, employing linear solution techniques to determine the model’s policy function. In Chapter 2, we address how to efficiently evaluate the likelihood of those models in terms of computational time required. Chapter 3 proposes a fast and reliable procedure for the maximum-likelihood estimation (MLE) in the context of a business cycle accounting (BCA) application in the spirit of Chari et al. (2007). Chapter 4 discusses a method known as the generalized polynomial chaos expansion (PCE) to obtain a (point-wise) approximation of a model’s quantity of interest (QoI) in terms of a series expansion of the model’s parameters. Analyzing the suitability of the PCE for DSGE models, we extend our analysis to non-linear solution techniques for the model’s policy function and limited-information methods to estimate its parameters.…