Periodic Reporting for period 2 - DSGE-RD (Estimating a DSGE Model with Rare Disasters)
Periodo di rendicontazione: 2019-08-01 al 2020-07-31
The global financial crisis in 2008 was followed by a severe economic recession, which is now called "The Great Recession". The Great Recession had three unique properties: First, it was unusually large in terms of the size of overall economic contraction. Second, by various measures (e.g. unemployment rate), the recovery period was particularly long, lasting roughly a decade. Third, standard monetary policy was useless, because interest rates fell to zero, so central banks could not lower their interest rates furthermore.
Before the crisis, the dominating macroeconomic theory focused on small economic fluctuations around a stable trend, generally called "Business Cycle Models". These models were used to understand regular cycles of booms and recessions, but they failed to explain the causes and consequences of large macroeconomic shocks such as the Great Recession.
The objective of the present project is to develop new models that are more suitable for studying large macroeconomic recessions. These models are inherently nonlinear. Namely, they generate different economic effects for large versus small shocks. For instance, the central bank can lower the interest rate in response to a small economic shock, but when the shock is too large the interest rate hits the zero bound and can no longer be used to stimulate the economy.
This type of nonlinearity is associated only with large shocks. Importantly, it introduces new computational challenges that have not yet been fully resolved. The traditional tools applied to the standard business cycle models were based on linear techniques. These methods are not suitable for models that feature large shocks such as the Great Recession.
The goal of this project is to develop new methodological tools that would enable to study macroeconomic models with large shocks. The project extends our previous work to two types of nonlinearities that have played a role in the Great Recession. The first kind of nonlinearity is called “Regime-Switching Parameters”. This concept refers to a temporary change in the model parameters that lasts for a certain period and then changes back. For instance, a decline in GDP growth for the duration of the recession and then a recovery of the growth rate to the pre-crisis level. The second form of nonlinearity is generated by “Occasionally Binding Constraints”. The most common example is a lower bound that restricts the interest rate from falling below zero. This constraint is ignored in regular times when the interest rate is high. However, in severe recessions the interest rate may fall to zero. Then, the constraint kicks in, making monetary policy useless.
The methodological contributions of this project would allow economists to study unusual economic conditions such as those prevailing in the Great Recession. Hopefully, our understanding of severe recessions would improve our economic system by designing policy tools to prevent those events and mitigate their impact once they occur.
Before the crisis, the dominating macroeconomic theory focused on small economic fluctuations around a stable trend, generally called "Business Cycle Models". These models were used to understand regular cycles of booms and recessions, but they failed to explain the causes and consequences of large macroeconomic shocks such as the Great Recession.
The objective of the present project is to develop new models that are more suitable for studying large macroeconomic recessions. These models are inherently nonlinear. Namely, they generate different economic effects for large versus small shocks. For instance, the central bank can lower the interest rate in response to a small economic shock, but when the shock is too large the interest rate hits the zero bound and can no longer be used to stimulate the economy.
This type of nonlinearity is associated only with large shocks. Importantly, it introduces new computational challenges that have not yet been fully resolved. The traditional tools applied to the standard business cycle models were based on linear techniques. These methods are not suitable for models that feature large shocks such as the Great Recession.
The goal of this project is to develop new methodological tools that would enable to study macroeconomic models with large shocks. The project extends our previous work to two types of nonlinearities that have played a role in the Great Recession. The first kind of nonlinearity is called “Regime-Switching Parameters”. This concept refers to a temporary change in the model parameters that lasts for a certain period and then changes back. For instance, a decline in GDP growth for the duration of the recession and then a recovery of the growth rate to the pre-crisis level. The second form of nonlinearity is generated by “Occasionally Binding Constraints”. The most common example is a lower bound that restricts the interest rate from falling below zero. This constraint is ignored in regular times when the interest rate is high. However, in severe recessions the interest rate may fall to zero. Then, the constraint kicks in, making monetary policy useless.
The methodological contributions of this project would allow economists to study unusual economic conditions such as those prevailing in the Great Recession. Hopefully, our understanding of severe recessions would improve our economic system by designing policy tools to prevent those events and mitigate their impact once they occur.
The first task performed during the outgoing phase was the development of a computational algorithm that would apply to the case of regime-switching parameters. This new algorithm has been successfully developed and tested on a number of models. It is implemented by a new computer code written by the PI (available at the project website). The code is accompanied by a new user guide that demonstrates its performance on economic models with large shocks that exhibit regime-switching parameters.
The PI has organized two workshops on this new codes, one at the University of Pennsylvania and the other at the Federal Reserve Bank of Philadelphia. In these workshops, Phd students and staff members of the Federal Reserve learned how to use the new algorithm. In addition, the PI presented his work at the ASSA annual meeting 2018, The Wharton School (UPENN), IDC Workshop, Penn-Wharton Budget model and McGill University.
The second goal of this project was to extend the algorithm to models with occasionally binding constraints. The PI has been adopting tools from continuous-time models, which are more suitable for this type of nonlinearity. To this end, the PI worked to adjust the existing code to continuous time models with occasionally binding constraints. The final computer code was tested successfully on a complex model with an occasionally binding constraint. The code and proper documentation will be posted soon on the project website.
The third goal of the project was to build and solve a DSGE model with a financial crisis. This was done during the third year of the project. The model is an extended version of He and Krishnamurthy (2012,2013) models. It merges a real business cycle model with a financial crisis. The model is set in continuous time and solved by the algorithm developed during the outgoing phase of the project. The codes and results are posted on the project website.
The PI has organized two workshops on this new codes, one at the University of Pennsylvania and the other at the Federal Reserve Bank of Philadelphia. In these workshops, Phd students and staff members of the Federal Reserve learned how to use the new algorithm. In addition, the PI presented his work at the ASSA annual meeting 2018, The Wharton School (UPENN), IDC Workshop, Penn-Wharton Budget model and McGill University.
The second goal of this project was to extend the algorithm to models with occasionally binding constraints. The PI has been adopting tools from continuous-time models, which are more suitable for this type of nonlinearity. To this end, the PI worked to adjust the existing code to continuous time models with occasionally binding constraints. The final computer code was tested successfully on a complex model with an occasionally binding constraint. The code and proper documentation will be posted soon on the project website.
The third goal of the project was to build and solve a DSGE model with a financial crisis. This was done during the third year of the project. The model is an extended version of He and Krishnamurthy (2012,2013) models. It merges a real business cycle model with a financial crisis. The model is set in continuous time and solved by the algorithm developed during the outgoing phase of the project. The codes and results are posted on the project website.
The present project is expected to deliver new computational tools for macroeconomic models with large shocks. The project extends previous methodological work to new types of nonlinearities, which are extremely difficult to address with the existing computational methods. The new tools will allow to study economic models with regime-switching parameters and occasionally binding constraints. Hopefully, the new tools would allow economists to study extreme episodes such as the Great Recession. To fully understand the causes and consequences of these episode, we need a new generation of economic models that can feature strong nonlinearities. This project makes a methodological contribution in this direction.