Economic models are limited by many factors, both theoretical and practical. Three fundamental hurdles of economic models are the presence of parameter instabilities, weak identification and mis-specification. "Parameter instabilities" in a model are when the models' parameters vary over time. For example, the volatility of output growth has increased in the late 1980s; it is unclear whether this increase may be explained by changes in monetary policy, or by changes in the private sector behavior, or simply by unknown causes. If an economic model is compatible with observed data for two different parameter values it is known as "weakly identified;" this implies that, with the available data, the true model of the economy may not be estimated precisely enough. "Mis-specification" means that the economic model used to estimate the impact of economic policies may be a fundamentally incorrect representation of reality. The presence of instabilities, weak identification, or mis-specification may invalidate the results from an estimated model, leading to incorrect conclusions by researchers and to incorrect policy decisions by policy makers.
The recent literature suggests that instabilities, weak identification and mis-specification are widespread phenomena. Clearly, a successful understanding of the economy and the implementation of economic policy in practice depend on understanding the causes of changes in these parameters. They also crucially depend on using reliable models where any weak identification or mis-specification issues have been resolved. As a result, much effort has been devoted to designing new and improved tests to address these issues, and researchers have paid more attention to such tools in their empirical works. However, notwithstanding the recent developments, we still lack methods to identify the exact sources of instabilities, as well as tests that are robust to weak identification and mis-specification, with the result that researchers are currently left with little guidance on how to deal with these issues in practice.
The PIs propose new and useful econometric methods in these realistic and empirically relevant situations. In the first research project, "Identifying the Sources of Instabilities in Macroeconomic Fluctuations", the PIs propose new methods for searching for and detecting the sources of instabilities in the data. The PIs overturn conventional findings by showing that previous results in the literature depend on imposing stability constraints on ad-hoc sets of parameters. In the second research project, "Testing for Weak Identification in Macroeconomic Models", the PIs propose a new test for identification whose novelty is the robustness to the presence of weak identification. In the third research project, "Selection Criteria for IRFME and Other Estimators of DSGE Models", the PIs propose new methodologies that guarantee consistent and efficient estimation of models' parameters even in the presence of mis-specification.
Broader impact: These methodologies will provide practical guidance to applied researchers, economists and central bankers on how to deal with such issues in practice. This proposal describes important economic applications where these methods will be useful. The methods will be shared with practitioners within the scientific community and students alike, and will contribute to understanding the role of instabilities in the economy.
In this research project, the PIs developed econometric methods for addressing important issues in empirical macroeconomics, such as: (i) instabilities; (ii) misspecification; and (iii) lack of identification. They found several important insights. (i) Instabilities are an important empirical issue that economists confront in practice. In fact, macroeconomists analyze the economy and prescribe policy recommendations based on a macroeconomic model. Models describe how the economy evolves over time and reacts to policies and are typically indexed by "parameters" that describe the behavior of consumers, firms, technologies, markets and policies, among other things. However, in the real world, institutions and economic environments are constantly changing. As a result, it is not surprising that models' parameters are also often found to be changing over time. This leads to a first issue studied by the PIs, namely instability. The PIs developed a methodology that sheds light on the source of the reduction in the uncertainty in the economy (also known as the great moderation), an important economic phenomenon in mid-1980s. Based on the new method developed by the PIs, the decline in the volatility was a combination of changes in the behavior of consumers, firms and the Central Bank, as well as a decline in shocks to the economy (such as unexpected oil price changes). The PIs also developed methods for improving forecasts in such uncertain environments. The method successfully improved forecasts of real output growth, whose model indeed suffers from instabilities. (ii) Model misspecification is also an important problem that economists confront. Macroeconomic models are often simplified versions of the complex real world. While this practice makes it possible to solve and estimate models as well as study and interpret their implications on the economy, they may be too simplified to adequately describe the real economy. When this is the case, the model is deemed misspecified. A very important form of misspecification of macroeconomic models is the imposition of a representative agent in the model. In a representative agent macroeconomic model, the population is assumed to be homogeneous and everyone is assumed to be affected by policies in exactly the same way. In the real world, however, the population is quite heterogeneous, for example in terms of education and income. How are poor people affected by a tax increase or by an increase in the government expenditure? Are they affected differently from wealthier individuals? The PIs find that individuals are indeed affected in different ways depending on their wealth, for example. The PIs also developed a method for identifying potential misspecification in a macroeconomic model. A macroeconomic model is composed of several "building blocks"; for example, a first block may describe the behavior of consumers; a second block may describe the behavior of policy institutions such as Central Banks. The PIs developed a method to identify which of the "building blocks" is a good description of the real world and which ones are not. Thus, the technique can be used to help researchers improve their models. The results suggest that the labor and financial markets of the state-of-art macroeconomic models may not be an adequate description of the real world. However, it is not always easy to modify the models. To help researchers in that situation, the PIs also developed methodologies to improve the estimation of potentially misspecified economic models. (iii) A third, important issue faced by researchers in practice is lack of parameter identification. Identification is a crucial assumption: when it fails, the estimated parameters may not correctly capture the features of the data that the researcher aimed to learn about. The PIs developed techniques to test whether the models' parameters are identified. The advantage of the methodology proposed by the PIs is that it can detect whenever the strength of identification is not strong which is not possible with most of the existing methods. When applied to the data, the techniques find that identification of the parameters is indeed an issue in a widely-used macroeconomic model.
