Skip to main content

New Approaches to the Identification of Macroeconomic Models

Periodic Reporting for period 3 - Macro Identification (New Approaches to the Identification of Macroeconomic Models)

Reporting period: 2018-09-01 to 2020-02-29

Macroeconomic data are largely non-experimental. Thus, causal inference in macroeconomics is largely based on assumptions about what aspects of the variation in the data are exogenous. This presents two major challenges, which this research addresses directly. First, few such assumptions are generally accepted. Second, conditional on any set of assumptions, identification of causal effects is often weak because there is little relevant variation in the data. To tackle these challenges, I propose two lines of enquiry to explore new sources of identification and develop the requisite econometric methods.

The first line aims at developing new methods for estimation and identification of the causal effects of macroeconomic policy when the policy instrument is subject to occasionally binding constraints. The leading application is on monetary policy where the policy instrument, the nominal interest rate set by a monetary authority, is constrained by a lower bound. This is known as “zero lower bound” problem, or ZLB for short. I will refer to this from now on as “the ZLB project” even though its scope is wider. This project will help us measure the impact of macroeconomic policy on the economy, and assess the efficacy of different policy in achieving their stated objectives. It is therefore of major importance to society at large. The overall objectives of the ZLB project are twofold: (i) to develop methods for analysing dynamic economic models for data that are subject to occasionally binding constraints; and (ii) to apply those methods to the ZLB problem in order to assess the efficacy of conventional and unconventional monetary policy.

The second line of inquiry contributes to the ongoing research on developing methods of inference that are robust to the problem of weak instruments. Weak instruments are pervasive in economics and threaten the validity of inference under any identification scheme. A burgeoning literature has developed in the last two decades whose objective is to develop methods of inference that are reliable and efficient even when there is little information in the data. It is therefore of central importance for causal inference using non-experimental data in economics and other fields. This line of inquiry has focused on the problem of inference on individual coefficients or subsets/subvectors of coefficients in models that contain several variables. The project will extend the state of the art by providing more efficient methods for subvector inference that are robust to weak instruments.
Three research papers have been completed so far since the beginning of the project, and ongoing collaborations have been established with researchers from several universities and policy institutions. Two of these papers are under review in leading international peer-reviewed journals. The results have been disseminated through presentations to several international conferences, universities and policy institutions.

The main results across the projects so far are the following. The ZLB can be used constructively to identify the causal effect of monetary policy on the economy. The requisite econometric methodology to estimate macroeconomic models with variables subject to occasionally binding constraints has been developed. As a bi-product, this methodology can also be used to test the efficacy of unconventional policies. Preliminary results from application of those methods to US and Japanese data suggest that the ZLB is a powerful tool for identification of the effects of monetary policy, and unconventional policies have not been fully effective in mitigating the constraints induced by the lower bound on interest rates.

A new test of hypotheses on subvectors of coefficients in the homoskedastic linear instrumental variables regression has been developed and is a significant improvement over the state of the art. The new test is easy to use and has a certain appealing optimality property: it is more powerful than all other existing methods in the case of just-identified models. This work has also established a new methodological framework for studying subvector inference that covers some forms of heteroskedasticity, thus greatly extending the applicability of these methods in practice.

Finally, a new method of inference has been developed for structural vector autoregressions that are identified by placing restrictions on the long-run effect of economic shocks. The method is robust to weak identification that arises when the data is very persistent, and sheds light to a long-standing debate on the effects of productivity shocks on employment.
The work on the ZLB has extended the state of the art by uncovering a heretofore unknown source of identification in macroeconomics, as well as by developing new methods for analysing macroeconomic models of data that are subject to occasionally binding constraints. Further work in this area will extend these methods in order to cover the most relevant macroeconomic applications (e.g. allowing for time-varying volatility of shocks, imposing additional identifying restrictions, performing Bayesian inference), and apply the methods to obtain empirical results on the efficacy of conventional and unconventional monetary policy across different countries.

The work on subvector inference has extended the state of the art by showing that existing methods of inference are inefficient and providing a new more powerful test. Further work in this area will extend the scope of the analysis to cover heteroskedasticity. Robustness to some form of heteroskedasticy is relatively straightforward, and I anticipate that this extension will prove useful in a variety of applications. This is largely an empirical question. Some progress has been made towards extending the results to a general form of heteroskedasticity, but this problem is considerably more difficult.