The Effects of Quantitative Easing: A Meta-Analysis

• Název práce v češtině: Účinky kvantitativního uvolňování: Metaanalýza
• Název v anglickém jazyce: The Effects of Quantitative Easing: A Meta-Analysis
• Klíčová slova: Meta-analysis, unconventional monetary policies, GDP, inflation, publication bias, model averaging
• Klíčová slova anglicky: Meta-analysis, unconventional monetary policies, GDP, inflation, publication bias, model averaging
• Akademický rok vypsání: 2021/2022
• Typ práce: bakalářská práce
• Jazyk práce: angličtina
• Ústav: Institut ekonomických studií (23-IES)
• Vedoucí / školitel: prof. PhDr. Tomáš Havránek, Ph.D.
• Řešitel: skrytý - zadáno vedoucím/školitelem
• Datum přihlášení: 19.04.2023
• Datum zadání: 19.04.2023
• Datum a čas obhajoby: 10.09.2024 09:00
• Místo konání obhajoby: Opletalova, O105, místnost č. 105
• Datum odevzdání elektronické podoby: 31.07.2024
• Datum proběhlé obhajoby: 10.09.2024
• Oponenti: Jan Mošovský, M.Sc.

Zásady pro vypracování

Start with the following (which ignores publication bias): https://www.sciencedirect.com/science/article/pii/S0304393221000398

Seznam odborné literatury

Andrews, Isaiah & Maximilian Kasy (2019): „Identication of and correction for publication bias." American Economic Review 109(8): pp. 2766-2794.

Bom, Pedro R. D. & Heiko Rachinger (2019): „A kinked meta-regression model for publication bias correction." Research Synthesis Methods 10(4): pp. 497-514.

Christensen, Garret, and Edward Miguel. 2018. "Transparency, Reproducibility, and the Credibility of Economics Research." Journal of Economic Literature, 56 (3): 920-80.

Christopher Martin , Costas Milas, Quantitative easing: a sceptical survey, Oxford Review of Economic Policy, Volume 28, Issue 4, WINTER 2012, Pages 750–764

Engen, E., Laubach, T., Reifschneider, D., 2015. The macroeconomic effects of the Federal Reserve's unconventional monetary policies. Finance and Economics Discussion Series 2015, 1-54.

Fabo, B., M. Jancokova, E. Kempf, and L. Pastor. (2021): Fifty Shades of QE: Comparing Findings of Central Bankers and Academics. Journal of Monetary Economics 120: 1–20.

Furukawa, C. (2021): „Publication bias under aggregation frictions: from communication model to new correction method." Working paper, MIT.

Ioannidis, J. P., T. D. Stanley, & H. Doucouliagos (2017): „The power of bias in economics research." The Economic Journal 127(605): pp. F236-F265.

Papadamou, S., Kyriazis, N.A., Tzeremes, P.G., 2019. Unconventional monetary policy effects on output and inflation: a meta-analysis. Int. Rev. Financ. Anal. 61, 295–305.

Pesaran, M.H., Smith, R.P., 2016. Counterfactual analysis in macroeconometrics: An empirical investigation into the effects of quantitative easing. Research in Economics 70, 262-280.

van Aert, R.C., M. and van Assen. 2021. Correcting for publication bias in a meta-analysis with the p-uniform* method. Working paper, Tilburg University.

Předběžná náplň práce

Research question and motivation

The main research question I intend to study is the effect of non-conventional monetary policy, primarily the effects of quantitative easing on the main macroeconomic indicators, namely inflation and GDP, by conducting a meta-analysis. Differences in reporting among people associated with central banks and academics will be discussed as well.

Quantitative easing has been a hot topic since the 2008 financial crisis. Now, with the unprecedented scale of QE during the Covid-19 pandemic, it has been brought back into the spotlight and its effects are still questioned in both academic and policy circles. A lot of the research on QE is created in central banks (Christopher et al. 2012). However, this characteristic of self-assessment combined with the fact that in general, findings with certain characteristics, such as statistical significance or being in line with theory get published more easily, may lead to bias.

Public policymakers, foundations and citizens rely on published research for guidance, particularly on policy-related matters. But if we only see a selective subset of the research, how valuable is the information that we obtain from published studies? Meta-analysis is a great tool to address this issue.

A meta-analysis on effects of unconventional monetary policies was conducted by Papadamou et al. (2019). However, the meta-analysis included only 16 studies which is generally insufficient for proper analysis of between-study heterogeneity, and it omits publication bias.

My main source will be the paper by Fabo et al. (2021), which finds that studies written by central bankers report stronger effects of QE on both output and inflation. They constructed a dataset of 54 studies analyzing the effects of unconventional monetary policy on output or inflation in the US, the UK, and the euro area. A small meta-analysis was conducted as a by-product, but without the usage of newer and more convenient techniques, and it does not address all the important aspects of a meta-analysis.

Contribution

A significant amount of research has been conducted on the effects of unconventional monetary policies, including the use of meta-analysis. However, my contribution to this field lies in the utilization of different methodologies. In particular, I aim to explicitly address and investigate the presence of publication bias. I plan to extensively search for this bias using various statistical tests and provide an explanation of how it could potentially influence the results. Additionally, I intend to construct a partially new dataset, which will enable us to examine heterogeneity in the literature. The results of this study could have practical implications as they will provide concrete conclusions that question the validity of previous findings and highlight possible biases and overestimations in the reported results. Additionally, this thesis could serve as a robustness check for existing studies, further strengthening the overall body of research.

Methodology

I intend to thoroughly examine the dataset constructed by Fabo et al. (2021) from studies before July 2018 and include as many sources as they have as well as include my own estimates from studies that might have been overlooked or studies that were published after July 2018. To achieve this, I will design an additional search query in Google Scholar using the same 40 keywords as the base paper. Once the data is collected and cleaned, I will determine the most suitable variables to be included in the final set for the meta-analysis.

Regarding publication bias, I will use three different approaches. Firstly, for the linear technique, I will utilize the FAT-PET with ordinary least squares (OLS), along with a regression weighted by the inverse variance, and the fixed-effects model. The idea is that in the absence of publication bias there should be no systematic relation between estimates and their standard errors which can be assessed visually in a so-called funnel plot. Secondly, recognizing that the relationship between the estimate and its standard error may not be linear, I will explore various non-linear techniques, including the ones proposed by Ioannidis et al. (2017), Furukawa (2021), Bom & Rachinger (2019), and the selection model by Andrews & Kasy (2019). Lastly, relaxing the exogeneity assumption I will implement the p-uniform* method by van Aert and van Assen (2021). The method does not assume any specific relationship between estimates and standard errors but relies on the principle that the distribution of p-values is uniform at the underlying mean effect size.

Furthermore, I will investigate heterogeneity within the data and studies, using Bayesian model averaging (BMA) as the primary tool to account for model uncertaity. Frequentist Model Averaging (FMA) as well as different priors or weights for BMA will be used as a robustness check.

Outline

Abstract
Introduction
a. Overview of the existing knowledge
b. My contribution
Data assembly
a. Literature search
b. Data collection
c. Initial analysis
Publication bias
a. Linear tests for the presence of publication bias
b. Non-linear tests for publication bias
c. Easing the exogeneity assumption and testing for publication bias
Heterogeneity
a. Variables and their construction
b. Model averaging
- Bayesian model averaging
- Robustness checks
The best-practice estimate analysis
a. Modelling the best-practice
Conclusion
a. Summary of the results
b. Implications and further research

Předběžná náplň práce v anglickém jazyce

Research question and motivation

The main research question I intend to study is the effect of non-conventional monetary policy, primarily the effects of quantitative easing on the main macroeconomic indicators, namely inflation and GDP, by conducting a meta-analysis. Differences in reporting among people associated with central banks and academics will be discussed as well.

Quantitative easing has been a hot topic since the 2008 financial crisis. Now, with the unprecedented scale of QE during the Covid-19 pandemic, it has been brought back into the spotlight and its effects are still questioned in both academic and policy circles. A lot of the research on QE is created in central banks (Christopher et al. 2012). However, this characteristic of self-assessment combined with the fact that in general, findings with certain characteristics, such as statistical significance or being in line with theory get published more easily, may lead to bias.

Public policymakers, foundations and citizens rely on published research for guidance, particularly on policy-related matters. But if we only see a selective subset of the research, how valuable is the information that we obtain from published studies? Meta-analysis is a great tool to address this issue.

A meta-analysis on effects of unconventional monetary policies was conducted by Papadamou et al. (2019). However, the meta-analysis included only 16 studies which is generally insufficient for proper analysis of between-study heterogeneity, and it omits publication bias.

My main source will be the paper by Fabo et al. (2021), which finds that studies written by central bankers report stronger effects of QE on both output and inflation. They constructed a dataset of 54 studies analyzing the effects of unconventional monetary policy on output or inflation in the US, the UK, and the euro area. A small meta-analysis was conducted as a by-product, but without the usage of newer and more convenient techniques, and it does not address all the important aspects of a meta-analysis.

Contribution

A significant amount of research has been conducted on the effects of unconventional monetary policies, including the use of meta-analysis. However, my contribution to this field lies in the utilization of different methodologies. In particular, I aim to explicitly address and investigate the presence of publication bias. I plan to extensively search for this bias using various statistical tests and provide an explanation of how it could potentially influence the results. Additionally, I intend to construct a partially new dataset, which will enable us to examine heterogeneity in the literature. The results of this study could have practical implications as they will provide concrete conclusions that question the validity of previous findings and highlight possible biases and overestimations in the reported results. Additionally, this thesis could serve as a robustness check for existing studies, further strengthening the overall body of research.

Methodology

I intend to thoroughly examine the dataset constructed by Fabo et al. (2021) from studies before July 2018 and include as many sources as they have as well. I will determine the most suitable variables to be included in the final set for the meta-analysis.

Regarding publication bias, I will use three different approaches. Firstly, for the linear technique, I will utilize the FAT-PET with ordinary least squares (OLS), along with a regression weighted by the inverse variance, and the fixed-effects model. The idea is that in the absence of publication bias there should be no systematic relation between estimates and their standard errors which can be assessed visually in a so-called funnel plot. Secondly, recognizing that the relationship between the estimate and its standard error may not be linear, I will explore various non-linear techniques, including the ones proposed by Ioannidis et al. (2017), Furukawa (2021), Bom & Rachinger (2019), and the selection model by Andrews & Kasy (2019). Lastly, relaxing the exogeneity assumption I will implement the p-uniform* method by van Aert and van Assen (2021). The method does not assume any specific relationship between estimates and standard errors but relies on the principle that the distribution of p-values is uniform at the underlying mean effect size.

Furthermore, I will investigate heterogeneity within the data and studies, using Bayesian model averaging (BMA) as the primary tool to account for model uncertaity. Frequentist Model Averaging (FMA) as well as different priors or weights for BMA will be used as a robustness check.

Outline

Abstract
Introduction
a. Overview of the existing knowledge
b. My contribution
Data assembly
a. Literature search
b. Data collection
c. Initial analysis
Publication bias
a. Linear tests for the presence of publication bias
b. Non-linear tests for publication bias
c. Easing the exogeneity assumption and testing for publication bias
Heterogeneity
a. Variables and their construction
b. Model averaging
- Bayesian model averaging
- Robustness checks
The best-practice estimate analysis
a. Modelling the best-practice
Conclusion
a. Summary of the results
b. Implications and further research