In their article, Tenreyro and Thwaites (2013) conduct research to explore the impact of monetary policy on real and nominal variables at various business cycle stages. They aim to determine whether the effects of monetary policy are symmetrical or asymmetrical throughout the business cycle and identify the sources of any asymmetry observed. In my view, considering these objectives, using the impulse response functions (IRFs) in the study is quite persuasive since it is combined with other methods to complement its limitation in answering the study’s objectives.
The authors employ IRFs to analyze the dynamic response of a variable system to a specific shock. These functions capture the average effect of the shock on the system’s variables based on the state of the economy when the shock occurs, encompassing its impact on future changes. By utilizing IRFs, the authors estimate how real and nominal variables respond to monetary policy shocks, facilitating the examination of response magnitude and timing. Additionally, IRFs allow them to study variations in these responses between economic expansions and recessions.
Despite their usefulness, IRFs have some limitations. First is their assumption of linearity and exogenous shocks, which may not always hold in real-world economic systems. Their effectiveness depends on the underlying model used to generate them, and different models can yield varying IRFs. Non-linearities or structural breaks, which are significant characteristics of variable relationships, are not accounted for in IRFs, lacking a comprehensive depiction of variables’ interconnectedness. Additionally, in cases where the data-generating process cannot be adequately approximated by a vector autoregression (VAR) process, IRFs derived from the model may lead to biased and misleading results. Therefore, it is essential to consider these limitations when interpreting the implications of IRFs in economic analysis.
While IRFs are valuable for studying the dynamic impact of shocks on variable systems, they may not be the sole or optimal approach for answering the questions of the study. Hence, it is essential to complement them with other analytical tools. In this case, it is my view that Tenreyro and Thwaites have effectively utilized a combination of methods to extend the IRFs’ analysis. To complement IRFs, the authors incorporate the local projection method (Jordà, 2005), and combine it with the smooth transition regression method (Granger & Terasvirta, 1994). This adaptation allows IRFs to be influenced by the state of the business cycle, offering a more comprehensive perspective on the effects of shocks. Moreover, this combined methodology improves the estimation of shocks and reduces susceptibility to measurement errors.
The article clearly explains that the local projection method offers several advantages for studying the impact variation of shocks over time. One key advantage is their focus on the state of the economy at the time of the shock’s occurrence, lending flexibility to accommodate a panel structure and reducing sensitivity to misspecification. Furthermore, the combination method of smooth transition regression local projection model (STLPM) effectively handles non-linearities, which are weaknesses of IRFs, and estimates the impulse response of real and nominal variables to monetary policy shocks during different stages of the business cycle. In contrast, IRFs assume a constant state of the economy when the shock hits, which becomes problematic when dealing with shocks that lead to significant real effects. Estimating the transition between different economic regimes caused by the policy shock in a regime-switching VAR model involves numerous modeling choices that can be prone to errors and controversies. By employing a regime-switching local projection model, researchers avoid the need to make assumptions about how the economy transitions between regimes, which is particularly beneficial when studying the effects of fiscal consolidations or other impactful policy shocks.
Nevertheless, it should be noted that employing the local projection IRFs method tends to exhibit increased bias and variance. As a result, the confidence intervals for impulse responses are generally less accurate and wider on average compared to appropriately designed intervals based on VAR models (Kilian & Kim, 2009). Hence, in the case of a finite sample in the subsamples data in the analysis, it could hypothetically result in a wider confidence interval of IRFs. It means that using confidence intervals for IRFs in the study could shed more information about IRFs’ estimation accuracy.
Moreover, it is imperative to consider that the article’s analysis focused on the United States as an industrialized and developed country. As a result, the applicability and generalizability of the findings to emerging economies may raise some questions. Thus, future research could explore these aspects in more detail to understand how the effects of monetary policy may vary in different economic contexts.
References
Granger, C., & Terasvirta, T. (1994). Modelling nonlinear economic relationships. International Journal of Forecasting, 10(1):169–171.
Jordà, Ò. (2005). Estimation and Inference of Impulse Responses by Local Projections. American Economic Review, 95 (1): 161-182.
Kilan, L., & Kim, Y. J. (2009). Do Local Projections Solve the Bias Problem in Impulse Response Inference?. CEPR Discussion Paper Series 7266. Tenreyro, S., & Thwaites, G. (2013). Pushing on a string: US monetary policy is less powerful in recessions. CEP Discussion Paper 1218.
hope. 