The reviewer wanted more detail about what you did and why you did it. Tell him/her. David Booth

Ibrahim Niftiyev , probably the reviewer wants to see not only whether a variable affects or not the other (i.e. the results of the Granger causality tests), but also to which extent (the magnitude and temporality of the dynamic relationship, something you can obtain from the IRFs of a VAR model). If you want to apply a VAR but you have a small sample size/data limitations, you may want to consider a Bayesian VAR. Bayesian VARs are very popular and Bayesian methods are valid in small samples.

Dear Ibrahim Niftiyev ,

Most time-series macroeconomic variables are potentially endogenous rather than exogenous and implicitly exhibit joint dynamic behaviour (Sims, 1980), and models that explicitly demonstrate causality have great potential of being mis-specified (Ganyaupfu, 2019). As an alternative, a VAR model is commonly used in consideration of the rationale that VAR allows variables to interact without imposing theoretical structures on estimates. Furthermore, a VAR model treats variables as endogenous, optimally describes dynamic behaviour of variables, and efficiently captures both short-run and long-run interrelations among the respective variables.

Now, given that a Granger-causality test is one of the major three types of structural analysis, where the other two major ones are the impulse response functions (IRFs) and forecast error variance decompositions; then your response to the Reviewer regarding performing VAR should strongly be guided by how you specified the main objective of your empirical study. If your study’s primary aim was NOT to assess the endogenous behaviour and relationships among multiple variables as they changed over time and forecast thereafter, then there is no rationale to apply a VAR model (special type of stochastic process model) in your analysis. Even if your empirical results show evidence of Granger causality between your variables, but if forecasting was not part of your study's primary objective (incl the reason of a small sample size you mentioned) then there is no justifiable reason to estimate a VAR model even if results indicate forecasting ability (Granger causality does not imply true causality, but only implies forecasting ability).

Dear Elvis Munyaradzi Ganyaupfu

Thank you for your detailed answer. I appreciate your help.

Best

Ibrahim

Not knowing how you specified your model, it's difficult to know what the reviewer meant. Usually, you need to check for the robustness of the specification, check for non-cointegration, and very frequently, perform impulse response analysis under several different ordering of the variables.

Thank you, dear Francis W Ahking

If you must apply VAR as suggested by your reviewer, certain conditions must be met, eg I(1) order of integration from the unit root test, non existence of cointegration.

if the conditions are ok, VAR will certainly give you more detailed results including the IRF, Variance decomposition, Impulse response etc

It is difficult to answer this question without knowing exactly what you are trying to model

I presume that theory implies and testing confirms that all your series are stationary. If this is the case

I would suspect that analyzing your VAR to provide answers to questions such as these might help in getting your paper accepted. These questions are also relevant if you are estimating impulse response functions or variance decompositions,

Thank you John C Frain