Jonathan

My comments#

In my view, the proposal is a pass. The introduction features a clear motivation of the topic and the present literature on QE, and identifies the gap which you want to fill.

You might want to also refer to the (possible) impact of QE on inequality to illustrate more societal relevance. I am not sure the graph says what you want it to say (divergence between financial markets and ec. growth), maybe if you cumulate both series, it will do a better job.

  • Figure 1: I know I encouraged you to put it in, but what do you want to show with the graph?

The literature has a clear structure and features all the elements you need to develop your hypotheses, although there is an edge towards empirical literature rather than theoretical papers attempting to outline the mechanisms you will likely observe the effects of. Also, the purpose of the section detailing the effect of recessions on the stock market is not clear to me yet.

  • Literature starts with empirical studies. It is not more logical to start with theoretical views and mechanisms?

The hypothesis development probably deserves a separate section.

The methodology section contains an outline of the approach that satisfies the requirements, but there are nevertheless some unclarities: for example, your p subscript corresponds to the lag, not to the index of the variable in the system. The non-normality test relates to normality in the residuals, not the residuals in your data, and similarly for the autocorrelation test. The table and regressions on pp. 21-22 can probably be made more concise.

Finally, the methodology section might be appended with a passage about the influence of multicollinearity on the VAR-estimator.

Issues#

I have made quite some progress on the empirical analysis and have been conducting my interpretations

  • Multicollinearity issues in VAR estimation

You write that “Given that the nature of the variables being tested is collinear we will jointly evaluate the coefficients that are obtained in order to make conclusive interpretations.”

But is there a consequence of multicollinearity for the (standard errors of) the estimates, like in the OLS case? Talk about this in the methods section.

  • Also talk about Eigenvalue stability results in the methods section! This relates to joint stationarity of the time series.

  • The results are reported too extensively (I think).

  • Structure: table $\rightarrow$ interpretation

  • “Using the Lagrange-multiplier test we find no evidence of autocorrelation in either of the two lags.”

    • Autocorrelation in the residuals when estimating a model with two lags

  • Very important: how do you pick the lags?

    • P-value, BIC or AIC elimination is the standard
    • Diagnostic tests could be integrated
    • Hausman tests