Ngozi ADELEYE,

By "Recursive", which of the two equations you consider? Residuals and coefficients of both the equations could exhibit changes or shift over time, but in different points. While dealing recursive type regressions for single equations are obvious in behavioral display, endogeneity incorporated system would need further assumptions. You should have more metric literature on it.

Dear Ngozi,

to me no recursive properties are visible. You have a 2-equation system, with two stochastic elements. I should be easy to substitute the first equation into the second. Doing this you would need to make sure that either 1) the two stochastic factors are uncorrelated, or 2) account for their eventual correlation. With time series you should always check for homoscedasticity and/or trends. On particular point: as you use two sets of exogeneous control variables, keep in mind to ensure the absence of multicollinearity!

Thanks Balaji and Frank, all points noted. It's actually a 2-equation recursive model because causality runs in one direction from 'X through BC to Inc'. 

My argument is that X has an indirect relationship with Inc (through BC) and this is what I did, so please correct me if I'm wrong (I'll appreciate that!)...I really need to get this right:

1) I did the preliminary unit root test for all the variables in the model (not all are stationary so I had to 1st difference those); also, there is no multicollinearity among the explanatory variables (both the financial and control variables):

2) using the ordinary least squares (OLS) method, I regressed the 1st equation and obtained the predicted value of BC

3) the coefficient of X obtained gives the direct impact on BC

4) I substituted the value of the predicted BC for the original BC into the 2nd equation

5) the coefficient of the predicted BC now gives the 'indirect' effect of X on Inc

...are these in order?

Thanks.

I think this thread addresses your query on test for long-term stationary position within VAR...

https://www.researchgate.net/post/Is_it_necessary_to_ensure_stationarity_of_all_time_series_variables_when_you_run_a_Vector_Autoregressive_VAR_Model

...thank you Lyndon, I'll read thru the thread...

I have found some interesting interaction effect between income inequality and financial development or credit growth in SSA with significant effect on the economy. That interaction could be treacherous based on the level of development or credit. This could help in the explanations of some of the things you'll be finding in your research. Good luck!