Calculating Dynamic Multipliers in a Combined ARDL and VECM Model with EViews 13

[Bob]

Hello everyone,

I am currently engaged in an analysis that combines two modeling approaches: an ARDL model and a VECM model, integrated into a single framework within EViews 13. My goal is to calculate dynamic multipliers to assess the impact of a shock in one specific variable (ger_3m_yield, an exogenous variable in the ARDL model) on the output of another variable (housing_credit, output of the VECM model), over a 24-month period (such as those calculated for an ARDL with code of the type mortgage_rate_ardl.dynmult(dmevolution,noci) 24)

Below is a brief overview of the commands used to set up the models and their integration:

pageselect mens
smpl 2010 2023
' ARDL
equation mortgage_rate_ardl.ardl(deplags=12, reglags=12, ic=aic, coef=ardl) mortgage_rate ger_3m_yield log(qe) mp_surprises france_inflation
' VECM
var housing_market_var.ec(rank=2, 2) 1 6 log(housing_credit) log(housing_transactions) log(housing_prices) @exogdual mortgage_rate log(hh_financial_wealth)
' Combining ARDL and VECM models
model pm_housing_market
pm_housing_market.merge mortgage_rate_ardl
pm_housing_market.merge housing_market_var
' Dynamic multipliers
?

I am encountering difficulties with programming the calculation of dynamic multipliers for the combined model. Please, could anyone guide me on the best way to proceed with EViews 13? Specific suggestions or code examples would be greatly appreciated.

Thank you in advance for your help and advice.

Best regards,
Bob

[Bob]

Hello,

In fact, since my question was poorly posed, I probably won't get an answer. It's poorly posed because I don't know all the possibilities of EViews.

For those interested, if EViews doesn't have a procedure, one must resort to more hardcore programming!
Theoretically, the combined model (ARDL; VECMX) should be able to be written with a state-space representation of the type:
Zt+1=AZt+BXt+ut
Yt=CZt
Where the matrices A, B, C depend on the estimated coefficients, the structuring of lags, and the error correction terms.

First difficulty: analyzing the ARDL and VECMX to intellectually construct the matrices (It's still complicated).

Second difficulty: programming Eviews to fill in the matrices with, among other things, the estimated coefficients (I clearly do not know how to do it).

After these 2 difficulties, calculating the dynamic multipliers for the horizon h (Mh) becomes child's play:
Mh=sum of i=0 to h of C*A^i*B

But I had imagined that EViews could spare me all this ;)
Best regards,
Bob

[Bob]

Conclusion and Aftermath of My Afternoon!

Attempting the transformation into state space turned out to be exceedingly challenging (feasibility? errors...) when dealing with cointegration equations. Therefore, I took a closer look at what EViews could offer and, indeed, it's a remarkable piece of software (once you get to know it, haha!)

In solving my particular problem, I learned that employing "scenario forecasting with alternate assumptions" and conducting a "Stochastic Simulation" with appropriate shocks are effective strategies. This technique is detailed in Chapter 51 of the "User Guide II."

However, I'd like to make an observation: while everything runs smoothly in deterministic simulation, in stochastic simulation, processes like Proc, Make Graph, and % deviations do not display!

Best regards,
Bob