Page 1 of 1

Interpreting the output for VAR with a dummy

Posted: Fri Nov 13, 2015 8:40 am
by yelena
Dear Tom,
I have run a VAR model with one dummy variable that stands for sanctions. In years >= 2014:2, the dummy variable called "sanctions" = 1, otherwise it is equal to 0. A part of the code is attached. The question I have is the following, "If I need to forecast macroeconomic variables, say, real GDP, I usually run the forecast procedure. But with a dummy variable included, do we need to report their forecasted values? What do they mean?

dummy(from=2014:2) sanc

@VARLagSelect(lags=12, cric = aic)
# DREALBRT_R DRGDPSA_2010 DRCPSA DRCGSA DRISA DRXSA DRMSA INFL DREER sanc

system 1 to 10
variables DREALBRT_R DRGDPSA_2010 DRCPSA DRCGSA DRISA DRXSA DRMSA INFL DREER sanc
lags 1
det constant
end(system)
estimate(print, outsigma=V) 1999:2 2014:4 1

forecast(print) 10 10 2015:1
#1 f_DREALBRT_R
#2 f_DRGDPSA_2010
#3 f_DRCPSA
#4 f_DRCGSA
#5 f_DRISA
#6 f_DRXSA
#7 f_DRMSA
#8 f_INFL
#9 f_DREER
#10 f_sanc

Output:

Re: Interpreting the output for VAR with a dummy

Posted: Fri Nov 13, 2015 9:49 am
by TomDoan
Yes. You need the dummy (or any exogenous variable) to extend to the end of the forecast period. What it "means" is up to you? You might need to do one set of forecasts with the dummy off through the forecast period (sanctions end) and one with it on (sanctions continue),

Re: Interpreting the output for VAR with a dummy

Posted: Fri Nov 13, 2015 11:27 am
by yelena
Dear Tom,

How should I extend the dummy variable? Say, I want to have a forecast for 9 consecutive quarters starting from 2014:3. What should I write in the program? Also, is it okay if the forecasted value for the dummy variable has values different than 0 and 1, like 1.5 or 1.4?

Thank you very much.

Yelena

Re: Interpreting the output for VAR with a dummy

Posted: Fri Nov 13, 2015 11:29 am
by yelena
Should I say,

dummy (from = 2014:2, to = 2017:2) sanc

Re: Interpreting the output for VAR with a dummy

Posted: Fri Nov 13, 2015 11:56 am
by TomDoan
The range over which the dummy is defined is set using the start and end options on the DUMMY instruction:

https://estima.com/ratshelp/index.html? ... ction.html

so

dummy (from=2014:2) sanc * 2017:2

However, isn't SANC in your data set? The instruction above would zero out the values of SANC through 2014:1. It sounds like you really want to do

set sanc 2014:3 2017:2 = 1 (or 0 depending upon whether you want it on or off)

to extend SANC out into the forecast range.

Re: Interpreting the output for VAR with a dummy

Posted: Fri Nov 13, 2015 12:54 pm
by yelena
Oh, fantastic! Thank you very much, Tom! It helps me sooo much!

Yelena