conditional frcts using both anticipated and unanticipated shocks

[Andrea Gerali]

Hi,

I want to obtain conditional frcts assuming that the conditioning path of one variable (say, short-term interest rate) is known to the agents in the model (i.e. using anticipated shocks) but, at the same time, the conditioning path of another variable (say, oil price) is NOT known to the agents in the model (i.e. there I want to use unanticipated shocks). 

I looked into the IRIS discussion forum, read the tutorials and the manual but I did not found any relevant examples for this. I know IRIS in principle can mix anticip. and unanticip. shocks, but I just could not figure out how to implement my "experiment".

Is it because this thing is not possible in IRIS?

Thanks,

Andrea

[Jaromir Benes]

Hi Andrea

Are you talking about conditioning or exogenizing? The two cases differ a little bit, so to keep the discussion more focused, I just want to make sure I know exactly what you're asking about.

Mirek

--
You received this message because you are subscribed to the Google Groups "iris-toolbox-discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to iris-toolbox...@googlegroups.com.
To post to this group, send email to iris-t...@googlegroups.com.
Visit this group at http://groups.google.com/group/iris-toolbox.
For more options, visit https://groups.google.com/groups/opt_out.

[Andrea Gerali]

I meant conditioning ...

[Andrea Gerali]

Hi Benes and progress on the issue I raised ?
Andrea

[Michael Johnston]

Hi Andrea, 

Jaromir is pretty busy at the moment so I will attempt to answer. There are two ways to potentially do this, as Jaromir mentioned earlier: conditioning and e/e swaps. To my knowledge it is not currently possible to use anticipated and unanticipated conditioning. 

With e/e swaps you can easily combine anticipated and unanticipated shock tunes using the fourth input argument to endogenise()/autoexogenise(). First you need to keep in mind that there is always a default anticipation method when calling jforecast()/simulate(): by default 'anticipate=',true, but you can obviously change this. When calling autoexogenise()/endogenise() use a complex number as the fourth input argument to indicate whether the swap uses the default method (real) or not (imaginary) or some combination of the two. E.g., the following are equivalent: 

p = plan(model,dates) ;

p = autoexogenise(p,'oil',edates,1i) ;

p = autoexogenise(p,'r',edates,1) ;

u = jforecast(model,database,range,'anticipate=',true,'plan=',p) ;

OR

p = plan(model,dates) ;

p = autoexogenise(p,'oil',edates,1) ;

p = autoexogenise(p,'r',edates,1i) ;

u = jforecast(model,database,range,'anticipate=',false,'plan=',p) ;'

Hope this helps. 

Michael

On Thu, Jan 30, 2014 at 10:51 PM, Andrea Gerali <andrea...@gmail.com> wrote:

Hi Benes and progress on the issue I raised ?
Andrea

[Andrea Gerali]

Thanks Michael,

now I know that "... it is not currently possible to use anticipated and unanticipated conditioning." 

Then, I have a related question (also posed to Mirek, but see if you can help me while he is busy):

  

when setting up a condition plan to be used in a "conditioning" exercise (not a e/e swap), is there a way to partition the set of N structural shocks into M (M< N) disjoint subsets to be used by my M conditioning variables? E.g.: the short-term rate conditioning path should only use:  E_R (innovation in the taylor rule) and E_PI_TARGET, innovation in the long-term inflation target shock. The oil price conditioning path can only use "oil-related" shocks (a couple in my model). And so on ... 

The <auto-exogenize> command does not do it because it only pairs one variable with one shock. I want a "correspondance"  between one variable and a "subset" of shocks. 

Thanks,

Andrea

2014-01-31 Michael Johnston <mkjo...@gmail.com>

[Jaromir Benes]

Hi Andrea

I finally have some spare time, and will try to answer the questions.

It is actually possible to combine anticipated and unanticipated conditioning (and exogenising as well, for that matter -- everything can be actually combined together in one run). Let me quickly code up a very basic example how to do this, and I'll email that to you in say 45 minutes. The basic idea is that you can assign std deviations of model shocks complex numbers, effectively thus introducing two shocks in the stead of each one -- an anticipated and an unanticipated.

Now, the "anticipation" and "unanticipation" describes the nature of the underlying shocks, and hence not necessarily the anticipation or unanticipation of the imposed endogenous variables. This closely relates to the other question about the partitioning of the structural shocks into disjoint subsets. I have actually given a lots of thoughts to exactly this question in the past, but haven't been able to come up with a satisfactory solution/design. I simply don't know if it's possible to do that in a simultaneous system... This question is, of course, independent of the anticipate/unanticipate issue, so let's say we have a static system with 2 variables (x,y) being determined by 4 shocks (a,b,c,d) -- think of it as a reduced form solution of a structural model

x = a + b + c + d;

y = 2*a - 3*b + 13*c - 5*d;

(the coefficients are totally irrelevant, of course:))

Whenever we use shocks a,b to condition x upon a particular value, this will automatically also change the conditional distribution for y, and hence influence the conditional distribution of c,d in case we use them to further condition y upon another particular value in the system. If we, on the other hand, do not condition x but we do y, then the conditional characteristics of the system with only y conditioned upon the same value as before will be different. In this sense, I don't see any possibility to construct disjoint sets...

Maybe there's a clever/simply way out of the simultaneity problem but I haven't arrived at it (yet?) :)).

Your thoughts are more than welcome.

M.

[Jaromir Benes]

Andrea

Attached is an example of a conditional forecast combining anticipated and unanticipated shocks. Let me know if it needs to be commented on more heavily for everything to be clear. And I'd really like to discuss further the issue of "disjoint sets". This is indeed something many other people ask about, and would be great to come up with a neat solution...

M.