Can one set up a stock-flow consistent model a la Godley and simulate it in Gretl ? And how ? Von meinem/meiner Galaxy gesendet _______________________________________________ Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/

Am 28.01.2023 um 19:23 schrieb Federico Giri: I have never tried to simulate a stock flow consistent model using hansl. On Gennaro zezza's website you can find a lot of example of R script on how tò build a SFC model. With a Little bit of patience you can try tò translate It in gretl Hi, let me expand a little both on the question and on the answer. Federico, I wasn't able to locate any code on Zezza's homepage. Do you mean this page? : https://www.unicas.it/didattica/docenti/schedadocente.aspx?UID=fa11c412-9... About what gretl/hansl can do in relation to SFC: I haven't dealt with SFC models myself, I only know a bunch of colleagues who have. So take my following remarks with some caution. In terms of econometrics, what you can do in gretl is to set up a general multiple-equation system in a 'system' block. Some documentation of this is available in chapter 34 of the user guide ("Multivariate models"). As explained in that chapter, this also allows to specify identities, which in my understanding is important to impose the restriction of being stock-flow consistent. (BTW, mainstream DSGE models are also stock-flow consistent, so I always found the name SFC a bit misleading, but let's not digress.) Of course you can also have contemporaneous relationships between endogenous variables in such a system, with the classical econometric complication that then you need appropriate instruments and you need to fulfill conditions for identification. There is no point-and-click interface for specifying a general system, so you would need to do a minimal amount of scripting. (There is a window that comes up if you select Model -> Multiple equation system in the GUI, and you have a little bit of mouse-connected helpers there, but in the end it's a 'system' block that you're writing there.) After estimation, you can also impose general coefficient restrictions (also cross-equation), with a 'restrict' block. After you're done estimating the (possibly restricted) system, you can also forecast with that. Section 35.6 of the user guide has some documentation about it, and it's only half a page, so I definitely recommend reading it. So, are "simulations" possible? This depends what you mean by that. In most cases, a simulation is just a conditional forecast, where the term "conditional" means for given future scenario values of the exogenous variables. This is certainly supported by gretl - of course you would have to provide the relevant values for the future exo series. If you run into problems there, feel free to ask for help here on the list. If, on the other hand, by "simulation" you mean that you want to produce a path for the endogenous variables that matches or approaches a certain pre-specified target, this is not something that gretl currently offers. It is certainly doable to write a hansl program that would do it (using the excellent numerical optimizers that gretl has), but it wouldn't be trivial. Finally, a remark about GMM estimation: AFAIK, a system-wide GMM estimation is not something that gretl currently offers. The single-equation 'tsls' command does offer a --gmm option, but it's not clear to me if or how that would be directly usable in the context of a 'system' block. Note that I'm not saying you need to do GMM - the offered true system estimators 3SLS and FIML are perfectly fine and, some would say, perhaps even better in practice. OK, that was a rather long message, but I hope that clarified some things. cheers sven _______________________________________________ Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/

Am 31.01.2023 um 10:50 schrieb Federico Giri: > This is the website I was referring to > > https://gennaro.zezza.it/ [1] > > However, the code are available in Eviews and not in R OK, thanks. Looking at the "simplest" model (https://gennaro.zezza.it/software/eviews/glch03.php [3]), there is no econometrics involved, it is "just" about simulation. It would be interesting to see whether the dynamic accounting relationships could be specified in gretl's system block as (dynamic) identities. It's not clear to me whether that is supported by gretl. (Also noting that the doc in ch. 34 says that identities only matter for FIML estimation, so another question is about their role in forecasting.)

> let me also add the personal page of Alessandro Caiani. Under > _teaching_, you can find models written in R > https://sites.google.com/view/alessandro-caiani/home [2] Actually, while the first entry there is with R code, further down you have Python codes and a Java-based toolkit (and also Mathematica, but that is again closed source, of course). > My original point is that an SFC model at the end of the day is a > system of dynamic backward-looking equations (similar to a svar or a > reduced form DSGE model) . I think that Gretl in principle can > handle that as long as you have some numerical solver available. Of > course, you have to build your own program but it can be done. Right. The relatively recent addition of lpsolve support (linear programming) might also be relevant here. (See Help / gretl + lpsolve).

Am 31.01.2023 um 12:10 schrieb Sven Schreiber: > "just" about simulation. It would be interesting to see whether the > dynamic accounting relationships could be specified in gretl's system > block as (dynamic) identities. It's not clear to me whether that is > supported by gretl. (Also noting that the doc in ch. 34 says that > identities only matter for FIML estimation, so another question is > about their role in forecasting.) OK, here's a partial answer with respect to gretl's handling of dynamic identities in forecasting from a system: It seems to be honored alright. A trivial and purely artificial but concrete example: