Suggestion: Temporal Disagregation (Litterman/Chow-Lin etc)

[fmramos]

Hi all!

On suggestion for next versions is a useful tool for economists: temporal disaggregation methods. In fact, there are some simple tools already in EV (conversion methods). But I'm talking about methods like Chow-Lin, Fernandez or Litterman.

Nowadays, when I need to implement them, I use Matlab/Eurostat/Nbb routines. I use them combined to EViews to seasonal adjust series.

This is useful to cope with variables as GDP and components in a quarterly or even monthly basis.

Tks.

Fabio

[QSnakecharmer]

I agree with Fabio, Chow-Lin and other interpolation methods would be an interesting option to add in future versions. I also use the frequency conversion tools of Eviews a lot.

[EViews Gareth]

Do you have any references you could cite on these methods?

[fmramos]

Tks for the answer. Sure I have.

I think the best way to understand how the programming works is to get the text "Software to perform temporal disaggregation of economic time series", by European Commision/Eurostat (2005). If u prefer, I can send u by email.

These methods (Chow-Lin, Litterman, Fernandez etc) comprehends stochastic ways to convert the frequency of time-series, avoiding problems such as breaks, but keeping sums or averages. Some methods are univariate (eg. Denton) and, others, multivariate. Some allow to disaggregate with constraints (eg. quarterly GDP by monthly industrial production but constraint by quarterly personal consumption, apparent consumption of machinery and external sector - each of them estimated by monthly indicators).

OECD has some material at http://www.oecd.org/document/47/0,2340, ... _1,00.html.

For economists, is really useful. Cause the lack of software to do it, its common to use simple tools as linear interpolations.


Tks for the interest.

Fabio

[fmgoto]

Anyone has these codes imlpemented in Eviews?

Gareth, can I solve non-linear equations with Eviews? Suppose I need to solve the following equation for "a":

q = (a^5 + 2*a^4 + ...)/(3 + 2*a^2+...)

Can I solve it within Eviews? How?

(Source: Chow and Lin, 1971, Review of Economics and Statistics, eq. 23)

Thank You, Fabio

[EViews Gareth]

You can estimate an implicit equation where you write an expression that is assumed equal to the residual. Thus something like:

ls ( q- (c(1)^5 + 2*c(1)^4...)/(3+2*c(1)^2...) )