< < <
Date Index
> > >
Re: Current position in the K-wave (corrected)
by SOncu
07 September 2001 22:40 UTC
< < <
Thread Index
> > >
In a message dated 01-09-06 19:05:21 EDT, malexan@net-link.net writes:

> Seasons can be visualized in GDP by dividing the GDP by its long term trend 
> value.  I obtain this trend by doing a running regression of ln(GDP) versus 
> year over a 100 year period, centered on the year of interest.  Thus, the 
> trend GDP for 1850 is obtained from the regression equation for the period 
> 1800-1900.  After 1950 a fixed regression equation obtained from the 1900-
> 2000 data is used, and before 1839 a fixed regression equation obtained 
from 
> the 1789-1889 data is used.  Here is what a plot of GDP/GDP trend looks 
like:
>  

Dear Mike,

I have not read your book so I am not very clear about the details of your 
model.  However, the above paragraph tells me that to characterize the 
present, say 1950 is the present, you are using future information spanning 
1951-2000. In 1950, this information was unavailable. If your objective is to 
characterize the past up until 1950, which is also the past of 2001, in some 
manner, this is fine. But if your objective is to make projections into the 
future of 1950, although I am not sure whether this is the case, the above 
would be a problem. 

I plan to read your book and hopefully carry more informed discussions with 
you after that in private but before that let me ask you whether you have 
done any predictive validity tests of your model. From the above and the rest 
of your mail I am under the impression that you are working with yearly data. 
For example, have you ever run yearly(or longer period)-rolling forecasts of 
some kind after calibrating your model using information only from the past 
of your forecast date and see whether your hit rate is statistically 
significant or not?  

Here is one more question pertaining what is below:

>The Depression and World War II really threw a wrench into the cycles.
>Government economic management that was introduced in response to 
>these events has produced the twentieth century price revolution and 
>practically eliminated the cycle in GDP.  The cycle in stocks is stronger 
>than ever though.  GDP is pretty useless for detecting the cycle today.  

The above is a description of a structural change in the system. Most likely, 
other structural changes had also occurred in the past. Indeed, there even 
were a few systemic changes, depending on how far back you go in the past. 
Have you ever looked at how such structural/systemic changes affect your 
model? 

These are a few questions that came to my mind before reading the book. 
Hopefully, I will be able to ask more intelligent questions after reading it. 

All the best,
Sabri Oncu 


< < <
Date Index
> > >
World Systems Network List Archives
at CSF
Subscribe to World Systems Network < < <
Thread Index
> > >