Lagging Independent Variables in Probit.

[nickij]

Hello,

I am doing a paper on term-spreads (difference in yields of 10 year and 3 month bonds) predictive power on the real-economic activity. I am using a Probit model to assess the probability of a recession.

What I want to check is to see the lagged term-spreads predictive power using various lags - the data is sampled quarterly.

The general model is specified as: F(x=1|Spreadt-k) or a + b*SPREADt-k.

I am confused regarding the method of lagging.

If I wish to lag the term-spread by 4 quarters (i.e. I wish to compare the state of the term-spread 4 quarters ago with todays general state of the economy). Do I specify the model as:

1) y = c + c(1)*SPREAD(-4)
or
2) y = c + c(1)*SPREAD(-1) + c(2)*SPREAD(-2) + c(3)*SPREAD(-3) + c(4)*SPREAD(-4)

Yet another confusing factor if I use the latter is the output eviews provides. Since I am specifying 4 variables (5 with the constant term) which one of them should I look at when I determine the 4-quarter lagged term-spread? Do I look at c(1), c(2), c(3) or c(4)?

The output of eviews is similar to this:
c - XXX
c(1) - XXX
c(2) - XXX
c(3) - XXX
c(4) - XXX

-Thanks

[EViews Gareth]

1)

[nickij]

1)

Thanks for the quick reply Gareth.

Do you have an intuitive explanation as to why 1) is to be chosen in favor 2)? I have tried searching the net, in addition to looking through relevant econometric litterature without any luck.

-Thanks again

[EViews Gareth]

I gave no judgement. Just answered your own question:

If I wish to lag the term-spread by 4 quarters (i.e. I wish to compare the state of the term-spread 4 quarters ago with todays general state of the economy). Do I specify the model as:

If you wish to view X against X four quarters ago, then in EViews you use X(-4).


Whether you want to view X against X(-4) AND X(-3) AND X(-2) AND X(-1) is another question.

[nickij]

Thanks again for your reply Gareth.

Yes, I want a static view of the term-spread (for example 4 quarters ago) and compare it to the current state of the real-economic activity.

I just read in my econometrics book (Gujarati & Porter) that if I wanted to lag variables I was supposed to do the (-1) + (-2) + (-3).... but it made no sense to me as to why I would have to do this if I just wanted a "snap-shot" of the prevailent state of the slope of the spread x-quarters ago, hence my question.

On a more technical note, if I specify the model as c(1)*SPREAD(-4) how does that relate to the sampling of the data? I read in the literature that if I have a series in a coloumn and lag it by 1, i.e. (-1), then the first number moves down to the second "spot":

Initial:
1 - A
2 - B
3 - C
4 - D
5 - E

After Lagging:
1
2 - A
3 - B
4 - C
5 - D

So two questions springs to mind here; 1) what happens to E'th value in this operation? 2) if I lag the variable as described above (-4), then does that mean that the number 1 (representing GDP) is "measured" against the fourth variable - D (representing the spread 4 quarters ago for example)?

Again, I really appriciate your help.
-Thanks

[EViews Gareth]

Just type:

Code: Select all

show X X(-4)

Where X is a variable, and see what it looks like.