marginal effect of binary x-variable in probit model

[archibaldthekoala]

Hi there,

I'm trying to get the marginal effects of dummy (i.e. zero-one) x-variables in my probit model. I.e. if the x-variable I'm interested in is x1, I need to find

G(B0 + x1 + B1x2 +...Bk.xk) - G(B0 + B1x2 +...Bk.xk).

I know EViews lets you get marginal effects of roughly continuous x-variables - is there any way I can estimate the above? (I would do it by hand, but the probit CDF is so complicated!)

I am using EViews 4, student version (at home) and EViews 6 (at university).

Thanks!

[EViews Glenn]

If you want to compute it for every observation in the sample, the easiest way is to first make a backup of your original dummy variable.

Then set it equal to 0. Click on the forecast button and compute the fitted probability.
Then set it equal to 1. Click on the forecast button and compute the fitted probability, saving the results to a different series.
Restore your original dummy variable
Compute the difference.

[archibaldthekoala]

Thanks for the reply.

Just to check - as I understand it, this is what I should do:

1) 'Set it equal to 0', i.e. change the dummy variable (or the backup of the dummy variable) so that every observation is 0 (i.e. x1=0 for each person in my sample)
2) Create another version of the same dummy in which each observation is set to 1
3) Re-estimate the regression when x1 = 1, and compute the fitted probability
4) Re-estimate the regression when x1 = 0, and compute the fitted probability
5) Get the difference of those fitted probabilities?

Is that right?

Also, is there any way to do this for lots of x-variables at once? Or do I just go through the same process with each dummy variable for which I want to get the marginal effect?

Thanks again!

[startz]

Not quite. You don't want to re-estimate. More like
series backup =0
x1=0
forecast
x1=1
forecast
x1=backup

[archibaldthekoala]

Sorry, could you elaborate a bit more on that?

[startz]

series backup =x1 ' first make a backup of your original dummy variable (my post had a typo here)
x1=0 'set it equal to 0
forecast 'Click on the forecast button and compute the fitted probability
x1=1 ' set it equal to 1
forecast 'Click on the forecast button and compute the fitted probability, saving the results to a different series
x1=backup 'Restore your original dummy variable
'Compute the difference