Re: st: Logit Model- Controlling for Differences Across Groups (Countries)

[Michael Weinberg <michaelhazeweinberg@gmail.com>]

Dear Dr. Hoaglin,

Thank you so much for responding to my Statalist question. I thought
I'd contact you directly to clear up some confusion I still have-- of
course if you don't have time, I will certainly understand.

My model is meant to explain factors that influence  'Preference for
Self-Employment' (a binary variable) among the 2000 respondents in the
5 countries. In addition to the individual variables (such as age,
gender, wealth, risk willingness, etc), I wanted to include several
macro factors, such as GDP per capita, inflation variability, taxes on
business profits, property rights, and several more. I believe that
such factors are theoretically important in explaining my dependent
variable, so I am eager to include them in the model.

As I understood from your advice, I can include this information only
by removing some or all country dummies, and then I'm still limited to
only four macro variable. I am unfortunately not an expert with
binary-response regressions, but this seems like a major limitation to
these models. Do you recommend any other way (perhaps another model
entirely?) which I can employ to capture the explanatory power of
these macro variables without forsaking the country dummies (which
seem, from your comments, to be rather essential to the model)?

Best,

Michael


On Mon, Apr 2, 2012 at 8:41 PM, David Hoaglin <dchoaglin@gmail.com> wrote:
> Michael,
>
> Your (four) country dummies are constant within country, and so is
> each of your macro variables.  As a result, the country dummies,
> together with the constant term, can account perfectly for the
> variation in any one of the macro variables.
>
> You could run the logit model and plot the coefficients of the country
> dummies and the constant (i.e., the reference country) against each of
> the macro variables.  If you find a strong systematic relation for a
> particular macro variable, you could consider using it in your logit
> model instead of the country dummies.  You did not say how many macro
> variables you are considering, but you have only 5 countries, so those
> plots will have only 5 points.  Because of the sizable within-country
> sample sizes, you may be able to consider using more than one macro
> variable instead of the country dummies.  Of course, if you have 5
> macro variables, and they are not collinear, they would capture all of
> the variation between countries, without necessarily adjusting for
> real effects.
>
> The country dummies (with the constant) will account for all observed
> variation among the countries, whether related to the macro variables
> or not.
>
> I hope this discussion helps.
>
> David Hoaglin
>
>> I am running a binary-response logit model, using 2010 survey data
>> which includes roughly 2000 respondents from 5 different countries. I
>> have two questions:
>>
>> -I want to control for variations across the 5 countries and capture
>> country-specific effects. What is the best way to do this? Should I
>> simply include country dummies for 4 of the 5 countries?
>>
>> -I also have some 'macro' data (e.g. gdp per capita) which is unique
>> between the countries but of course does not vary within them. I
>> believe it's useful to include these variables. If I want to include
>> the country dummies as well, STATA throws out my macro data due to
>> collinearity. What is the best way to proceed?
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/