st: Best strategy: Multilevel model or other?

[Richard Moverare <richard.moverare@gmail.com>]

Dear statalist,

I have some struggles estimating a model, or rather I'm not sure I am
heading the right way. To some extent this is more a statistical
question than a Stata question, but any help would be greatly
appreciated.

I want to estimate election results for political party j in district
i and I want to see the effect of party j using a specific strategy in
district i. The variable referring to strategy is dichotomous (1/0).
The election results are by county and each county have several
districts. I also have set of control variables that for each district
refer to income level, educational level and so forth. These are
characteristics known to effect the election results of different
parties.

I want to control for the fact that in some counties some parties are
very small, and that could be the reason why they are not using the
strategy.

So far, I have calculated the difference in election results between
district i and the average over all districts in the county that
district i belongs to. And likewise for the other control variables.

But I assume that using a multilevel model might be a better strategy,
or? And if I use -xtmixed- with county as a random effects, should I
still use differences between districts and counties as y and x
variables, or is that not necessary when I have a multilevel model?
And what is the best way to set this up in Stata?

All the best,
Richard
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/